ML data that does not decrypt is recorded with its reason (closes #198)

`fetchMLDataBatch` dropped an entry that did not decrypt, gunzip or parse,
so the file looked like one with no ML data and the run exited 0. It now
returns the reason beside the payloads, also for ML data that is not a
JSON object. The library gives the reasons from its last ML data fetch
that succeeded in `status().mlDataErrors` and the `fetchMLData` "done"
event; a fetch that fails leaves them as they were. `quak backup` writes
the reason as `mlDataError` and records the file in `failures.json`;
`backup-metadata` writes `mlDataError` and exits 1.

Judgement call: a file whose ML data is cached keeps it in its JSON when a
newer copy does not decode, as when the fetch fails.

Model: opus-5-5

README.md

@@ -19,6 +19,10 @@ encrypting and uploading them back to the server.
 
 ## Getting Started
 
+The commands below need git, node 22 or later, and yarn. `make install`, which
+builds the single binary that "Running the backup from cron" uses, also needs
+[nix](https://nixos.org/download).
+
 ```bash
 git clone https://git.eeqj.de/sneak/quak.git
 cd quak
@@ -86,9 +90,12 @@ enumeration/download calls) is exported too and documented under Design below.
 ## Running the backup from cron
 
 `quak backup` never prompts, so cron can run it. `make install` builds quak as a
-single binary and copies it to `~/bin/quak`. Log in once with it, as the user
-the cron job will run as; the session is saved in that user's data directory
-(see "Session handling"):
+single binary and copies it to `~/bin/quak`. Besides what Getting Started needs,
+it needs [nix](https://nixos.org/download), even where bun is already installed:
+`make install` runs `make build-bin`, which bundles the built `dist/` with bun
+run through `nix-shell -p bun`. Log in once with `~/bin/quak`, as the user the
+cron job will run as; the session is saved in that user's data directory (see
+"Session handling"):
 
 ```bash
 ~/bin/quak login
@@ -127,19 +134,34 @@ matters.
 
 ### Exit codes
 
-| Code | Meaning                                                                                                                                                                           |
-| ---- | --------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- |
-| `0`  | The backup is complete: every original is at its save path, and no file failed.                                                                                                   |
-| `1`  | The run finished with files in `failures.json` or an album it could not decrypt, or it stopped on another error, printed as one line `quak: <message>`. The next run tries again. |
-| `2`  | Another backup of the same directory is running. This run sent no request and changed nothing.                                                                                    |
-| `3`  | There is no usable session: none is saved, the saved one is corrupt, or the server no longer accepts it. Run `quak login` as the user the cron job runs as.                       |
+| Code | Meaning                                                                                                                                                                                                        |
+| ---- | -------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- |
+| `0`  | The backup is complete: every original is at its save path, and no file failed.                                                                                                                                |
+| `1`  | The run finished with files in `failures.json` or albums it could not decrypt or write (see "Backup layout"), or it stopped on another error, printed as one line `quak: <message>`. The next run tries again. |
+| `2`  | Another backup of the same directory is running. This run sent no request and changed nothing.                                                                                                                 |
+| `3`  | There is no usable session: none is saved, the saved one is corrupt, or the server no longer accepts it. Run `quak login` as the user the cron job runs as.                                                    |
+
+`<message>` is the error's message, followed by the message of each error that
+caused it. Under Node, as `yarn quak` runs quak, `fetch` reports every failure
+below HTTP as just `fetch failed` and gives the reason as its cause, so a run
+that could not reach the server prints, for example,
+`quak: fetch failed: connect ECONNREFUSED 1.2.3.4:443`. When the server has
+several addresses and none of them could be reached, the line gives the reason
+for each, separated by `, `. A file whose download failed that way is in
+`failures.json` with that error, as a transient failure. So is each file with no
+ML data cached when the ML data fetch failed that way, its error starting
+`ML data: `. The `~/bin/quak` that `make install` builds runs under bun, and
+prints bun's own wording for such a failure, with no reason after it, such as
+`quak: Unable to connect. Is the computer able to access the url?` for a refused
+connection. It records such a failure in `failures.json` as transient too.
 
 After a `1`, the next run fetches each missing original again, fetches the ML
-data that is not cached, and rebuilds the album links. A file or album quak
-could not decrypt keeps the exit code at `1` until the server sends a copy that
-decrypts, or deletes it (see "Backup layout"). An original that a `--verify` run
-could not read, or put back with bytes that still do not match, stays at its
-save path. Only a later `--verify` run checks it again: a run without `--verify`
+data that is not cached, and rebuilds the album links. An album quak could not
+decrypt, or a file in `failures.json` because quak could not decrypt it, keeps
+the exit code at `1` until the server sends a copy that decrypts, or deletes it
+(see "Backup layout"). An original that a `--verify` run could not read, or
+could not replace with a download that matches, stays at its save path as it
+was. Only a later `--verify` run checks it again: a run without `--verify`
 leaves it as it is and takes the file out of `failures.json`, so that run can
 exit 0.
 
@@ -347,7 +369,8 @@ quak/
         exif.ts         EXIF read from an image's bytes with exifreader
         mldata-fetch.ts fetch + decrypt per-file ML data
         filename.ts     safe file names from server metadata
-        errors.ts       error types shared across layers
+        errors.ts       error types shared across layers, and how an error
+                        is described
         retry.ts        retry classifier + exponential backoff with jitter
         thumbnails.ts   detect + regenerate missing thumbnails
         client.ts       high-level Client class assembled from the above
@@ -474,6 +497,17 @@ request is repeated only when repeating it could produce a different answer:
   declining to retry a transient one costs a single file that the next run picks
   up.
 
+The `~/bin/quak` that `make install` builds runs under bun, whose `fetch`
+rejects with a plain `Error` that has no `cause` and carries a code of bun's
+own. quak retries a refused connection, a name that does not resolve, a reset
+connection and a deadline under bun as it does under Node. Bun reports a refused
+connection, and a TLS handshake that failed, as `ConnectionRefused`. For a name
+that does not resolve it gives `ConnectionRefused` on some requests and
+`FailedToOpenSocket` on others. A connection reset or closed before the response
+ended is `ECONNRESET`, as under Node, and a deadline is the same `TimeoutError`.
+A certificate bun rejects, reported with a code such as
+`DEPTH_ZERO_SELF_SIGNED_CERT`, is not retried.
+
 Backoff is exponential with full jitter: the delay before retry _n_ is
 `random() * min(maxDelayMs, baseDelayMs * 2 ** (n - 1))`. The exponential term
 is the ceiling and the wait is drawn below it, so a client that lost many
@@ -529,9 +563,13 @@ routing errnos `EHOSTUNREACH`, `ENETUNREACH` and `ENETDOWN` are excluded for the
 same reason, despite looking like connect-time failures: on Linux an ICMP
 unreachable arriving mid-flight, or a local interface going down after the
 request was written, delivers them on an already-established socket. They stay
-retryable for the idempotent calls. `putFile` is exempt: a presigned PUT stores
-one whole object at one key in one request, so replaying it has no partial state
-to damage.
+retryable for the idempotent calls. So do bun's `ConnectionRefused` and
+`FailedToOpenSocket`, on which a `POST` or `PUT` is not replayed either. When a
+connection bun kept open from an earlier request closes before any answer, bun
+sends the request again on a new connection, and if that one fails it gives one
+of these codes although the server may already have the request. `putFile` is
+exempt: a presigned PUT stores one whole object at one key in one request, so
+replaying it has no partial state to damage.
 
 A download is retried as a whole — request, stream consumption, and decryption —
 because a socket reset after the response headers have arrived surfaces in the
@@ -626,9 +664,9 @@ answer, so they report current account state rather than whatever the cache last
 held. If that round-trip fails, the command prints the error on one line and
 exits 1, or 3 when the server no longer accepts the saved session (see "Session
 handling"). An album or file quak cannot decrypt does not fail it; the command
-answers without it (see "Default reads vs. fresh reads"). `--cache-dir`
-overrides where the cache lives; without it each account gets its own directory
-under the per-user cache path.
+answers from what did decrypt (see "Default reads vs. fresh reads").
+`--cache-dir` overrides where the cache lives; without it each account gets its
+own directory under the per-user cache path.
 
 An ID given to `files --collection`, `get`, `get-thumb` or
 `helper fix-missing-thumbnails --file` is written in decimal digits, no larger
@@ -655,13 +693,15 @@ reads no tag from is recorded, base64, as `exifRaw`, with the reason in
 and `helper fix-missing-thumbnails` take `--json` for machine-readable output.
 
 `backup --verify` also hashes the originals already in the backup and replaces
-any that do not match the content hash Ente records; one it cannot replace goes
-into `failures.json` (see "Backup layout").
+any that do not match the content hash Ente records; one it cannot replace stays
+as it was and goes into `failures.json` (see "Backup layout").
 
 `backup-metadata` fetches ML data in requests of up to 200 files. When a request
 fails, the error is logged, each of its files is written with the reason in an
-`mlDataError` field instead of `mlData`, and the dump goes on. The exit code is
-non-zero if any ML data request failed.
+`mlDataError` field instead of `mlData`, and the dump goes on. A file whose ML
+data does not decrypt, gunzip or parse is logged and written the same way, with
+why in `mlDataError`. The exit code is non-zero if any file was written with
+`mlDataError`.
 
 `helper fix-missing-thumbnails` regenerates thumbnails for JPEG images only,
 because the bundled decoder (`jpeg-js`) decodes only JPEG. A non-JPEG image
@@ -726,8 +766,15 @@ A file's JSON holds Ente's ML data for it (its faces and its CLIP embedding) as
 for has no `mlData`. The backup waits for the library's ML data fetch to finish
 before it writes the JSON files. If that fetch fails, each file whose ML data is
 not in the cache gets the reason in `mlDataError` instead and counts as failed,
-and the next run fetches it again. The JSON files are rewritten on every run, so
-ML data that arrived since the last run appears.
+and the next run fetches it again. The same goes for a file with no ML data in
+the cache whose ML data the server sent but quak could not decrypt, gunzip or
+parse. Its `mlDataError` says why, and `failures.json` records it with an error
+starting `ML data: `. quak keeps that reason in memory only and caches nothing
+for the file, so the next run asks for its ML data again. An ML data fetch that
+fails, such as one the library's refresh starts while the backup writes the JSON
+files, leaves the reasons as the last fetch that succeeded left them. The JSON
+files are rewritten on every run, so ML data that arrived since the last run
+appears.
 
 A file's JSON holds, as `imageMetadata`, what `backup-metadata --exif` records
 from its original, or from a live photo's image: `format`, `width` and `height`
@@ -745,23 +792,25 @@ succeeds, or once it is no longer in the library or in the backup's scope. The
 library's `lib.backup({ includeThumbnails: true })` also writes
 `thumbnails/<fileID>.jpg` beside `collections/`; `quak backup` does not.
 
-A file the library could not decrypt (see "Default reads vs. fresh reads") goes
-into `failures.json` with an empty title, since its title is encrypted too, and
-an error that starts `could not be decrypted:` and gives the reason. A file in
-two albums is backed up as usual if its copy in the other album decrypted. An
-album the library could not decrypt has no name quak can read and no files it
-can list. A run without `--json` prints one line for it,
+A file in the backup's scope that the library could not decrypt (see "Default
+reads vs. fresh reads") goes into `failures.json` with an error that starts
+`could not be decrypted:` and gives the reason. If a copy of it in another album
+decrypted, it does not: its original is stored from that copy, and the album
+where it does not decrypt neither links nor lists it. An album the library could
+not decrypt has no name quak can read and no files it can list. A run without
+`--json` prints one line for it,
 `FAILED album <id>: could not be decrypted: <reason>`, and `--json` and the
-`BackupResult` list its ID in `undecryptableAlbums`. Every other file is backed
-up. Either kind keeps the exit code at 1, run after run, until the server sends
-a copy that decrypts or deletes it; the next run then drops it.
+`BackupResult` list its ID in `undecryptableAlbums`. Such a file in
+`failures.json`, or such an album, keeps the exit code at 1, run after run,
+until the server sends a copy that decrypts or deletes it; the next run then
+drops it.
 
 `backup.lock` keeps two backups of the same directory from running at once, such
 as a cron run that starts while the previous one is still going. A backup
 creates `<dir>` if it is missing and takes the lock before its refresh, and
-removes the lock when it ends, whether it succeeds or fails; `quak backup` takes
-it before it opens its library, so a refused run sends no request. The lock is a
-directory that
+removes the lock when it ends, whether it succeeds or fails, unless it lost the
+lock (see below); `quak backup` takes it before it opens its library, so a
+refused run sends no request. The lock is a directory that
 [proper-lockfile](https://github.com/moxystudio/node-proper-lockfile) creates
 and keeps touching while the backup runs. A second backup of the directory, from
 another process or the same one, fails at once: `quak backup` prints
@@ -773,13 +822,38 @@ over, so nobody has to remove it. The lock is outside the date folders and
 `collections/`, so it is never taken for an original, and the removal of old
 album directories never touches it.
 
+A running backup loses its lock when someone removes `backup.lock`, or when
+another backup takes it over because the running one left it untouched for 10
+seconds, as a stopped or stalled process does. The backup finds this out the
+next time it goes to touch the lock, which it does every 5 seconds, and then
+stops before it starts another file. The originals it is downloading at that
+moment are still stored. It leaves `failures.json` and `backup.lock` as they
+are, since another backup may be using them. `quak backup` prints
+`quak: lost the lock on <dir>: <reason>` as one line and exits 1, and
+`lib.backup()` rejects with `lost the lock on <dir>: <reason>`.
+
 A collection's directory and JSON are named after the collection, and a symlink
 after the file's title, both with unsafe characters replaced. When two
 collections would get the same name, or two symlinks in one collection the same
 name (ignoring case in both), each of them gets its ID added: two albums named
 `Trip` become `Trip (10)/` and `Trip (11)/`, and two files titled `IMG_0001.JPG`
-become `IMG_0001 (12345).JPG` and `IMG_0001 (12346).JPG`. IDs never change, so a
-name stays the same from run to run until such a clash appears or goes away.
+become `IMG_0001 (12345).JPG` and `IMG_0001 (12346).JPG`. A collection whose
+directory would have the name of another collection's JSON, again ignoring case,
+gets its ID added as well: beside an album `x`, whose JSON is `x.json`, an album
+named `x.json` gets the directory `x.json (12)/`. A name too long for a file
+name, which most file systems limit to 255 bytes, gets its ID added too. A name
+with its ID added is cut short between two characters in the part before the ID,
+just far enough that it fits. When a symlink's extension alone leaves no room
+for the ID, the extension is cut short and the part before the ID keeps only the
+bytes that cut leaves over. A collection's directory name also leaves room for
+the `.json` its JSON adds. IDs never change, so a name stays the same from run
+to run until such a clash appears or goes away.
+
+A collection whose directory or JSON cannot be written, such as one whose
+directory name is taken by a file you put in `collections/`, does not stop the
+run. A run without `--json` prints `FAILED album <name>: <reason>` for it,
+`--json` and the `BackupResult` list its ID in `failedAlbums`, and the exit code
+is 1. The other collections and `failures.json` are still written.
 
 A live photo, which Ente stores as one ZIP of its image and its video, is stored
 as those two files, which a photo viewer can open: each is
@@ -792,9 +866,12 @@ both, each named after the title with that file's extension (`IMG_0001.heic` and
 
 Each run removes the symlinks into the date folders that no longer belong in
 their collection's directory, and the directories (and JSON) of collections that
-were deleted or renamed. Nothing else in `collections/` is touched: a file or a
-symlink you put there stays, and a directory that still holds one after its
-symlinks are removed stays too, with its JSON.
+were deleted or renamed. An album quak can no longer decrypt loses its directory
+and JSON the same way, and a file whose copy in an album quak can no longer
+decrypt loses its symlink there and its entry in that album's JSON; the
+originals stay at their save paths. Nothing else in `collections/` is touched: a
+file or a symlink you put there stays, and a directory that still holds one
+after its symlinks are removed stays too, with its JSON.
 
 Each file is downloaded exactly once regardless of how many collections it
 appears in, and written once: straight to its save path, with no copy left in
@@ -803,40 +880,49 @@ subsequent runs, existing originals are skipped. If a download fails, the error
 is logged and the backup continues with the next file. The exit code is non-zero
 if any files failed. `quak backup` opens its library with the thumbnail and
 originals precache off, so the only file content it fetches is the originals the
-backup stores.
+backup stores. It also opens it with no background timer, so it asks the server
+for changes only as it starts, not again while it downloads.
 
 With `--verify`, or `lib.backup({ verify: true })`, a run also hashes each
 original already at its save path the way a download is checked (see "On-disk
 cache layout" below): its bytes, read in chunks, or a live photo's image and
 video, joined as `<imageHash>:<videoHash>`. An original that matches the content
 hash its metadata records is left as it is. One that does not is logged on one
-line naming the file, deleted (a live photo's image and video both), and put
-back in the same run like a missing one: downloaded, or copied from the cache if
-the cache holds it. What is put back is hashed too, because a copy from the
-cache is not checked as a download is. If it still does not match, it stays at
-its save path and the file goes into `failures.json`, as it does when the
-download fails. A file whose metadata records no hash is left as it is and
-counted as unchecked. A stored original that cannot be read is left as it is and
-counts as failed. The summary and `--json` add the counts `verified`,
-`mismatched` and `unchecked`, all of originals that were already stored; one
-first downloaded in this run is in none of them. A mismatch that was put back
-with matching bytes does not make the exit code non-zero. Without `--verify`
-nothing is hashed, the summary is unchanged, and the three counts are 0 in
-`--json`.
+line naming the file, and replaced once a matching copy is downloaded, in the
+same run. The copy is downloaded even when the cache holds the original, because
+a copy from the cache is not checked as a download is. It is written to a
+temporary file beside the original, a live photo's image and video each to its
+own, hashed the same way, and renamed over the original only when it matches; a
+live photo's image and video are both hashed before either is renamed. Until
+then the original stays as it was: when the download fails, or what it brings
+does not match either, the temporary files are deleted, the original keeps its
+bytes and its album links, and the file goes into `failures.json`. A run that
+stops between renaming a live photo's new image and its new video leaves the new
+image beside the old video, which a later `--verify` run replaces if the video
+was the part that did not match. A file whose metadata records no hash is left
+as it is and counted as unchecked. A stored original that cannot be read is left
+as it is and counts as failed. The summary and `--json` add the counts
+`verified`, `mismatched` and `unchecked`, all of originals that were already
+stored; one first downloaded in this run is in none of them. A mismatch that was
+replaced with matching bytes does not make the exit code non-zero. Without
+`--verify` nothing is hashed, the summary is unchanged, and the three counts are
+0 in `--json`.
 
 Each original is written to a temporary file in the same directory, synced to
 disk, and renamed into place, so an original is either complete or absent, even
 after a power cut. A downloaded original's temporary file is named
 `.quak-<pid>-<random>.tmp`, one copied from the cache
-`.quak-backup-YYYY-MM-DD.<fileID>.<ext>-<pid>-<random>.tmp`. A run that is
-killed can leave one of these temporary files behind; the next backup deletes
-those whose process is no longer running. The content cache uses the same
-scheme, and opening a library deletes the temporary files in the cache whose
-process is no longer running, so a download another process has in progress in
-the same cache is left alone. The rename replaces whatever was at the
-destination rather than writing through it: a symlink there is replaced, not
-followed, and the new file has the temporary file's permissions, not those of
-the file it replaced.
+`.quak-backup-YYYY-MM-DD.<fileID>.<ext>-<pid>-<random>.tmp`. A copy that
+`--verify` downloads to replace an original is written to such a
+`.quak-<pid>-<random>.tmp` too, a live photo's image and video to
+`.quak-<pid>-<random>.<ext>`, each with its own extension. A run that is killed
+can leave one of these temporary files behind; the next backup deletes those
+whose process is no longer running. The content cache uses the same scheme, and
+opening a library deletes the temporary files in the cache whose process is no
+longer running, so a download another process has in progress in the same cache
+is left alone. The rename replaces whatever was at the destination rather than
+writing through it: a symlink there is replaced, not followed, and the new file
+has the temporary file's permissions, not those of the file it replaced.
 
 ## TODO
 
@@ -884,7 +970,7 @@ background, so an unreachable server does not block opening.
 | `client`                 | required                                           | the account client (a `Client`, or any `LibraryClient`)              |
 | `cacheDirectory`         | `quak/<userID>` under the per-user cache directory | where `metadata.json` and the content cache live                     |
 | `downloadDirectory`      | `photos` in the working directory                  | root of the save paths; an original stored there counts as cached    |
-| `refreshIntervalSeconds` | `3`                                                | background refresh cadence                                           |
+| `refreshIntervalSeconds` | `3`                                                | background refresh cadence; `Infinity` for no background timer       |
 | `precacheThumbnails`     | `true`                                             | prefetch every thumbnail, newest first                               |
 | `precacheOriginals`      | `true`                                             | prefetch the favorites album and the latest-window originals         |
 | `precacheOriginalsDays`  | `7`                                                | length in days of that latest window                                 |
@@ -910,6 +996,14 @@ resolves once that refresh (including its cache write), the ML data fetch and
 the precache fetches already running have all finished, so the cache directory
 can then be removed.
 
+`lib.status().mlDataErrors` lists, as `{ fileID, error }`, each file whose ML
+data the last ML data fetch that succeeded got but could not decrypt, gunzip or
+parse; the `fetchMLData` event with status `"done"` carries the same list. A
+fetch that is still running leaves the list as it is until it ends, and a fetch
+that fails leaves it as it was. Such ML data is not cached, so the next ML data
+fetch asks for it again. The reasons are kept in memory only, never in the
+cache.
+
 ### Default reads vs. fresh reads
 
 Default reads — `lib.albums`, `lib.photos`, `lib.timeline` — answer
@@ -1054,21 +1148,25 @@ photos newest first). `lib.subscribe({ onChange })` delivers a `LibraryChange`
   directory, and fails at once with an error whose `code` is `ELOCKED` while
   another backup of it runs. A caller can instead take the lock itself, before
   it opens its library, as `quak backup` does: `await lockBackupDirectory(dir)`
-  takes it, failing the same way, and returns the function that releases it, and
-  the caller passes `lockHeld: true` to the backup. It waits for a refresh as
-  `fresh()` does, puts every in-scope original not already at its save path
-  there as `photo.download()` does (and, with `includeThumbnails`, fetches
-  thumbnails) through the content cache, waits for an ML data fetch, and
-  rebuilds the on-disk backup tree with a durable failure ledger. Each file's
-  JSON holds its ML data and, for an image (for a live photo, its image), its
-  original's EXIF and XMP, with its dimensions for a JPEG only; a video's JSON
-  holds none of these three. A fetched original is written straight to its save
-  path and not into the cache, which then counts it as present; one the cache
-  already held is copied from there. `BackupOptions`: `downloadDirectory` (falls
-  back to the library's), `includeOriginals` (default `true`),
-  `includeThumbnails` (default `false`), `onlyAlbumNames`, `verify` (default
-  `false`), `onProgress`, and `lockHeld` (default `false`). See Backup layout
-  above for the tree it writes.
+  takes it, failing the same way, and returns a `BackupLock`, whose `release()`
+  removes the lock unless it was lost and whose `lost` holds why it was lost.
+  The caller passes it to the backup as `lock`, and releases it once the backup
+  ends. It waits for a refresh as `fresh()` does, puts every in-scope original
+  not already at its save path there as `photo.download()` does (and, with
+  `includeThumbnails`, fetches thumbnails) through the content cache, waits for
+  an ML data fetch, and rebuilds the on-disk backup tree with a durable failure
+  ledger. Each file's JSON holds its ML data and, for an image (for a live
+  photo, its image), its original's EXIF and XMP, with its dimensions for a JPEG
+  only; a video's JSON holds none of these three. A fetched original is written
+  straight to its save path and not into the cache, which then counts it as
+  present; one the cache already held is copied from there. `BackupOptions`:
+  `downloadDirectory` (falls back to the library's), `includeOriginals` (default
+  `true`), `includeThumbnails` (default `false`), `onlyAlbumNames`, `verify`
+  (default `false`), `onProgress`, and `lock`. See Backup layout above for the
+  tree it writes, and for the albums `BackupResult` lists in
+  `undecryptableAlbums` and `failedAlbums`. Whether the backup took the lock or
+  its caller did, once the lock is lost the backup rejects with
+  `lost the lock on <dir>: <reason>` before it starts another file.
 
 ### Request pools
 
@@ -1077,6 +1175,11 @@ photos newest first). `lib.subscribe({ onChange })` delivers a `LibraryChange`
 Each pool orders on-demand work ahead of background/precache work and dedups
 in-flight fetches by key, and an idle pool never lends its slots to a busy one.
 
+`lib.backup()`, and so `quak backup`, works on up to the content pool's limit of
+originals at once, 5 unless the library was opened with other `pools`. The limit
+covers everything it does for one original, hashing it for `verify` included, so
+however large the account is, no more originals than that are open at once.
+
 ### On-disk cache layout
 
 Under `cacheDirectory`:
@@ -1128,7 +1231,7 @@ from a very old client, is stored unchecked.
 - `src/library/mlsearch.ts`: `MLDataAPI`, `SimilarResult`
 - `src/exif.ts`: `ExifTags`, `PhotoExif`
 - `src/library/pools.ts`: `RequestPools`, `RequestPoolsOptions`, `BoundedPool`
-- `src/backup.ts`: `BackupOptions`, `BackupResult`, `BackupError`
+- `src/backup.ts`: `BackupLock`, `BackupOptions`, `BackupResult`, `BackupError`
 - `src/client.ts`: `Client`, `LoginOptions`, `ClientSnapshot`
 - `src/api/client.ts`: `ApiClient`, `ApiClientOptions`, `StreamOptions`
 - `src/errors.ts`: `ApiError`, `TruncatedStreamError`

TODO.md

@@ -17,24 +17,120 @@ pre-1.0
 
 # Next Step
 
-None: no implementation work is open. The cache design,
-https://git.eeqj.de/sneak/quak/issues/36, waits on sneak's review.
+One bug is open: https://git.eeqj.de/sneak/quak/issues/205. The check of quak
+against a live Ente account, https://git.eeqj.de/sneak/quak/issues/187, needs an
+account only sneak has, so it waits on him.
 
 Tagging and releases are decided by sneak alone, and happen only when he
 declares one.
 
 # Completed Steps
 
-- 2026-10-13: An album or file quak cannot decrypt, or whose metadata is not
+- 2026-10-16: ML data the server sends that does not decrypt, gunzip or parse is
+  no longer taken for no ML data (issue 198). `fetchMLDataBatch` returns why for
+  each such file beside the payloads; ML data that parses to something other
+  than a JSON object, such as `null`, is one of them. The library keeps the
+  reasons from its last ML data fetch that succeeded in memory, not in the
+  cache, and gives them in `status().mlDataErrors` and in the `fetchMLData`
+  event with status `"done"`; a fetch still running or one that fails does not
+  change them. `quak backup` writes the reason as `mlDataError` in the file's
+  JSON when no ML data is cached for it, and records the file in `failures.json`
+  with an error starting `ML data: `; `backup-metadata` writes `mlDataError`
+  too. Both exit 1. Tests serve ML data encrypted under a key that is not the
+  file's.
+
+- 2026-10-16: With `--verify`, an original whose hash does not match is no
+  longer removed before its replacement is downloaded (issue 199).
+  `replaceOriginal` in `src/library/content.ts` downloads the replacement to a
+  temporary file beside it, a live photo's image and video each to its own,
+  hashes it and renames it over the original only when it matches. When the
+  download fails or does not match either, the original keeps its bytes and
+  links, and the file goes into `failures.json`. The replacement is always
+  downloaded, through the new `fetchOriginal` of the content cache, never copied
+  from the cache.
+
+- 2026-10-16: The `~/bin/quak` that `make install` builds retries a refused
+  connection and a name that does not resolve, and the backup records either as
+  transient (issue 206). Bun's `fetch` reports them with codes of its own,
+  `ConnectionRefused` and `FailedToOpenSocket`, which `src/retry.ts` now counts
+  as transport failures. A `POST` or `PUT` is not replayed on them, because bun
+  also gives them when it resends a request on a new connection after a
+  connection it kept open closed, and the server may already have the request.
+  Bun already gave `ECONNRESET` for a reset and a `TimeoutError` for a deadline.
+  `classify` in `src/backup.ts` now records any `TimeoutError` as transient, so
+  a stalled download, whose error says only that it stalled, is no longer
+  `unknown`. The shapes were found by running the binary against local
+  listeners.
+
+- 2026-10-15: An album whose folder or JSON `quak backup` cannot write no longer
+  ends the run before the other albums and `failures.json` are written (issue
+  197). The run prints `FAILED album <name>: <reason>`, lists the album's ID in
+  `failedAlbums` in the result and in `--json`, goes on, and exits 1. An album
+  whose folder would have the name of another album's JSON, as `x.json` beside
+  `x` would, gets its ID added. A folder or link name over 255 bytes is cut
+  short between two characters, by `shortenToBytes` in `src/filename.ts`, and
+  gets the ID.
+
+- 2026-10-15: A network failure is printed and recorded with its reason, not as
+  just `fetch failed` (issue 196). `describeError` in `src/errors.ts` gives an
+  error's message followed by the message of each error on its `cause` chain,
+  such as `fetch failed: connect ECONNREFUSED 1.2.3.4:443`, and each address's
+  reason when a host has several and none could be reached. `run` prints that,
+  and the backup logs it and records it in `failures.json`, where `classify` now
+  calls the failure transient when the chain holds an errno that
+  `hasTransportCode` in `src/retry.ts` recognises, for a failed ML data fetch
+  too. The walk along the chain moved from `src/retry.ts` to `src/errors.ts`,
+  which both use. Tests refuse every connection of a backup, then only one
+  file's download, and fail the ML data fetch.
+
+- 2026-10-15: The README says that `make install` needs nix (issue 200), in
+  Getting Started and in "Running the backup from cron": `make build-bin` runs
+  bun through `nix-shell -p bun`. The `Makefile` is unchanged.
+
+- 2026-10-14: `lib.backup()`, and so `quak backup`, works on up to the content
+  pool's limit of originals at once, 5 by default, instead of one at a time
+  (issue 182). That many workers take files from one shared list, each doing all
+  the work for one original, its `--verify` hashing included, so no more
+  originals than that are open at once. `BackupLibrary` says the limit as
+  `contentConcurrency`. A test holds each download at the fake Ente server to
+  count those in flight. Once a worker has thrown, as one does when `onProgress`
+  throws, or the lock is lost, no worker takes another file; the backup fails
+  once the originals under way are in place, and a lost lock stops it before it
+  fetches ML data.
+
+- 2026-10-14: Next Step said the cache design, issue 36, waits on sneak's
+  review, though that issue closed on 2026-10-03 (issue 194). It now names the
+  issues open on the tracker.
+
+- 2026-10-14: A backup whose lock is lost no longer dies with a stack trace
+  (issue 184). `lockBackupDirectory` gives proper-lockfile an `onCompromised`
+  that records why the lock was lost instead of throwing from its timer, and
+  returns the lock, which `quak backup` passes to `lib.backup()` as `lock`. The
+  backup then stops before it starts another file with
+  `lost the lock on DIR: REASON`, which `quak backup` prints as one line before
+  it exits 1. Releasing a lost lock leaves `backup.lock` alone. Tests in
+  `test/cli/backup-end-to-end.test.ts` remove the lock, or put another backup's
+  in its place, while a download is held open.
+
+- 2026-10-14: An album or file quak cannot decrypt, or whose metadata is not
   JSON, no longer fails the refresh, so one bad record no longer stops
   `quak backup` from storing anything (issue 181). The refresh leaves it out, an
   album with its files, and the library keeps it with the reason in
-  `metadata.json` until the server sends a copy that decrypts or deletes it. The
-  backup stores every other file, records each such file in `failures.json` with
-  an error starting `could not be decrypted:`, prints one line naming each such
-  album's ID, lists those IDs in `undecryptableAlbums`, and exits 1 while any
-  remain. A file whose copy in another album decrypts is backed up as usual. The
-  read commands answer from what decrypted.
+  `metadata.json` until the server sends a copy that decrypts or deletes it.
+  `quak backup` stores every other file, records each such file in
+  `failures.json` with an error starting `could not be decrypted:`, names each
+  such album's ID in a printed line, or with `--json` in `undecryptableAlbums`,
+  and exits 1 while any remain. A file whose copy in another album decrypts is
+  not recorded: its original is stored from that copy, and the album where it
+  does not decrypt neither links nor lists it. The read commands answer from
+  what decrypted.
+
+- 2026-10-14: `quak backup` no longer asks the server for changes every 3
+  seconds while it runs (issue 183). It opens its library with
+  `refreshIntervalSeconds: Infinity`, which the library now takes as no
+  background timer, so the backup refreshes only as it starts and no timed
+  refresh starts another ML data pass during the run. A test holds a backup's
+  download open while two hours pass and counts its `/collections/v2` requests.
 
 - 2026-10-14: `test/cli/backup-end-to-end.test.ts` runs `quak backup` as cron
   does (issue 186): `backupCommand` and `run`, the real `loadSession` reading a

src/backup.ts

@@ -1,13 +1,14 @@
 // The backup command, rebuilt on the library API (issue #51).
 //
 // `lib.backup()` takes the lock in `downloadDirectory` (unless its caller
-// already holds it), and fails at once when another backup of it holds the
-// lock. It waits for a completed refresh of the library (a failed one fails the
+// passes one in), and fails at once when another backup of it holds the
+// lock; when the lock is lost while it runs, it fails before it starts another
+// file. It waits for a completed refresh of the library (a failed one fails the
 // backup before any file is touched), then, for every file in scope, puts its
 // original at its save path under `downloadDirectory`, as `Photo.download()`
-// does, waits for an ML data fetch, and rebuilds the derived views (per-file
-// sidecars, per-collection symlink trees, per-collection JSON) from the model.
-// The on-disk layout:
+// does, up to the content pool's limit of them at once, waits for an ML data
+// fetch, and rebuilds the derived views (per-file sidecars, per-collection
+// symlink trees, per-collection JSON) from the model. The on-disk layout:
 //
 //     <downloadDirectory>/
 //         YYYY/YYYY-MM/YYYY-MM-DD/
@@ -40,8 +41,9 @@
 //
 // With `verify`, each original already at its save path is hashed as the
 // download check hashes it, and one that does not match the content hash its
-// metadata records is removed and fetched again in the same run. What is put
-// back is hashed too, and recorded as failed if it still does not match.
+// metadata records is replaced once a matching copy is downloaded, in the same
+// run. Until then it stays as it is: a download that fails, or that does not
+// match either, leaves it in place and is recorded as failed.
 //
 // Resilience (issue #8): no per-file condition aborts the run. A failed
 // download, a failed symlink, or ML data missing because the ML data fetch
@@ -49,7 +51,9 @@
 // running attempt count, and the last-tried time, and the run continues.
 // A file the library could not decrypt is recorded there too, unless a copy of
 // it in another album decrypted; an album it could not decrypt is logged by ID
-// and listed in `result.undecryptableAlbums`, as its name is not known.
+// and listed in `result.undecryptableAlbums`, as its name is not known. An
+// album whose folder or JSON cannot be written is logged by name and
+// listed in `result.failedAlbums`, and the run goes on to the next album.
 // `result.failed` — and thus the CLI's exit code — stays non-zero while any
 // failure remains unresolved and clears once every one succeeds. Each run
 // reconciles the ledger against the files it attempted, so an entry for a file
@@ -58,7 +62,6 @@
 // code.
 
 import {
-    createReadStream,
     lstatSync,
     mkdirSync,
     readdirSync,
@@ -74,28 +77,39 @@ import { readFile } from "node:fs/promises";
 import { dirname, extname, join, relative, resolve } from "node:path";
 import lockfile from "proper-lockfile";
 
-import {
-    chunkHashFinal,
-    chunkHashInit,
-    chunkHashUpdate,
-    init,
-} from "./crypto/index.js";
 import { removeLeftoverTempFiles } from "./download/index.js";
-import { sanitizeFileName, withExtension } from "./filename.js";
+import { describeError } from "./errors.js";
+import {
+    MAX_NAME_BYTES,
+    sanitizeFileName,
+    shortenToBytes,
+    withExtension,
+} from "./filename.js";
 import {
     copyAtomic,
     placeOriginal,
+    replaceOriginal,
     savePath,
     storedAtSavePath,
+    storedHash,
 } from "./library/content.js";
 import { representative } from "./library/records.js";
 import type { Undecryptable } from "./library/store.js";
 import { extractImageMetadata } from "./metadata-backup.js";
 import type { MLData } from "./mldata-fetch.js";
 import type { Collection, EnteFile, FileMetadata } from "./model/types.js";
+import { hasTransportCode } from "./retry.js";
 
 export type ProgressCallback = (message: string) => void;
 
+// The lock in a backup directory, as `lockBackupDirectory` returns it.
+export interface BackupLock {
+    // Removes `backup.lock`, unless the lock was lost.
+    release(): Promise<void>;
+    // Why the lock was lost, once it is.
+    lost?: Error;
+}
+
 export interface BackupOptions {
     // Where the backup tree lives. `lib.backup()` defaults it to the library's
     // download directory; `runBackup` with none throws before any network
@@ -112,10 +126,11 @@ export interface BackupOptions {
     // bytes do not match the content hash its metadata records. Default false.
     verify?: boolean;
     onProgress?: ProgressCallback;
-    // The caller already holds the lock in `downloadDirectory`, taken with
-    // `lockBackupDirectory`, and releases it itself, so the backup does not
-    // take it. `quak backup` takes it before it opens its library.
-    lockHeld?: boolean;
+    // The lock in `downloadDirectory`, when the caller took it with
+    // `lockBackupDirectory` and releases it itself. The backup then does not
+    // take the lock, and fails once this one is lost. `quak backup` takes it
+    // before it opens its library.
+    lock?: BackupLock;
 }
 
 export interface BackupError {
@@ -133,9 +148,9 @@ export interface BackupResult {
     // Originals already at their save path and left untouched.
     skipped: number;
     // With `verify`, the originals already at their save path whose hash
-    // matched, those whose hash did not (each removed and fetched again), and
-    // those whose metadata records no hash (left as they are). All three are
-    // zero without `verify`.
+    // matched, those whose hash did not (each replaced once a matching copy is
+    // downloaded), and those whose metadata records no hash (left as they
+    // are). All three are zero without `verify`.
     verified: number;
     mismatched: number;
     unchecked: number;
@@ -148,6 +163,9 @@ export interface BackupResult {
     // The IDs of the albums the library could not decrypt, whose files it
     // could not list; the CLI exits non-zero while there are any.
     undecryptableAlbums: number[];
+    // The IDs of the albums whose folder or JSON this run could not write;
+    // the CLI exits non-zero while there are any.
+    failedAlbums: number[];
 }
 
 // The slice of the library that backup drives. `Library` implements it; a test
@@ -170,7 +188,19 @@ export interface BackupLibrary {
         fileID: number,
         destination: string,
     ): Promise<{ path: string; videoPath?: string }>;
+    // Fetch an original's bytes to `destination` through the content pool,
+    // a live photo as its image and its video beside it, even when the cache
+    // or the library's save path already holds it, and return where they
+    // landed. With `verify`, an original whose hash does not match is replaced
+    // with what this fetches.
+    fetchOriginal(
+        fileID: number,
+        destination: string,
+    ): Promise<{ path: string; videoPath?: string }>;
     thumbnail(fileID: number): Promise<{ path: string }>;
+    // How many originals the backup works on at once: the limit of the
+    // library's content pool.
+    contentConcurrency: number;
     // Wait for an ML data fetch to complete, joining one already running or
     // starting one. Rejects with the reason when it fails; resolves at once
     // when the library cannot fetch ML data.
@@ -178,6 +208,11 @@ export interface BackupLibrary {
     // A file's cached ML data, as `lib.mldata.forFile()` returns it, or
     // undefined when none is cached.
     mlData(fileID: number): Promise<MLData | undefined>;
+    // Why the ML data the last ML data fetch that succeeded got for a file did
+    // not decrypt, gunzip or parse, or undefined when it did or the fetch got
+    // none. A fetch still running does not change it. A fetch that fails
+    // leaves it as it was.
+    mlDataError(fileID: number): unknown;
 }
 
 type FailureClass = "transient" | "permanent" | "unknown";
@@ -206,11 +241,16 @@ const isPresent = (path: string): boolean => {
 
 // Best-effort classification for the ledger. Retryable server/network problems
 // are transient; refusals and local filesystem/decrypt errors are permanent;
-// anything else is unknown. Both the error code and message are inspected.
+// anything else is unknown. The error code and the error as `describeError`
+// gives it are inspected, and so is each errno on the error's cause chain,
+// where Node's `fetch` puts it. An error named `TimeoutError` is transient
+// whatever its message says, as `isRetryable` retries it: each of quak's
+// deadlines aborts with one, and the download's says only that it stalled.
 const classify = (err: unknown): FailureClass => {
+    if (hasTransportCode(err)) return "transient";
     const e = err as NodeJS.ErrnoException;
-    const text =
-        `${e?.code ?? ""} ${err instanceof Error ? err.message : String(err)}`.toLowerCase();
+    if (e?.name === "TimeoutError") return "transient";
+    const text = `${e?.code ?? ""} ${describeError(err)}`.toLowerCase();
     if (
         /timeout|timed out|econnreset|econnrefused|econnaborted|network|socket|eai_again|throttl|temporarily|429|500|502|503|504/.test(
             text,
@@ -228,9 +268,6 @@ const classify = (err: unknown): FailureClass => {
     return "unknown";
 };
 
-const errorMessage = (err: unknown): string =>
-    err instanceof Error ? err.message : String(err);
-
 // Ensure `linkPath` is a symlink to `target`, rebuilding a missing, wrong, or
 // non-symlink entry. Throws on failure (a directory in the way, no permission)
 // so the caller records it and moves on rather than aborting the run.
@@ -250,31 +287,47 @@ const rebuildSymlink = (linkPath: string, target: string): void => {
 // The on-disk names for the entries of one directory, in entry order. Each
 // name is used as is unless another entry would get the same name, ignoring
 // case (two names that differ only in case are one entry on a case-insensitive
-// file system); then every entry sharing it gets ` (<id>)`, before the
-// extension when `beforeExtension` is set. A name with an ID added can match
-// another entry's own name (`IMG (6).JPG`), so this repeats until no name is
-// shared. IDs are stable, so the names are too.
+// file system), or it is too long for a file name; then it gets ` (<id>)`,
+// right before the extension when `beforeExtension` is set. To fit, the part
+// before the ID is cut short as far as it must be. When the extension alone
+// leaves no room for the ID, the extension is cut short and the part before
+// the ID keeps only the bytes that cut leaves over. With `withJSON`, each
+// entry also has the JSON file `<name>.json` beside it, as an album does: a
+// name then leaves room for `.json`, and one that is another entry's JSON name
+// gets its ID added too.
+// Two names with different IDs added never match, but one can match another
+// entry's own name (`IMG (6).JPG`), so this repeats until no name is shared.
+// IDs are stable, so the names are too.
 const uniqueNames = (
     entries: { id: number; name: string }[],
-    beforeExtension: boolean,
+    opts: { beforeExtension?: boolean; withJSON?: boolean },
 ): string[] => {
+    const maxBytes = MAX_NAME_BYTES - (opts.withJSON ? ".json".length : 0);
     const withID = (id: number, name: string): string => {
-        const ext = beforeExtension ? extname(name) : "";
-        const stem = name.slice(0, name.length - ext.length);
-        return `${stem} (${id})${ext}`;
+        const tag = ` (${id})`;
+        const fullExt = opts.beforeExtension ? extname(name) : "";
+        const stem = name.slice(0, name.length - fullExt.length);
+        const ext = shortenToBytes(fullExt, maxBytes - Buffer.byteLength(tag));
+        const room = maxBytes - Buffer.byteLength(tag + ext);
+        return shortenToBytes(stem, room) + tag + ext;
     };
     const names = entries.map((e) => e.name);
     const suffixed = new Set<number>();
     for (;;) {
         const counts = new Map<string, number>();
-        for (const name of names) {
+        const count = (name: string): void => {
             const key = name.toLowerCase();
             counts.set(key, (counts.get(key) ?? 0) + 1);
+        };
+        for (const name of names) {
+            count(name);
+            if (opts.withJSON) count(`${name}.json`);
         }
         let changed = false;
         for (const [i, { id, name }] of entries.entries()) {
             if (suffixed.has(i)) continue;
-            if (counts.get(name.toLowerCase()) === 1) continue;
+            const fits = Buffer.byteLength(name) <= maxBytes;
+            if (fits && counts.get(name.toLowerCase()) === 1) continue;
             names[i] = withID(id, name);
             suffixed.add(i);
             changed = true;
@@ -401,26 +454,6 @@ const saveLedger = (path: string, ledger: Map<number, FailureEntry>): void => {
     );
 };
 
-// The content hash of the original stored at `stored`, computed as the download
-// check computes it: over each file's bytes, read in chunks, and for a live
-// photo `<imageHash>:<videoHash>`.
-const storedHash = async (stored: {
-    path: string;
-    videoPath?: string;
-}): Promise<string> => {
-    await init();
-    const hashFile = async (path: string): Promise<string> => {
-        const state = chunkHashInit();
-        for await (const chunk of createReadStream(path)) {
-            chunkHashUpdate(state, chunk as Buffer);
-        }
-        return chunkHashFinal(state);
-    };
-    const hash = await hashFile(stored.path);
-    if (stored.videoPath === undefined) return hash;
-    return `${hash}:${await hashFile(stored.videoPath)}`;
-};
-
 // A file's EXIF, XMP and dimensions as its JSON holds them: what
 // `extractImageMetadata` found in its original, or why the original could not
 // be read.
@@ -457,7 +490,7 @@ const imageMetadataFor = async (
         const bytes = await readFile(originalPath);
         return { imageMetadata: extractImageMetadata(bytes) ?? {} };
     } catch (err) {
-        return { imageMetadataError: errorMessage(err) };
+        return { imageMetadataError: describeError(err) };
     }
 };
 
@@ -510,11 +543,14 @@ const writeAlbumJSON = (
     writeFileSync(path, JSON.stringify(album, null, 2));
 };
 
-// The backup itself, which `runBackup` below runs while the lock is held.
+// The backup itself, which `runBackup` below runs while `lock` is held. Once
+// the lock is lost, another backup may be writing to `downloadDirectory`, so
+// this one fails before it starts another file.
 const runLockedBackup = async (
     lib: BackupLibrary,
     opts: BackupOptions,
     downloadDirectory: string,
+    lock: BackupLock,
 ): Promise<BackupResult> => {
     const includeOriginals = opts.includeOriginals ?? true;
     const includeThumbnails = opts.includeThumbnails ?? false;
@@ -524,6 +560,7 @@ const runLockedBackup = async (
 
     log("Refreshing library...");
     await lib.refresh();
+    if (lock.lost) throw lock.lost;
 
     const collectionsDir = join(downloadDirectory, "collections");
     const thumbnailsDir = join(downloadDirectory, "thumbnails");
@@ -593,7 +630,7 @@ const runLockedBackup = async (
         // and thumbnail both fail this run must not double its attempt count
         // or appear twice in errors.
         if (failedThisRun.has(fileID)) return;
-        const error = errorMessage(err);
+        const error = describeError(err);
         errors.push({ fileID, title, collection, error });
         const prior = ledger.get(fileID);
         ledger.set(fileID, {
@@ -638,76 +675,30 @@ const runLockedBackup = async (
     // through the content cache/pools, as `Photo.download()` does, and fetch
     // the optional thumbnails; a present file is left as is. With `verify`, a
     // present original is hashed first, and one that does not match the hash
-    // its metadata records is removed and fetched like a missing one. One
-    // that cannot be read is recorded as failed and left where it is.
-    if (includeOriginals) {
-        for (const [fileID, file] of distinct) {
-            let stored = storedAtSavePath(downloadDirectory, file);
-            let mismatch = false;
-            if (stored !== undefined && verify) {
-                try {
-                    if (file.metadata.hash === undefined) {
-                        unchecked++;
-                    } else if (
-                        (await storedHash(stored)) === file.metadata.hash
-                    ) {
-                        verified++;
-                    } else {
-                        log(
-                            `MISMATCH original ${file.metadata.title} (${fileID}): its bytes do not match its content hash`,
-                        );
-                        mismatched++;
-                        mismatch = true;
-                        rmSync(stored.path);
-                        if (stored.videoPath !== undefined) {
-                            rmSync(stored.videoPath);
-                        }
-                        stored = undefined;
-                    }
-                } catch (err) {
-                    log(
-                        `FAILED verifying original ${file.metadata.title}: ${errorMessage(err)}`,
-                    );
-                    recordFailure(
-                        file.id,
-                        file.metadata.title,
-                        collectionName.get(file.collectionID) ?? "",
-                        err,
-                    );
-                    continue;
-                }
-            }
-            if (stored !== undefined) {
-                skipped++;
-                continue;
-            }
+    // its metadata records is downloaded again and replaced by
+    // `replaceOriginal`, which leaves it as it was when that download fails
+    // or does not match either. One that cannot be read is recorded as failed
+    // and left where it is. `storeOriginal` does this for one original, and
+    // records its failure.
+    const storeOriginal = async (file: EnteFile): Promise<void> => {
+        const stored = storedAtSavePath(downloadDirectory, file);
+        let mismatch = false;
+        if (stored !== undefined && verify) {
             try {
-                log(`Fetching original ${file.metadata.title} (${fileID})...`);
-                // A fetched original is written straight to its save path (a
-                // live photo beside it); only one that was already cached
-                // elsewhere is copied.
-                const placed = await placeOriginal(
-                    downloadDirectory,
-                    file,
-                    (dest) => lib.original(fileID, dest),
-                );
-                storedThisRun.add(fileID);
-                // A copy from the cache is not checked as a download is and can
-                // hold the same bad bytes, so what is put back after a
-                // mismatch is hashed too. A bad copy stays where it is and the
-                // file fails.
-                if (
-                    mismatch &&
-                    (await storedHash(placed)) !== file.metadata.hash
-                ) {
-                    throw new Error(
-                        "the original put back does not match its content hash either",
+                if (file.metadata.hash === undefined) {
+                    unchecked++;
+                } else if ((await storedHash(stored)) === file.metadata.hash) {
+                    verified++;
+                } else {
+                    log(
+                        `MISMATCH original ${file.metadata.title} (${file.id}): its bytes do not match its content hash`,
                     );
+                    mismatched++;
+                    mismatch = true;
                 }
-                downloaded++;
             } catch (err) {
                 log(
-                    `FAILED original ${file.metadata.title}: ${errorMessage(err)}`,
+                    `FAILED verifying original ${file.metadata.title}: ${describeError(err)}`,
                 );
                 recordFailure(
                     file.id,
@@ -715,12 +706,77 @@ const runLockedBackup = async (
                     collectionName.get(file.collectionID) ?? "",
                     err,
                 );
+                return;
             }
         }
+        if (stored !== undefined && !mismatch) {
+            skipped++;
+            return;
+        }
+        try {
+            log(`Fetching original ${file.metadata.title} (${file.id})...`);
+            if (mismatch) {
+                await replaceOriginal(downloadDirectory, file, (dest) =>
+                    lib.fetchOriginal(file.id, dest),
+                );
+            } else {
+                // A fetched original is written straight to its save path (a
+                // live photo beside it); only one that was already cached
+                // elsewhere is copied.
+                await placeOriginal(downloadDirectory, file, (dest) =>
+                    lib.original(file.id, dest),
+                );
+            }
+            storedThisRun.add(file.id);
+            downloaded++;
+        } catch (err) {
+            log(
+                `FAILED original ${file.metadata.title}: ${describeError(err)}`,
+            );
+            recordFailure(
+                file.id,
+                file.metadata.title,
+                collectionName.get(file.collectionID) ?? "",
+                err,
+            );
+        }
+    };
+
+    // Up to the content pool's limit of originals are worked on at once,
+    // their `verify` hashing included, so however large the account is, no
+    // more than that many are open. Each worker takes the next file from the
+    // iterator they all share until none is left. Once a worker has thrown,
+    // as one does when `onProgress` throws, or the lock is lost, no worker
+    // takes another file. The backup fails only once every worker has
+    // stopped, so none writes into the backup after it has ended. It checks
+    // the lock then too, so it does not go on to the thumbnails or the ML
+    // data fetch without it.
+    if (includeOriginals) {
+        const pending = distinct.values();
+        let workerThrew = false;
+        const worker = async (): Promise<void> => {
+            for (const file of pending) {
+                if (workerThrew || lock.lost) return;
+                try {
+                    await storeOriginal(file);
+                } catch (err) {
+                    workerThrew = true;
+                    throw err;
+                }
+            }
+        };
+        const workers = await Promise.allSettled(
+            Array.from({ length: lib.contentConcurrency }, worker),
+        );
+        for (const w of workers) {
+            if (w.status === "rejected") throw w.reason;
+        }
+        if (lock.lost) throw lock.lost;
     }
 
     if (includeThumbnails) {
         for (const [fileID, file] of distinct) {
+            if (lock.lost) throw lock.lost;
             const dest = join(thumbnailsDir, `${fileID}.jpg`);
             if (isPresent(dest)) continue;
             try {
@@ -740,19 +796,21 @@ const runLockedBackup = async (
     // Phase 2: rebuild the derived views from the model. Sidecars first, for
     // every present original (this repairs stale ones), each with the file's
     // ML data once an ML data fetch has completed, and its image metadata.
-    // When the fetch fails, a file with no cached ML data gets the reason
-    // instead and is recorded as failed. The next run fetches its ML data
-    // again because none is cached.
+    // When the fetch got a file's ML data but could not decrypt, gunzip or
+    // parse it, or when the fetch failed, a file with no cached ML data gets
+    // the reason instead and is recorded as failed. The next run fetches its
+    // ML data again because none is cached.
     if (includeOriginals) {
-        let mlDataError: string | undefined;
+        let mlDataFailure: unknown;
         try {
             log("Fetching ML data...");
             await lib.fetchMLData();
         } catch (err) {
-            mlDataError = errorMessage(err);
-            log(`FAILED ML data: ${mlDataError}`);
+            mlDataFailure = err;
+            log(`FAILED ML data: ${describeError(err)}`);
         }
         for (const file of distinct.values()) {
+            if (lock.lost) throw lock.lost;
             const stored = storedAtSavePath(downloadDirectory, file);
             if (stored === undefined) continue;
             const path = withExtension(
@@ -766,13 +824,19 @@ const runLockedBackup = async (
                 storedThisRun.has(file.id),
             );
             const mlData = await lib.mlData(file.id);
-            if (mlData === undefined && mlDataError !== undefined) {
-                writeSidecar(path, file, { mlDataError }, image);
+            const failure = lib.mlDataError(file.id) ?? mlDataFailure;
+            if (mlData === undefined && failure !== undefined) {
+                writeSidecar(
+                    path,
+                    file,
+                    { mlDataError: describeError(failure) },
+                    image,
+                );
                 recordFailure(
                     file.id,
                     file.metadata.title,
                     collectionName.get(file.collectionID) ?? "",
-                    new Error(`ML data: ${mlDataError}`),
+                    new Error("ML data", { cause: failure }),
                 );
             } else {
                 writeSidecar(path, file, { mlData }, image);
@@ -791,11 +855,12 @@ const runLockedBackup = async (
             id: c.id,
             name: sanitizeFileName(c.name, `collection-${c.id}`),
         })),
-        false,
+        { withJSON: true },
     );
     const albumDirNames = new Map(
         allCollections.map((c, i) => [c.id, dirNames[i]!]),
     );
+    if (lock.lost) throw lock.lost;
     try {
         removeStaleAlbumDirs(
             collectionsDir,
@@ -803,60 +868,72 @@ const runLockedBackup = async (
             downloadDirectory,
         );
     } catch (err) {
-        log(`FAILED removing old album directories: ${errorMessage(err)}`);
+        log(`FAILED removing old album directories: ${describeError(err)}`);
     }
 
+    // An album whose folder or JSON cannot be written, such as one whose
+    // folder name is taken by a file the user put there, is logged and listed
+    // in the result, and the run goes on to the next album.
+    const failedAlbums: number[] = [];
     for (const c of collections) {
-        const colDirName = albumDirNames.get(c.id)!;
-        const colDir = join(collectionsDir, colDirName);
-        mkdirSync(colDir, { recursive: true });
-
-        // Every album links the one original, saved from the file's entry in
-        // `distinct`.
-        const files = filesByCollection.get(c.id) ?? [];
-        const links = files.flatMap((f) =>
-            linksFor(
-                f,
-                storedAtSavePath(downloadDirectory, distinct.get(f.id)!),
-            ),
-        );
-        const linkNames = uniqueNames(links, true);
+        if (lock.lost) throw lock.lost;
         try {
-            removeStaleLinks(colDir, new Set(linkNames), downloadDirectory);
-        } catch (err) {
-            log(`FAILED removing old links in ${c.name}: ${errorMessage(err)}`);
-        }
+            const colDirName = albumDirNames.get(c.id)!;
+            const colDir = join(collectionsDir, colDirName);
+            mkdirSync(colDir, { recursive: true });
 
-        const metaFiles = files.map((f) => ({
-            id: f.id,
-            metadata: f.metadata,
-        }));
-        for (const [i, link] of links.entries()) {
-            if (!includeOriginals || link.target === undefined) continue;
-            const linkName = linkNames[i]!;
+            // Every album links the one original, saved from the file's entry
+            // in `distinct`.
+            const files = filesByCollection.get(c.id) ?? [];
+            const links = files.flatMap((f) =>
+                linksFor(
+                    f,
+                    storedAtSavePath(downloadDirectory, distinct.get(f.id)!),
+                ),
+            );
+            const linkNames = uniqueNames(links, { beforeExtension: true });
             try {
-                rebuildSymlink(
-                    join(colDir, linkName),
-                    relative(colDir, link.target),
-                );
+                removeStaleLinks(colDir, new Set(linkNames), downloadDirectory);
             } catch (err) {
                 log(
-                    `FAILED symlink ${c.name}/${linkName}: ${errorMessage(err)}`,
-                );
-                recordFailure(
-                    link.file.id,
-                    link.file.metadata.title,
-                    c.name,
-                    err,
+                    `FAILED removing old links in ${c.name}: ${describeError(err)}`,
                 );
             }
-        }
 
-        writeAlbumJSON(
-            join(collectionsDir, `${colDirName}.json`),
-            c,
-            metaFiles,
-        );
+            const metaFiles = files.map((f) => ({
+                id: f.id,
+                metadata: f.metadata,
+            }));
+            for (const [i, link] of links.entries()) {
+                if (!includeOriginals || link.target === undefined) continue;
+                const linkName = linkNames[i]!;
+                try {
+                    rebuildSymlink(
+                        join(colDir, linkName),
+                        relative(colDir, link.target),
+                    );
+                } catch (err) {
+                    log(
+                        `FAILED symlink ${c.name}/${linkName}: ${describeError(err)}`,
+                    );
+                    recordFailure(
+                        link.file.id,
+                        link.file.metadata.title,
+                        c.name,
+                        err,
+                    );
+                }
+            }
+
+            writeAlbumJSON(
+                join(collectionsDir, `${colDirName}.json`),
+                c,
+                metaFiles,
+            );
+        } catch (err) {
+            log(`FAILED album ${c.name}: ${describeError(err)}`);
+            failedAlbums.push(c.id);
+        }
     }
 
     // Reconcile the ledger against what this run actually attempted: an entry
@@ -864,7 +941,9 @@ const runLockedBackup = async (
     // its failure resolved; a file gone from the library (deleted) or outside
     // this run's scope is not something this run can resolve, so keeping its
     // stale entry would keep the exit code non-zero forever — a single
-    // since-deleted photo would fail every future scheduled backup.
+    // since-deleted photo would fail every future scheduled backup. A backup
+    // whose lock was lost leaves the ledger as it is.
+    if (lock.lost) throw lock.lost;
     for (const fileID of [...ledger.keys()]) {
         if (!failedThisRun.has(fileID)) ledger.delete(fileID);
     }
@@ -880,23 +959,42 @@ const runLockedBackup = async (
         failed: ledger.size,
         errors,
         undecryptableAlbums,
+        failedAlbums,
     };
 };
 
 // Only one backup of a directory runs at a time, in this process or another.
 // Creates `downloadDirectory` if it is missing, takes the lock in it and
-// returns the function that releases it; while another backup holds the lock,
-// fails at once with an error whose `code` is `ELOCKED`. The lock is the
-// directory `backup.lock`, whose modification time proper-lockfile keeps
-// current while it is held. One it has not touched for 10 seconds was left by
-// a run that could not remove it, and is taken over.
+// returns it; while another backup holds the lock, fails at once with an error
+// whose `code` is `ELOCKED`. The lock is the directory `backup.lock`, whose
+// modification time proper-lockfile keeps current while it is held. One it has
+// not touched for 10 seconds was left by a run that could not remove it, and
+// is taken over.
+//
+// proper-lockfile finds the lock lost when `backup.lock` is gone or is another
+// backup's, and calls `onCompromised` from a timer, where throwing would end
+// the process. The `onCompromised` given here sets the lock's `lost` instead,
+// and the backup holding the lock stops before it starts another file (see
+// `runLockedBackup`). Releasing a lost lock leaves `backup.lock` alone, since
+// it may be another backup's.
 export const lockBackupDirectory = async (
     downloadDirectory: string,
-): Promise<() => Promise<void>> => {
+): Promise<BackupLock> => {
     mkdirSync(downloadDirectory, { recursive: true });
+    let release: () => Promise<void>;
+    const lock: BackupLock = {
+        release: async () => {
+            if (lock.lost === undefined) await release();
+        },
+    };
     try {
-        return await lockfile.lock(downloadDirectory, {
+        release = await lockfile.lock(downloadDirectory, {
             lockfilePath: join(downloadDirectory, "backup.lock"),
+            onCompromised: (err) => {
+                lock.lost = new Error(
+                    `lost the lock on ${downloadDirectory}: ${err.message}`,
+                );
+            },
         });
     } catch (err) {
         if ((err as NodeJS.ErrnoException).code !== "ELOCKED") throw err;
@@ -905,10 +1003,11 @@ export const lockBackupDirectory = async (
             { code: "ELOCKED" },
         );
     }
+    return lock;
 };
 
 // Runs the backup holding the lock, which it takes and releases itself unless
-// the caller already holds it (`opts.lockHeld`).
+// the caller passes the lock it took (`opts.lock`).
 export const runBackup = async (
     lib: BackupLibrary,
     opts: BackupOptions,
@@ -920,13 +1019,21 @@ export const runBackup = async (
                 "open the library with one)",
         );
     }
-    if (opts.lockHeld) {
-        return runLockedBackup(lib, opts, downloadDirectory);
+    // As `BoundedPool` does. A limit below 1, or one that is not a number,
+    // would back up no original and end with no error.
+    const { contentConcurrency } = lib;
+    if (!Number.isInteger(contentConcurrency) || contentConcurrency < 1) {
+        throw new RangeError(
+            `contentConcurrency must be a positive integer, got ${contentConcurrency}`,
+        );
     }
-    const release = await lockBackupDirectory(downloadDirectory);
+    if (opts.lock) {
+        return runLockedBackup(lib, opts, downloadDirectory, opts.lock);
+    }
+    const lock = await lockBackupDirectory(downloadDirectory);
     try {
-        return await runLockedBackup(lib, opts, downloadDirectory);
+        return await runLockedBackup(lib, opts, downloadDirectory, lock);
     } finally {
-        await release();
+        await lock.release();
     }
 };

src/cli-commands.ts

@@ -20,7 +20,7 @@ import {
     type ClientSnapshot,
     type LoginOptions,
 } from "./client.js";
-import { lockBackupDirectory } from "./backup.js";
+import { lockBackupDirectory, type BackupLock } from "./backup.js";
 import { init } from "./crypto/index.js";
 import {
     defaultCacheDirectory,
@@ -394,11 +394,11 @@ export const backupMetadataCommand = async (
         // Refresh first so the dump holds current account state, not what the
         // cache last held; a failed refresh throws.
         await lib.fresh();
-        const { failedMLBatches } = await runMetadataBackup(lib, client, dir, {
+        const result = await runMetadataBackup(lib, client, dir, {
             exif: opts.exif || opts.all,
             onProgress: (msg) => ctx.stderr.write(msg + "\n"),
         });
-        return failedMLBatches > 0 ? 1 : 0;
+        return result.filesWithMLDataError > 0 ? 1 : 0;
     } finally {
         await lib.close();
     }
@@ -416,9 +416,9 @@ export const backupCommand = async (
     ctx.stderr.write("Starting backup...\n");
     // The lock is taken before the library opens and starts its refresh, so a
     // run refused while another backup of `dir` runs sends no request.
-    let release: () => Promise<void>;
+    let lock: BackupLock;
     try {
-        release = await lockBackupDirectory(dir);
+        lock = await lockBackupDirectory(dir);
     } catch (err) {
         if ((err as NodeJS.ErrnoException).code !== "ELOCKED") throw err;
         ctx.stderr.write(`quak: ${(err as Error).message}\n`);
@@ -427,17 +427,20 @@ export const backupCommand = async (
     try {
         // The precache is off: the backup fetches what it needs, and must not
         // also fill the cache with every thumbnail and the recent originals.
+        // No refresh starts on a timer: the backup lists its files once, from
+        // the refresh it waits for, and can run for hours.
         const lib = await Library.open({
             client,
             downloadDirectory: dir,
             cacheDirectory: ctx.cacheDir,
+            refreshIntervalSeconds: Infinity,
             precacheThumbnails: false,
             precacheOriginals: false,
         });
         try {
             const result = await lib.backup({
                 downloadDirectory: dir,
-                lockHeld: true,
+                lock,
                 verify: opts.verify,
                 onProgress: (msg) => {
                     if (!opts.json) ctx.stderr.write(msg + "\n");
@@ -467,14 +470,16 @@ export const backupCommand = async (
                 }
             }
 
-            return result.failed > 0 || result.undecryptableAlbums.length > 0
+            return result.failed > 0 ||
+                result.undecryptableAlbums.length > 0 ||
+                result.failedAlbums.length > 0
                 ? 1
                 : 0;
         } finally {
             await lib.close();
         }
     } finally {
-        await release();
+        await lock.release();
     }
 };
 

src/cli-run.ts

@@ -2,14 +2,17 @@
 
 import type { Writable } from "node:stream";
 import { ApiError } from "./api/client.js";
+import { describeError } from "./errors.js";
 
 // Run a command and exit with its code once stdout/stderr have drained.
 // Exiting before the drain can truncate piped output, and the library can keep
 // the event loop alive after a command returns, so a plain return could hang.
 // An error the command throws is printed as one `quak: MESSAGE` line, without
-// the stack trace, and exits 1. A 401 from the server means it no longer
-// accepts the saved session: that prints one line saying to log in again and
-// exits 3, as a missing or corrupt session file does.
+// the stack trace, and exits 1. MESSAGE is the error as `describeError` gives
+// it, so it carries the reason Node's `fetch` puts on `cause`. A 401 from the
+// server means it no longer accepts the saved session: that prints one line
+// saying to log in again and exits 3, as a missing or corrupt session file
+// does.
 export const run = async (
     command: Promise<number>,
     stdout: Writable,
@@ -26,9 +29,7 @@ export const run = async (
             );
             code = 3;
         } else {
-            stderr.write(
-                `quak: ${err instanceof Error ? err.message : String(err)}\n`,
-            );
+            stderr.write(`quak: ${describeError(err)}\n`);
             code = 1;
         }
     }

src/client.ts

@@ -373,13 +373,17 @@ export class Client {
     }
 
     // Fetch machine-learning data (face detections + CLIP embeddings) for up
-    // to a batch of files, each decrypted with its own key. One request; the
-    // library batches at `MLDATA_BATCH_SIZE` and schedules each batch through
-    // its metadata request pool.
+    // to a batch of files, each decrypted with its own key, and for each file
+    // whose ML data did not decrypt, gunzip or parse, why not. One request;
+    // the library batches at `MLDATA_BATCH_SIZE` and schedules each batch
+    // through its metadata request pool.
     async fetchMLData(args: {
         fileIDs: number[];
         fileKeys: Map<number, Uint8Array>;
-    }): Promise<Map<number, MLData>> {
+    }): Promise<{
+        payloads: Map<number, MLData>;
+        errors: Map<number, unknown>;
+    }> {
         this.assertLoggedIn();
         return fetchMLDataBatch(this.api, args.fileIDs, args.fileKeys);
     }

src/download/index.ts

@@ -135,15 +135,17 @@ const streamDecrypt = async (
                     // cannot tell a partial chunk from a corrupt one, so this
                     // is reported as the truncation it almost always is, with
                     // the authentication failure kept as the error's cause.
-                    // Only the pull is guarded: a sink failure on a chunk
-                    // that did authenticate is a disk error, not a
+                    // The message leaves that failure to the cause, which
+                    // `describeError` appends to what `quak backup` prints and
+                    // records. Only the pull is guarded: a sink failure on a
+                    // chunk that did authenticate is a disk error, not a
                     // truncation.
                     let pulled;
                     try {
                         pulled = pullStreamChunk(state, buffer);
                     } catch (err) {
                         throw new TruncatedStreamError(
-                            `download: stream truncated: response body ended with ${buffer.length} trailing bytes that did not authenticate as a final chunk (transfer stopped mid-chunk, or the data is corrupt)`,
+                            `download: stream truncated: response body ended with ${buffer.length} trailing bytes (transfer stopped mid-chunk, or the data is corrupt)`,
                             { cause: err },
                         );
                     }
@@ -197,11 +199,13 @@ const isRunning = (pid: number): boolean => {
 };
 
 // Delete the temp files a killed process left in `dir`: the writer's
-// `.quak-<pid>-<random>.tmp` and the backup copy's
-// `.quak-backup-<name>-<pid>-<random>.tmp`. Only files whose process is no
-// longer running are removed, so another process writing into the same
-// directory keeps its own. A reused process ID can only keep a leftover a while
-// longer, never remove a live one.
+// `.quak-<pid>-<random>.tmp`, the backup copy's
+// `.quak-backup-<name>-<pid>-<random>.tmp`, and the image and video of a live
+// photo downloaded to replace a stored one, `.quak-<pid>-<random>.<ext>` (see
+// `replaceOriginal`). Only files whose process is no longer running are
+// removed, so another process writing into the same directory keeps its own. A
+// reused process ID can only keep a leftover a while longer, never remove a
+// live one.
 export const removeLeftoverTempFiles = (dir: string): void => {
     let names: string[];
     try {
@@ -210,7 +214,9 @@ export const removeLeftoverTempFiles = (dir: string): void => {
         return;
     }
     for (const name of names) {
-        const match = /^\.quak-(?:.*-)?(\d+)-[0-9a-z]*\.tmp$/.exec(name);
+        const match = /^\.quak-(?:.*-)?(\d+)-[0-9a-z]*\.[0-9A-Za-z]+$/.exec(
+            name,
+        );
         if (match && !isRunning(Number(match[1]))) {
             rmSync(join(dir, name), { force: true });
         }

src/errors.ts

@@ -1,10 +1,13 @@
-// Error types that more than one layer of quak needs to recognise.
+// Error types that more than one layer of quak needs to recognise, and how an
+// error is described in what quak prints and in `failures.json`.
 //
-// They live here rather than beside the code that throws them so that the
+// The types live here rather than beside the code that throws them so that the
 // retry classifier can identify them without importing the HTTP client or the
 // download layer — both of which import the classifier. `ApiError` is
 // re-exported from `src/api/client.ts`, which is where callers have always
-// imported it from and where it still belongs conceptually.
+// imported it from and where it still belongs conceptually. The walk along an
+// error's `cause` chain is here too, because the retry classifier reads the
+// chain for errnos and `describeError` for messages.
 
 export class ApiError extends Error {
     readonly status: number;
@@ -43,3 +46,68 @@ export class TruncatedStreamError extends Error {
         this.name = "TruncatedStreamError";
     }
 }
+
+// `cause` is an arbitrary user-settable property and nothing prevents it from
+// forming a cycle, so the walk is bounded. Hanging the process would be worse
+// than misclassifying an error or describing only part of it.
+const MAX_CAUSE_DEPTH = 8;
+
+// The error and each error on its `cause` chain, top first, at most
+// `MAX_CAUSE_DEPTH` of them. `complete` is false when the walk stopped at the
+// depth limit with more of the chain still below it.
+export const causeChain = (
+    err: unknown,
+): { chain: object[]; complete: boolean } => {
+    const chain: object[] = [];
+    let current: unknown = err;
+    for (let depth = 0; depth < MAX_CAUSE_DEPTH; depth++) {
+        if (current === null || typeof current !== "object") {
+            return { chain, complete: true };
+        }
+        chain.push(current);
+        const { cause } = current as { cause?: unknown };
+        if (cause === current) return { chain, complete: true };
+        current = cause;
+    }
+    return { chain, complete: current === null || typeof current !== "object" };
+};
+
+// An error as quak prints it and records it in `failures.json`: its message,
+// then the message of each error on its `cause` chain, joined with ": ". Node's
+// `fetch` rejects with just "fetch failed" and puts the reason on `cause`, so
+// this gives "fetch failed: connect ECONNREFUSED 1.2.3.4:443". When the host
+// has several addresses and every one fails, the cause is an `AggregateError`
+// with an empty message and one error per address in `errors`, so an
+// `AggregateError` is given as the messages of its errors, joined with ", ".
+// A cause that is not an object, such as a thrown string, is given as
+// `String(cause)`. Each message is trimmed of the white space around it, as
+// OpenSSL ends its own with a newline. A message the text already holds is
+// left out.
+export const describeError = (err: unknown): string => {
+    if (!(err instanceof Error)) return String(err);
+    const { chain } = causeChain(err);
+    const messages = chain
+        .slice(1)
+        .map((cause) =>
+            cause instanceof AggregateError
+                ? cause.errors
+                      .map((e: unknown) =>
+                          (e instanceof Error ? e.message : String(e)).trim(),
+                      )
+                      .join(", ")
+                : (cause as { message?: unknown }).message,
+        );
+    // `causeChain` stops at a cause that is not an object, so that one is read
+    // from the last error on the chain.
+    const last = chain[chain.length - 1] as { cause?: unknown };
+    if (last.cause !== undefined && typeof last.cause !== "object") {
+        messages.push(String(last.cause));
+    }
+    let text = err.message.trim();
+    for (const message of messages) {
+        if (typeof message !== "string") continue;
+        const trimmed = message.trim();
+        if (!text.includes(trimmed)) text += `: ${trimmed}`;
+    }
+    return text;
+};

src/filename.ts

@@ -4,7 +4,8 @@
 // hands us, and quak does not trust the server. Any name taken from them and
 // used on disk goes through here, so it can only ever name one file inside the
 // directory the caller chose: never a path, never `..`, never hidden, never a
-// Windows device name.
+// Windows device name. Such a name can still be too long for a file name, and
+// `shortenToBytes` cuts one short.
 //
 // A path the user typed (`--out`, `outPath`) is not passed through here: the
 // caller is trusted, the server is not.
@@ -39,3 +40,20 @@ export const safeExtension = (title: string): string => {
 // `name` with its extension, if it has one, replaced by `ext` (".mov").
 export const withExtension = (name: string, ext: string): string =>
     name.slice(0, name.length - extname(name).length) + ext;
+
+// The longest file name, in bytes, that ext4 and most other file systems allow.
+export const MAX_NAME_BYTES = 255;
+
+// `name` cut short to at most `maxBytes` bytes of UTF-8, the encoding a file
+// name is written to disk in, between two characters, never inside one. A name
+// that fits comes back unchanged.
+export const shortenToBytes = (name: string, maxBytes: number): string => {
+    let bytes = 0;
+    let end = 0;
+    for (const character of name) {
+        bytes += Buffer.byteLength(character);
+        if (bytes > maxBytes) break;
+        end += character.length;
+    }
+    return name.slice(0, end);
+};

src/index.ts

@@ -69,6 +69,7 @@ export {
     type EnsureEvent,
     lockBackupDirectory,
     runBackup,
+    type BackupLock,
     type BackupOptions,
     type BackupResult,
     type BackupError,

src/library/content.ts

@@ -13,6 +13,8 @@
 // originals, the thumbnail pool for thumbnails. The pool limits concurrency,
 // orders on-demand work ahead of background, and dedups by key so a fileID
 // requested twice while the first is still in flight downloads once.
+// `fetchOriginal` passes no key, so it is not shared with a fetch of the same
+// file already in flight.
 //
 // Integrity. The reused streaming decrypt is the enforced guarantee: every
 // chunk is authenticated and the writer renames the file into place only once
@@ -27,9 +29,10 @@
 // It also names each original's save path under the download directory
 // (`savePath`), where `Photo.download()` and `lib.backup()` put it. A copy in
 // the cache does not count as saved there, but is copied there rather than
-// fetched again.
+// fetched again. An original stored there whose bytes do not match its content
+// hash is replaced only by a download that does (`replaceOriginal`).
 
-import { existsSync, readFileSync, statSync } from "node:fs";
+import { createReadStream, existsSync, readFileSync, statSync } from "node:fs";
 import {
     chmod,
     copyFile,
@@ -44,6 +47,12 @@ import {
 import { basename, dirname, extname, join } from "node:path";
 
 import type { ApiClient } from "../api/client.js";
+import {
+    chunkHashFinal,
+    chunkHashInit,
+    chunkHashUpdate,
+    init,
+} from "../crypto/index.js";
 import {
     downloadFile,
     downloadThumbnail,
@@ -349,6 +358,26 @@ export const storedAtSavePath = (
     return storedOriginal(dirname(path), basename(path, extname(path)), file);
 };
 
+// The content hash of the original stored at `stored`, computed as the download
+// check computes it: over each file's bytes, read in chunks, and for a live
+// photo `<imageHash>:<videoHash>`.
+export const storedHash = async (stored: {
+    path: string;
+    videoPath?: string;
+}): Promise<string> => {
+    await init();
+    const hashFile = async (path: string): Promise<string> => {
+        const state = chunkHashInit();
+        for await (const chunk of createReadStream(path)) {
+            chunkHashUpdate(state, chunk as Buffer);
+        }
+        return chunkHashFinal(state);
+    };
+    const hash = await hashFile(stored.path);
+    if (stored.videoPath === undefined) return hash;
+    return `${hash}:${await hashFile(stored.videoPath)}`;
+};
+
 // Copy bytes into `dest` via a temp file in the same directory plus rename, so
 // `dest` appears only once it is whole ("present means complete"). As in the
 // download writer, the temp file is fsynced before the rename and the directory
@@ -405,6 +434,62 @@ export const placeOriginal = async (
     return { path, videoPath };
 };
 
+// Replace the original of `file` stored at its save path under `root` with a
+// new copy, keeping the stored one until the copy is complete and its hash,
+// computed as `storedHash` computes it, is the content hash the file's metadata
+// records. `get` is given a temporary file beside the save path and writes the
+// copy there, a live photo as its image and its video beside that file, each
+// with its own extension. Only a copy that matches is renamed over the stored
+// one, and a live photo's image and video are both hashed before either is
+// renamed. The image is renamed first: should the run stop, or the video's
+// rename fail, before the video is renamed, the new image is left beside the
+// old video, and a later backup with `verify` replaces the video if the video
+// was the part that did not match. When `get` fails or the copy does not match,
+// the temporary files are removed and the stored original stays as it was.
+export const replaceOriginal = async (
+    root: string,
+    file: EnteFile,
+    get: (dest: string) => Promise<{ path: string; videoPath?: string }>,
+): Promise<{ path: string; videoPath?: string }> => {
+    const dest = savePath(root, file);
+    const dir = dirname(dest);
+    // Named as the download writer names its temporary files; a live photo's
+    // image and video get this name with their own extensions. The next
+    // backup removes any that a killed run leaves (`removeLeftoverTempFiles`).
+    const tmp = join(
+        dir,
+        `.quak-${process.pid}-${Math.random().toString(36).slice(2)}.tmp`,
+    );
+    let got: { path: string; videoPath?: string } | undefined;
+    try {
+        got = await get(tmp);
+        if ((await storedHash(got)) !== file.metadata.hash) {
+            throw new Error(
+                "the downloaded original does not match its content hash either",
+            );
+        }
+        if (got.videoPath === undefined) {
+            await rename(got.path, dest);
+            await fsyncPath(dir);
+            return { path: dest };
+        }
+        const path = withExtension(dest, extname(got.path));
+        const videoPath = withExtension(dest, extname(got.videoPath));
+        await rename(got.path, path);
+        await rename(got.videoPath, videoPath);
+        const name = basename(dest, extname(dest));
+        await writeLivePhotoJSON(dir, name, { path, videoPath });
+        return { path, videoPath };
+    } finally {
+        // Whatever was not renamed into place. A failure to remove it must
+        // not replace the error that explains what went wrong.
+        for (const path of [tmp, got?.path, got?.videoPath]) {
+            if (path === undefined) continue;
+            await rm(path, { force: true }).catch(() => undefined);
+        }
+    }
+};
+
 export class ContentCache implements PhotoContent, ThumbnailsAPI {
     private readonly pools: RequestPools;
     private readonly source: ContentSource;
@@ -543,6 +628,35 @@ export class ContentCache implements PhotoContent, ThumbnailsAPI {
         };
     }
 
+    // Fetch an original through the content pool to `destination`, a live
+    // photo as its image and its video beside it, and return where it landed.
+    // Unlike `backupOriginal`, it fetches even when the cache or the save path
+    // already holds the original, and it records nothing: what it writes is
+    // the caller's. It is not shared with a fetch of the same file already
+    // under way, which writes somewhere else.
+    async fetchOriginal(
+        fileID: number,
+        destination: string,
+    ): Promise<{ path: string; videoPath?: string }> {
+        const file = this.getFile(fileID);
+        if (!file) throw new Error(`content cache: unknown file ${fileID}`);
+        return this.pools.content.run(
+            async () => {
+                const fetched = await this.fetchInto(
+                    file,
+                    destination,
+                    "original",
+                    undefined,
+                );
+                for (const path of [fetched.path, fetched.videoPath]) {
+                    if (path !== undefined) await chmod(path, FILE_MODE);
+                }
+                return fetched;
+            },
+            { priority: "on-demand" },
+        );
+    }
+
     async ensure(args: EnsureOptions): Promise<EnsureResult[]> {
         return this.ensureThumbnails(args);
     }

src/library/index.ts

@@ -89,6 +89,7 @@ export {
 } from "./content.js";
 export { type MLDataAPI, type SimilarResult } from "./mlsearch.js";
 import type { CollectionsPage, FilesPage } from "../client.js";
+import { describeError } from "../errors.js";
 import { MLDATA_BATCH_SIZE, type MLData } from "../mldata-fetch.js";
 import type { Collection, EnteFile } from "../model/types.js";
 import { runBackup, type BackupOptions, type BackupResult } from "../backup.js";
@@ -96,6 +97,7 @@ import { runBackup, type BackupOptions, type BackupResult } from "../backup.js";
 export {
     lockBackupDirectory,
     runBackup,
+    type BackupLock,
     type BackupOptions,
     type BackupResult,
     type BackupError,
@@ -137,12 +139,16 @@ export interface LibraryClient {
         sinceTime: number;
     }): Promise<FilesPage>;
     // Fetch ML data (face detections + CLIP embeddings) for up to a batch of
-    // files. Optional: a client without it simply disables ML fetching, leaving
-    // the metadata refresh untouched.
+    // files, and for each file whose ML data did not decrypt, gunzip or parse,
+    // why not. Optional: a client without it simply disables ML fetching,
+    // leaving the metadata refresh untouched.
     fetchMLData?(args: {
         fileIDs: number[];
         fileKeys: Map<number, Uint8Array>;
-    }): Promise<Map<number, MLData>>;
+    }): Promise<{
+        payloads: Map<number, MLData>;
+        errors: Map<number, unknown>;
+    }>;
     // The byte source for the on-disk content cache. Optional so a mock client
     // that only serves metadata still satisfies the interface; when absent (and
     // no explicit `contentSource` is passed to `open`) the content cache is
@@ -153,9 +159,11 @@ export interface LibraryClient {
 // A progress event for one unit of background work. A metadata "refresh" or an
 // ML "fetchMLData" pass each fire "started" before their network work and then
 // exactly one of "done" or "failed"; "failed" carries the error message and
-// an ML "done" reports how many payloads it stored. The precache fills
-// ("precacheThumbnails"/"precacheOriginals", #48) fire "started"/"done" around
-// each sweep that has work, "done" reporting the count newly cached.
+// an ML "done" reports how many payloads it stored and, in `mlDataErrors`,
+// each file whose ML data did not decrypt, gunzip or parse, with the reason.
+// The precache fills ("precacheThumbnails"/"precacheOriginals", #48) fire
+// "started"/"done" around each sweep that has work, "done" reporting the count
+// newly cached.
 export interface RefreshEvent {
     operation:
         | "refresh"
@@ -165,6 +173,7 @@ export interface RefreshEvent {
     status: "started" | "done" | "failed";
     error?: string;
     fetched?: number;
+    mlDataErrors?: { fileID: number; error: string }[];
 }
 
 export type RefreshProgressCallback = (event: RefreshEvent) => void;
@@ -179,6 +188,9 @@ export interface LibraryOptions {
     // directory at open. The content cache treats an original already stored
     // at its save path as present.
     downloadDirectory?: string;
+    // How long after one refresh ends the background timer starts the next.
+    // `Infinity` starts none: the library then refreshes only as it opens and
+    // when `fresh()` or `backup()` asks.
     refreshIntervalSeconds?: number;
     onProgress?: RefreshProgressCallback;
     // The bounded request pools (issue #45), shared by the ML-data fetch (the
@@ -221,6 +233,11 @@ export interface LibraryStatus {
     lastMLFetchAt?: number;
     // The most recent ML fetch pass's error, set only while it failed.
     lastMLError?: string;
+    // Each file whose ML data the last ML fetch pass that succeeded got but
+    // could not decrypt, gunzip or parse, with the reason. A pass still
+    // running leaves it as it is. A pass that fails leaves it as it was. Such
+    // ML data is not cached, so the next pass asks for it again.
+    mlDataErrors: { fileID: number; error: string }[];
     // ML payloads stored on disk and CLIP embeddings in the index; undefined
     // when ML fetching is disabled.
     mlStored?: number;
@@ -288,6 +305,12 @@ export class Library {
     private lastError?: string;
     private lastMLFetchAt?: number;
     private lastMLError?: string;
+    // For each file whose ML data the last ML fetch pass that succeeded got
+    // but could not decrypt, gunzip or parse, why not, by fileID. A pass
+    // gathers its own and puts them here when it ends without failing, so
+    // `status()` and `backup()` never see the list of a pass that is still
+    // running or that failed. Kept in RAM only, never in the cache.
+    private mlDataErrors = new Map<number, unknown>();
     // The plain-record projection as of the last refresh, and the GUI change
     // subscribers. A refresh that alters the projection notifies each with the
     // delta; `lastRecords` is kept current every refresh so a subscriber that
@@ -538,6 +561,7 @@ export class Library {
             lastError: this.lastError,
             lastMLFetchAt: this.lastMLFetchAt,
             lastMLError: this.lastMLError,
+            mlDataErrors: this.describeMLDataErrors(),
             mlStored: ml?.stored,
             mlIndexed: ml?.indexed,
             originalsUsedBytes: originals?.usedBytes,
@@ -567,9 +591,10 @@ export class Library {
 
     // Back up every in-scope file to `opts.downloadDirectory`, or else the
     // library's, each original at its save path, with a durable failure
-    // ledger (issue #51). Takes the lock in that directory first, unless
-    // `opts.lockHeld` says the caller holds it, failing at once while another
-    // backup of it runs (see `runBackup`). Waits for a
+    // ledger (issue #51). Takes the lock in that directory first, unless the
+    // caller passes the lock it took as `opts.lock`, failing at once while
+    // another backup of it runs, and before it starts another file once the
+    // lock is lost (see `lockBackupDirectory`). Waits for a
     // completed refresh, as `fresh()` does, joining one already running, and
     // rejects before touching any file but the lock when it fails. Then puts
     // pending originals at their save paths as `Photo.download()` does (and
@@ -601,9 +626,13 @@ export class Library {
                 undecryptable: () => this.store.listUndecryptable(),
                 original: (fileID, destination) =>
                     cache!.backupOriginal(fileID, destination),
+                fetchOriginal: (fileID, destination) =>
+                    cache!.fetchOriginal(fileID, destination),
                 thumbnail: (fileID) => cache!.thumbnail(fileID),
+                contentConcurrency: this.pools.content.concurrency,
                 fetchMLData: () => this.fetchMLDataNow(),
                 mlData: (fileID) => this.mldata.forFile({ fileID }),
+                mlDataError: (fileID) => this.mlDataErrors.get(fileID),
             },
             { ...opts, downloadDirectory },
         );
@@ -628,7 +657,8 @@ export class Library {
     }
 
     private scheduleNext(): void {
-        if (this.closed) return;
+        // `setTimeout` would turn an infinite delay into 1 ms.
+        if (this.closed || this.intervalMs === Infinity) return;
         this.timer = setTimeout(() => {
             void this.runRefresh().then(() => this.scheduleNext());
         }, this.intervalMs);
@@ -829,8 +859,11 @@ export class Library {
 
     // One ML fetch pass: fetch, decrypt and store the ML data for every file
     // the store knows about that is not cached (or whose `updationTime` has
-    // advanced), through the metadata pool, and update the CLIP index. A
-    // failure is reported through `status()` and `onProgress`, then thrown.
+    // advanced), through the metadata pool, and update the CLIP index. For a
+    // file whose ML data does not decrypt, gunzip or parse, it gathers the
+    // reason instead, and puts the reasons it gathered in `mlDataErrors` when
+    // it ends without failing. A failure leaves `mlDataErrors` as it was and
+    // is reported through `status()` and `onProgress`, then thrown.
     private async runMLFetch(): Promise<void> {
         const mldata = this.mlStore;
         // Bind so the call keeps the client as its receiver when invoked
@@ -840,7 +873,14 @@ export class Library {
 
         const files = this.uniqueFiles();
         const needed = mldata.neededFor(files);
-        if (needed.length === 0) return;
+        // This pass's reasons, put in `this.mlDataErrors` when it ends without
+        // failing. A file whose ML data did not decode in the pass before is
+        // still among `needed`, so this pass gives its reason anew.
+        const mlDataErrors = new Map<number, unknown>();
+        if (needed.length === 0) {
+            this.mlDataErrors = mlDataErrors;
+            return;
+        }
 
         this.emit({ operation: "fetchMLData", status: "started" });
         try {
@@ -855,20 +895,25 @@ export class Library {
             for (let i = 0; i < needed.length; i += MLDATA_BATCH_SIZE) {
                 if (this.closed) break;
                 const batch = needed.slice(i, i + MLDATA_BATCH_SIZE);
-                const payloads = await this.pools.metadata.run(
+                const { payloads, errors } = await this.pools.metadata.run(
                     () => fetchMLData({ fileIDs: batch, fileKeys }),
                     { priority: "background" },
                 );
+                for (const [id, error] of errors) {
+                    mlDataErrors.set(id, error);
+                }
                 stored += (await mldata.storeFetched(payloads, updation))
                     .stored;
             }
 
+            this.mlDataErrors = mlDataErrors;
             this.lastMLFetchAt = Date.now();
             this.lastMLError = undefined;
             this.emit({
                 operation: "fetchMLData",
                 status: "done",
                 fetched: stored,
+                mlDataErrors: this.describeMLDataErrors(),
             });
         } catch (err) {
             const error = err instanceof Error ? err.message : String(err);
@@ -878,6 +923,14 @@ export class Library {
         }
     }
 
+    // `mlDataErrors` as `status()` and the ML "done" event give it.
+    private describeMLDataErrors(): { fileID: number; error: string }[] {
+        return [...this.mlDataErrors].map(([fileID, err]) => ({
+            fileID,
+            error: describeError(err),
+        }));
+    }
+
     // The distinct files the store holds, one entry per fileID (a file in
     // several collections shares its ML data), each carrying the key and the
     // newest `updationTime` seen across its memberships.

src/library/store.ts

@@ -119,7 +119,12 @@ export class MetadataStore {
                 };
                 store.files.set(fileKey(file.collectionID, file.id), file);
             }
-            store.undecryptable = parsed.undecryptable ?? [];
+            for (const record of parsed.undecryptable) {
+                if (typeof record.collectionID !== "number") {
+                    throw new Error("damaged record of what did not decrypt");
+                }
+                store.undecryptable.push(record);
+            }
         } catch {
             // Any corruption discards the partial result: a half-read cache is
             // worse than an empty one, since the refresh unit will rebuild it.

src/metadata-backup.ts

@@ -2,6 +2,7 @@ import { mkdirSync, readFileSync, writeFileSync } from "node:fs";
 import { join } from "node:path";
 import * as jpeg from "jpeg-js";
 import type { Client } from "./client.js";
+import { describeError } from "./errors.js";
 import { readExifTags } from "./exif.js";
 import type { Library, Photo } from "./library/index.js";
 import { sanitizeFileName } from "./filename.js";
@@ -91,14 +92,15 @@ const extractExif = async (
 // of plain JSON: account, per-collection, and per-file records including the
 // private and public magic metadata and (by default) the ML data. Collections
 // and files are enumerated from the library's cache, which the caller refreshes
-// first. Returns how many ML data requests failed; their files are still
+// first. Returns how many files have no ML data because their ML data request
+// failed or their ML data did not decrypt, gunzip or parse; they are still
 // written, with `mlDataError` in place of `mlData`.
 export const runMetadataBackup = async (
     lib: Library,
     client: Client,
     outDir: string,
     opts?: MetadataBackupOptions,
-): Promise<{ failedMLBatches: number }> => {
+): Promise<{ filesWithMLDataError: number }> => {
     const log = opts?.onProgress ?? (() => {});
     const wantExif = opts?.exif ?? false;
 
@@ -164,24 +166,28 @@ export const runMetadataBackup = async (
     }
 
     // One failed request (retries exhausted) must not end the dump: its files
-    // get the reason in `mlDataError` and the other batches go on.
+    // get the reason in `mlDataError` and the other batches go on. So does a
+    // file whose ML data does not decrypt, gunzip or parse.
     log("Fetching ML data (face detections, CLIP embeddings)...");
     const mlDataMap = new Map<number, MLData>();
     const mlDataErrors = new Map<number, string>();
-    let failedMLBatches = 0;
     const fileIDs = [...fileKeys.keys()];
     for (let i = 0; i < fileIDs.length; i += MLDATA_BATCH_SIZE) {
         const batch = fileIDs.slice(i, i + MLDATA_BATCH_SIZE);
         try {
-            const result = await fetchMLDataBatch(
+            const { payloads, errors } = await fetchMLDataBatch(
                 client.getApiClient(),
                 batch,
                 fileKeys,
             );
-            for (const [id, payload] of result) mlDataMap.set(id, payload);
+            for (const [id, payload] of payloads) mlDataMap.set(id, payload);
+            for (const [id, err] of errors) {
+                const reason = describeError(err);
+                log(`ML data of file ${id} could not be read: ${reason}`);
+                mlDataErrors.set(id, reason);
+            }
         } catch (err) {
             const reason = err instanceof Error ? err.message : String(err);
-            failedMLBatches++;
             log(
                 `ML data request for ${batch.length} file(s) failed: ${reason}`,
             );
@@ -229,5 +235,5 @@ export const runMetadataBackup = async (
     }
 
     log("Metadata backup complete.");
-    return { failedMLBatches };
+    return { filesWithMLDataError: mlDataErrors.size };
 };

src/mldata-fetch.ts

@@ -28,43 +28,52 @@ interface RawRemoteFileData {
     updatedAt?: number;
 }
 
-// Decrypt one entry with its file key and gunzip the JSON payload. Returns
-// undefined when the key is unknown or the entry does not decrypt/parse, so one
-// corrupt file never fails a whole batch.
-const decodeEntry = (
-    entry: RawRemoteFileData,
-    key: Uint8Array | undefined,
-): MLData | undefined => {
-    if (!key) return undefined;
-    try {
-        const decrypted = decryptBlob(
-            fromBase64(entry.encryptedData),
-            fromBase64(entry.decryptionHeader),
-            key,
-        );
-        const json = gunzipSync(Buffer.from(decrypted)).toString("utf-8");
-        return JSON.parse(json) as MLData;
-    } catch {
-        return undefined;
+// Decrypt one entry with its file key and gunzip the JSON payload. Throws why
+// when the entry does not decrypt, gunzip or parse, or parses to something
+// other than a JSON object, such as `null`, a number or an array.
+const decodeEntry = (entry: RawRemoteFileData, key: Uint8Array): MLData => {
+    const decrypted = decryptBlob(
+        fromBase64(entry.encryptedData),
+        fromBase64(entry.decryptionHeader),
+        key,
+    );
+    const json = gunzipSync(Buffer.from(decrypted)).toString("utf-8");
+    const payload: unknown = JSON.parse(json);
+    if (
+        typeof payload !== "object" ||
+        payload === null ||
+        Array.isArray(payload)
+    ) {
+        throw new Error("not a JSON object");
     }
+    return payload as MLData;
 };
 
 // Fetch ML data for up to `MLDATA_BATCH_SIZE` ids in a single request. This is
 // the unit the request pools schedule; callers with more ids split them into
-// batches and submit each batch to the pool.
+// batches and submit each batch to the pool. Returns the payload of each file
+// whose entry decodes and, in `errors`, why not for each file whose entry does
+// not, so one corrupt file never fails a whole batch. An entry for a file
+// whose key is not known, which quak did not ask for, is skipped.
 export const fetchMLDataBatch = async (
     api: ApiClient,
     fileIDs: number[],
     fileKeys: Map<number, Uint8Array>,
-): Promise<Map<number, MLData>> => {
+): Promise<{ payloads: Map<number, MLData>; errors: Map<number, unknown> }> => {
     const { data } = await api.postJSON<{ data: RawRemoteFileData[] }>(
         "/files/data/fetch",
         { type: "mldata", fileIDs },
     );
-    const result = new Map<number, MLData>();
+    const payloads = new Map<number, MLData>();
+    const errors = new Map<number, unknown>();
     for (const entry of data ?? []) {
-        const payload = decodeEntry(entry, fileKeys.get(entry.fileID));
-        if (payload) result.set(entry.fileID, payload);
+        const key = fileKeys.get(entry.fileID);
+        if (!key) continue;
+        try {
+            payloads.set(entry.fileID, decodeEntry(entry, key));
+        } catch (err) {
+            errors.set(entry.fileID, err);
+        }
     }
-    return result;
+    return { payloads, errors };
 };

src/retry.ts

@@ -10,7 +10,7 @@
 // declining to retry a transient one costs a single file that the next run
 // picks up anyway. When the evidence is ambiguous, fail fast.
 
-import { ApiError, TruncatedStreamError } from "./errors.js";
+import { ApiError, causeChain, TruncatedStreamError } from "./errors.js";
 
 export interface RetryOptions {
     // Total calls, not retries: `attempts: 1` disables retrying.
@@ -58,6 +58,14 @@ export const resolveRetryOptions = (
 });
 
 // Transport failures: the request did not complete, for reasons below HTTP.
+//
+// The errnos are Node's. The binary `make install` builds runs under bun,
+// whose `fetch` rejects with a plain `Error` that has no `cause` and a `code`
+// of bun's own. Bun gives `ConnectionRefused` for a refused connection and for
+// a TLS handshake that failed, and `ConnectionRefused` or `FailedToOpenSocket`
+// for a name that does not resolve. For a connection reset or closed before
+// the response ended, it gives Node's `ECONNRESET`. These were found by
+// running that binary, built with bun 1.3.13, against listeners on 127.0.0.1.
 const TRANSPORT_CODES = new Set([
     "ECONNRESET",
     "ECONNABORTED",
@@ -70,6 +78,9 @@ const TRANSPORT_CODES = new Set([
     "ENETUNREACH",
     "ENETRESET",
     "ENETDOWN",
+    // Bun's.
+    "ConnectionRefused",
+    "FailedToOpenSocket",
 ]);
 
 // The subset of the above that can only be reported before a TCP connection
@@ -87,13 +98,14 @@ const TRANSPORT_CODES = new Set([
 // ambiguity this set exists to exclude. They stay in `TRANSPORT_CODES`, so
 // they remain retryable for idempotent calls; only replay eligibility is
 // narrowed. See `isSafeToReplay`.
+//
+// Bun's `ConnectionRefused` and `FailedToOpenSocket` are absent too, so a
+// `POST` or `PUT` is not replayed on them. When a connection bun kept open from
+// an earlier request closes before any answer, bun sends the request again on
+// a new connection, and if that one fails it gives one of these codes although
+// the server may already have the request.
 const CONNECT_CODES = new Set(["ENOTFOUND", "EAI_AGAIN", "ECONNREFUSED"]);
 
-// `cause` is an arbitrary user-settable property and nothing prevents it from
-// forming a cycle, so the walk is bounded. Hanging the process would be a
-// worse outcome than any misclassification.
-const MAX_CAUSE_DEPTH = 8;
-
 // Collect every `code` in an error's cause chain. undici does not put the
 // errno on the error it throws — it hangs the underlying socket error off
 // `cause`, sometimes more than one level down — so a classifier that only read
@@ -101,20 +113,21 @@ const MAX_CAUSE_DEPTH = 8;
 // connection permanent. `complete` is false when the walk stopped at the
 // depth limit with more of the chain still below it.
 const causeCodes = (err: unknown): { codes: string[]; complete: boolean } => {
+    const { chain, complete } = causeChain(err);
     const codes: string[] = [];
-    let current: unknown = err;
-    for (let depth = 0; depth < MAX_CAUSE_DEPTH; depth++) {
-        if (current === null || typeof current !== "object") {
-            return { codes, complete: true };
-        }
-        const { code, cause } = current as { code?: unknown; cause?: unknown };
+    for (const link of chain) {
+        const { code } = link as { code?: unknown };
         if (typeof code === "string") codes.push(code);
-        if (cause === current) return { codes, complete: true };
-        current = cause;
     }
-    return { codes, complete: current === null || typeof current !== "object" };
+    return { codes, complete };
 };
 
+// Does the error's cause chain hold the code of a transport failure, one of
+// `TRANSPORT_CODES`? `isRetryable` retries such an error, and `classify` in
+// `src/backup.ts` calls it transient.
+export const hasTransportCode = (err: unknown): boolean =>
+    causeCodes(err).codes.some((code) => TRANSPORT_CODES.has(code));
+
 const isAbort = (err: unknown): boolean => {
     if (err === null || typeof err !== "object") return false;
     const { name } = err as { name?: unknown };
@@ -156,9 +169,10 @@ export const isRetryable = (err: unknown): boolean => {
     // deliberately literal. Demanding a recognised `cause` instead would
     // classify real network failures as permanent and fail backups that should
     // have succeeded; the cost of the imprecision is bounded by the attempt
-    // count.
+    // count. Bun's fetch rejects with a plain `Error` instead, which is
+    // recognised by its code (see `TRANSPORT_CODES`).
     if (err instanceof TypeError) return true;
-    return causeCodes(err).codes.some((code) => TRANSPORT_CODES.has(code));
+    return hasTransportCode(err);
 };
 
 // Could the first attempt already have taken effect on the server?
@@ -174,11 +188,13 @@ export const isRetryable = (err: unknown): boolean => {
 //
 // Everything else is ambiguous. A 5xx proves the server did process the
 // request. A reset or a broken pipe can arrive after it was fully sent and
-// acted on. A routing errno can be delivered on an established socket. A
-// deadline says nothing at all about the server's state. So every errno in the
-// cause chain must be a connect errno: one other errno anywhere in the chain
-// is doubt, and doubt is not replayed. A chain longer than the walk is doubt
-// too: the links below the limit were never read.
+// acted on. A routing errno can be delivered on an established socket. Bun's
+// `ConnectionRefused` and `FailedToOpenSocket` can come from a new connection
+// bun opened to resend a request that already went out on a connection it kept
+// open. A deadline says nothing at all about the server's state. So every code in the
+// cause chain must be one of `CONNECT_CODES`: one other code anywhere in the
+// chain is doubt, and doubt is not replayed. A chain longer than the walk is
+// doubt too: the links below the limit were never read.
 export const isSafeToReplay = (err: unknown): boolean => {
     const { codes, complete } = causeCodes(err);
     return (

test/cli/backup-concurrency.test.ts

@@ -0,0 +1,608 @@
+/**
+ * How many originals `lib.backup()` works on at once.
+ *
+ * The backup puts up to the content pool's limit of originals at their save
+ * paths at the same time: 5 for a library opened with the default
+ * `RequestPools`, as `quak backup` opens it, or the `contentConcurrency` of the
+ * pools a library was opened with. Everything the backup does for one original
+ * counts against that limit, hashing it for `verify` included, so however large
+ * the account is, the backup has no more than that many originals open at once.
+ *
+ * As in `backup-end-to-end.test.ts`, the real `Client` and `Library` run
+ * against `buildMetaFetch` from `test/ente-server.ts`, a fake of the Ente HTTP
+ * API serving data this file encrypts with libsodium. Here a test can also hold
+ * each original's download until it lets it go, to see how many the backup has
+ * in flight at once.
+ *
+ * The account, user 42, has one album of its own, Holiday, of eight photos,
+ * all taken at noon on 1 March 2024, each with an original of its own.
+ */
+
+import {
+    existsSync,
+    lstatSync,
+    mkdtempSync,
+    readdirSync,
+    readFileSync,
+    readlinkSync,
+    rmSync,
+    writeFileSync,
+} from "node:fs";
+import { tmpdir } from "node:os";
+import { join } from "node:path";
+import sodium from "libsodium-wrappers-sumo";
+import { afterAll, beforeAll, describe, expect, it, vi } from "vitest";
+
+import type { BackupOptions } from "../../src/backup.js";
+import { Client, type ClientSnapshot } from "../../src/client.js";
+import { init, toBase64 } from "../../src/crypto/index.js";
+import { Library } from "../../src/library/index.js";
+import { RequestPools } from "../../src/library/pools.js";
+import {
+    buildAccount,
+    buildMetaFetch,
+    encryptSecretbox,
+    encryptStreamBlob,
+} from "../ente-server.js";
+import { blake2b } from "../live-photo.js";
+
+const EMAIL = "backup@example.com";
+const PASSWORD = "backup-password";
+// The fake server logs every account in as user 42.
+const USER_ID = 42;
+
+// Noon on 1 March 2024 in the time zone of the machine running the test, in
+// microseconds, as Ente records times.
+const TAKEN = new Date(2024, 2, 1, 12).getTime() * 1000;
+
+// The eight photos' IDs. Photo `id` is titled `photo-<id>.jpg`, and its
+// original is a few bytes naming it.
+const PHOTO_IDS = [101, 102, 103, 104, 105, 106, 107, 108];
+const originalOf = (id: number): Uint8Array =>
+    new TextEncoder().encode(`the photo ${id}`);
+
+// Where a backup to `dir` saves photo `id`'s original.
+const savedAt = (dir: string, id: number): string =>
+    join(dir, "2024/2024-03/2024-03-01", `2024-03-01.${id}.jpg`);
+
+// The fake server, the session it logs the account in with, and the
+// directory every backup below is made in.
+let server: typeof globalThis.fetch;
+let session: ClientSnapshot;
+let root: string;
+
+// The fake server, holding the account encrypted as Ente's clients upload it.
+const buildServer = async () => {
+    const account = await buildAccount(PASSWORD);
+
+    // The album as `/collections/v2` lists it: its key encrypted with the
+    // account's master key, its name with its own key.
+    const albumKey = sodium.crypto_secretbox_keygen();
+    const encryptedAlbumKey = encryptSecretbox(albumKey, account.masterKey);
+    const encryptedName = encryptSecretbox(
+        new TextEncoder().encode("Holiday"),
+        albumKey,
+    );
+    const album = {
+        id: 10,
+        owner: { id: USER_ID },
+        encryptedKey: toBase64(encryptedAlbumKey.ciphertext),
+        keyDecryptionNonce: toBase64(encryptedAlbumKey.nonce),
+        encryptedName: toBase64(encryptedName.ciphertext),
+        nameDecryptionNonce: toBase64(encryptedName.nonce),
+        type: "album",
+        updationTime: TAKEN,
+    };
+
+    // Each photo as the album's `/collections/v2/diff` lists it: its key
+    // encrypted with the album's key, and its metadata, holding the hash of
+    // its original, encrypted with its own key. Its original, encrypted with
+    // its own key too, is what the files CDN serves for it.
+    const ciphertexts: Record<number, Uint8Array> = {};
+    const photos = PHOTO_IDS.map((id) => {
+        const key = sodium.crypto_secretstream_xchacha20poly1305_keygen();
+        const original = encryptStreamBlob(originalOf(id), key);
+        ciphertexts[id] = original.ciphertext;
+        const encryptedKey = encryptSecretbox(key, albumKey);
+        const metadata = encryptStreamBlob(
+            new TextEncoder().encode(
+                JSON.stringify({
+                    title: `photo-${id}.jpg`,
+                    fileType: 0,
+                    hash: blake2b(originalOf(id)),
+                    creationTime: TAKEN,
+                    modificationTime: TAKEN,
+                }),
+            ),
+            key,
+        );
+        return {
+            id,
+            collectionID: 10,
+            ownerID: USER_ID,
+            encryptedKey: toBase64(encryptedKey.ciphertext),
+            keyDecryptionNonce: toBase64(encryptedKey.nonce),
+            metadata: {
+                encryptedData: toBase64(metadata.ciphertext),
+                decryptionHeader: toBase64(metadata.header),
+            },
+            file: { decryptionHeader: toBase64(original.header) },
+            thumbnail: {
+                decryptionHeader: toBase64(sodium.randombytes_buf(24)),
+            },
+            updationTime: TAKEN,
+        };
+    });
+
+    return buildMetaFetch({
+        verifier: account.verifier,
+        srpAttributes: account.srpAttributes,
+        keyAttributes: account.keyAttributes,
+        encryptedToken: account.encryptedToken,
+        collections: [album],
+        filesByCollection: { 10: photos },
+        encryptedMLData: {},
+        fileCiphertexts: ciphertexts,
+    });
+};
+
+beforeAll(async () => {
+    await init();
+    await sodium.ready;
+    root = mkdtempSync(join(tmpdir(), "quak-backup-concurrency-"));
+    server = await buildServer();
+    const client = await Client.login({
+        email: EMAIL,
+        password: PASSWORD,
+        apiOptions: { fetch: server },
+    });
+    session = client.toJSON();
+});
+
+afterAll(() => {
+    rmSync(root, { recursive: true, force: true });
+});
+
+// The ID of the photo whose original a request downloads, or undefined for any
+// other request. quak downloads an original from
+// `https://files.ente.io/?fileID=<id>`.
+const downloadOf = (input: RequestInfo | URL): number | undefined => {
+    const url = new URL(String(input));
+    return url.hostname === "files.ente.io"
+        ? Number(url.searchParams.get("fileID"))
+        : undefined;
+};
+
+// The fake server, except that it answers 404 to every download of photo
+// `failing`'s original.
+const failingFor =
+    (failing: number): typeof globalThis.fetch =>
+    async (input, init) =>
+        downloadOf(input) === failing
+            ? new Response("not found", { status: 404 })
+            : server(input, init);
+
+// The fake server, holding each download of an original until the test lets
+// it go. `held` lists the downloads waiting, the first to arrive first;
+// `most()` is the largest number that ever waited at once. A held download of
+// photo `failing`'s original is answered 404 once it is let go.
+const holdingServer = (failing?: number) => {
+    const held: { id: number; release: () => void }[] = [];
+    let most = 0;
+    const fetch: typeof globalThis.fetch = async (input, init) => {
+        const id = downloadOf(input);
+        if (id !== undefined) {
+            await new Promise<void>((release) => {
+                held.push({ id, release });
+                most = Math.max(most, held.length);
+            });
+            if (id === failing) {
+                return new Response("not found", { status: 404 });
+            }
+        }
+        return server(input, init);
+    };
+    return { fetch, held, most: () => most };
+};
+
+// Let the held downloads go one at a time, the first to arrive first, each
+// once `limit` of them wait, or all the downloads still to come when there
+// are fewer. Each of the eight photos is downloaded once. With a backup that
+// downloads fewer than `limit` at once, this fails waiting for them. Each wait
+// allows five seconds, for a loaded machine's disk to sync the files written.
+const releaseOneAtATime = async (
+    held: { release: () => void }[],
+    limit: number,
+): Promise<void> => {
+    for (let left = PHOTO_IDS.length; left > 0; left--) {
+        await vi.waitFor(
+            () => expect(held).toHaveLength(Math.min(limit, left)),
+            { interval: 5, timeout: 5000 },
+        );
+        held.shift()!.release();
+    }
+};
+
+// A library of the account opened as `quak backup` opens one, with no
+// precache and no background refresh, sending its requests through `fetch`.
+// It backs up to `dir` and keeps its cache beside it. Its request pools are
+// `pools`, or by default the library's own.
+const openLibrary = (
+    dir: string,
+    fetch: typeof globalThis.fetch,
+    pools?: RequestPools,
+): Promise<Library> =>
+    Library.open({
+        client: Client.fromJSON(session, { fetch }),
+        downloadDirectory: dir,
+        cacheDirectory: `${dir}-cache`,
+        refreshIntervalSeconds: Infinity,
+        precacheThumbnails: false,
+        precacheOriginals: false,
+        pools,
+    });
+
+// Every entry under `dir`, relative to it, sorted, a symlink as
+// `name -> target`.
+const listTree = (dir: string): string[] =>
+    readdirSync(dir, { recursive: true })
+        .sort()
+        .map((name) => {
+            const path = join(dir, name);
+            return lstatSync(path).isSymbolicLink()
+                ? `${name} -> ${readlinkSync(path)}`
+                : name;
+        });
+
+const expectStored = (dir: string, ids: number[]): void => {
+    for (const id of ids) {
+        expect(readFileSync(savedAt(dir, id))).toEqual(
+            Buffer.from(originalOf(id)),
+        );
+    }
+};
+
+describe("lib.backup()", () => {
+    // A library opened with the default pools, as `quak backup` opens one,
+    // downloads up to five originals at once. As soon as one finishes, the
+    // backup starts on the next photo, so five stay in flight until fewer
+    // than five are left.
+    it("downloads five originals at once, and never more, with the default pools", async () => {
+        const dir = mkdtempSync(join(root, "backup-"));
+        const holding = holdingServer();
+        const lib = await openLibrary(dir, holding.fetch);
+        try {
+            const backup = lib.backup();
+            await releaseOneAtATime(holding.held, 5);
+            const result = await backup;
+
+            expect(holding.most()).toBe(5);
+            expect(result).toMatchObject({
+                totalFiles: 8,
+                downloaded: 8,
+                failed: 0,
+            });
+            expectStored(dir, PHOTO_IDS);
+        } finally {
+            await lib.close();
+        }
+    });
+
+    // The limit is the content pool's, so a library opened with pools of
+    // another `contentConcurrency` backs up that many at once.
+    it("downloads as many originals at once as the content pool of the pools it was opened with allows", async () => {
+        const dir = mkdtempSync(join(root, "backup-"));
+        const holding = holdingServer();
+        const lib = await openLibrary(
+            dir,
+            holding.fetch,
+            new RequestPools({ contentConcurrency: 2 }),
+        );
+        try {
+            const backup = lib.backup();
+            await releaseOneAtATime(holding.held, 2);
+            const result = await backup;
+
+            expect(holding.most()).toBe(2);
+            expect(result).toMatchObject({ downloaded: 8, failed: 0 });
+            expectStored(dir, PHOTO_IDS);
+        } finally {
+            await lib.close();
+        }
+    });
+
+    // With `verify`, hashing an original already stored is part of the work
+    // on it, and counts against the limit too: a large backup never has more
+    // than five originals open for hashing. Here every stored original has
+    // changed on disk, so each is hashed, found not to match, and downloaded
+    // again. While five downloads wait, the backup has hashed only those five
+    // photos' originals; it opens the next one once a download finishes.
+    it("with verify, hashes no more originals ahead of their downloads than the limit", async () => {
+        const dir = mkdtempSync(join(root, "backup-"));
+        const first = await openLibrary(dir, server);
+        try {
+            await first.backup();
+        } finally {
+            await first.close();
+        }
+        for (const id of PHOTO_IDS) {
+            writeFileSync(savedAt(dir, id), "not the photo any more");
+        }
+
+        const holding = holdingServer();
+        const lib = await openLibrary(dir, holding.fetch);
+        const lines: string[] = [];
+        const mismatches = () =>
+            lines.filter((line) => line.startsWith("MISMATCH"));
+        try {
+            const backup = lib.backup({
+                verify: true,
+                onProgress: (line) => lines.push(line),
+            });
+            await vi.waitFor(() => expect(holding.held).toHaveLength(5), {
+                interval: 5,
+                timeout: 5000,
+            });
+            expect(mismatches()).toHaveLength(5);
+
+            await releaseOneAtATime(holding.held, 5);
+            const result = await backup;
+
+            expect(holding.most()).toBe(5);
+            expect(mismatches()).toHaveLength(8);
+            expect(result).toMatchObject({
+                downloaded: 8,
+                verified: 0,
+                mismatched: 8,
+                failed: 0,
+            });
+            expectStored(dir, PHOTO_IDS);
+        } finally {
+            await lib.close();
+        }
+    });
+
+    // The server answers 404 for photo 103's original. A 404 is not retried,
+    // so that download fails at once, while other downloads are still in
+    // flight. The backup records the failure and goes on: the downloads
+    // beside it finish, the photos after it are downloaded, and only photo
+    // 103 is missing from the backup and in failures.json.
+    it("goes on with the other originals when one fails", async () => {
+        const dir = mkdtempSync(join(root, "backup-"));
+        const holding = holdingServer(103);
+        const lib = await openLibrary(dir, holding.fetch);
+        try {
+            const backup = lib.backup();
+            await releaseOneAtATime(holding.held, 5);
+            const result = await backup;
+
+            expect(result).toMatchObject({
+                totalFiles: 8,
+                downloaded: 7,
+                failed: 1,
+            });
+            expect(result.errors.map((e) => e.fileID)).toEqual([103]);
+            expectStored(
+                dir,
+                PHOTO_IDS.filter((id) => id !== 103),
+            );
+            expect(existsSync(savedAt(dir, 103))).toBe(false);
+            const failures = JSON.parse(
+                readFileSync(join(dir, "failures.json"), "utf-8"),
+            );
+            expect(Object.keys(failures.files)).toEqual(["103"]);
+        } finally {
+            await lib.close();
+        }
+    });
+
+    // A throw from `onProgress` fails the backup: `lib.backup()` rejects with
+    // it. From then on the backup starts no further original. It finishes the
+    // ones it is working on, and rejects only once they are in place, so
+    // nothing is written into the backup after it has ended and removed its
+    // lock. Here the first five downloads wait. Photo 103's is let go first
+    // and answered 404, and the line reporting that failure throws while the
+    // other four wait. Only those five downloads are held: a backup that went
+    // on would store the other three photos.
+    it("when onProgress throws, starts no further original, and rejects once the originals it was working on are in place", async () => {
+        const dir = mkdtempSync(join(root, "backup-"));
+        const downloads: number[] = [];
+        const held: { id: number; release: () => void }[] = [];
+        const fetch: typeof globalThis.fetch = async (input, init) => {
+            const id = downloadOf(input);
+            if (id !== undefined) {
+                downloads.push(id);
+                if (downloads.length <= 5) {
+                    await new Promise<void>((release) =>
+                        held.push({ id, release }),
+                    );
+                }
+                if (id === 103) {
+                    return new Response("not found", { status: 404 });
+                }
+            }
+            return server(input, init);
+        };
+        const lib = await openLibrary(dir, fetch);
+        const lines: string[] = [];
+        try {
+            const backup = lib.backup({
+                onProgress: (line) => {
+                    lines.push(line);
+                    if (line.startsWith("FAILED")) {
+                        throw new Error("the progress output is closed");
+                    }
+                },
+            });
+            // The photos whose originals are in place, and whether the lock
+            // is there, at the moment `lib.backup()` rejects.
+            let storedWhenRejected: number[] = [];
+            let lockedWhenRejected = true;
+            const rejection = backup.catch((err: unknown) => {
+                storedWhenRejected = PHOTO_IDS.filter((id) =>
+                    existsSync(savedAt(dir, id)),
+                );
+                lockedWhenRejected = existsSync(join(dir, "backup.lock"));
+                return err;
+            });
+
+            await vi.waitFor(() => expect(held).toHaveLength(5), {
+                interval: 5,
+                timeout: 5000,
+            });
+            const at103 = held.findIndex((h) => h.id === 103);
+            held.splice(at103, 1)[0]!.release();
+            await vi.waitFor(
+                () =>
+                    expect(
+                        lines.some((line) => line.startsWith("FAILED")),
+                    ).toBe(true),
+                { interval: 5, timeout: 5000 },
+            );
+            for (const download of held) download.release();
+
+            expect(await rejection).toEqual(
+                new Error("the progress output is closed"),
+            );
+            expect(downloads.sort()).toEqual([101, 102, 103, 104, 105]);
+            expect(storedWhenRejected).toEqual([101, 102, 104, 105]);
+            expect(lockedWhenRejected).toBe(false);
+        } finally {
+            await lib.close();
+        }
+    });
+
+    // Once the backup's lock is lost, another backup may be writing to the
+    // directory, so the backup starts no further original. It finishes the
+    // ones it is working on, and `lib.backup()` rejects once they are all in
+    // place. Here `backup.lock` is removed while the first five downloads
+    // wait, and proper-lockfile's timer finds it gone. As in
+    // `backup-end-to-end.test.ts`, only `setTimeout` and `clearTimeout` are
+    // faked, so the test can fire that timer without waiting 5 seconds. Only
+    // the first five downloads are held: a backup that went on would store
+    // the other three photos.
+    it("when its lock is lost, starts no further original, and rejects once the originals it was working on are in place", async () => {
+        const dir = mkdtempSync(join(root, "backup-"));
+        const lockPath = join(dir, "backup.lock");
+        const downloads: number[] = [];
+        const held: (() => void)[] = [];
+        const fetch: typeof globalThis.fetch = async (input, init) => {
+            const id = downloadOf(input);
+            if (id !== undefined) {
+                downloads.push(id);
+                if (downloads.length <= 5) {
+                    await new Promise<void>((release) => held.push(release));
+                }
+            }
+            return server(input, init);
+        };
+        const lib = await openLibrary(dir, fetch);
+        vi.useFakeTimers({ toFake: ["setTimeout", "clearTimeout"] });
+        try {
+            const backup = lib.backup();
+            // The photos whose originals are in place at the moment
+            // `lib.backup()` rejects.
+            let storedWhenRejected: number[] = [];
+            const rejection = backup.catch((err: unknown) => {
+                storedWhenRejected = PHOTO_IDS.filter((id) =>
+                    existsSync(savedAt(dir, id)),
+                );
+                return err;
+            });
+
+            await vi.waitFor(() => expect(held).toHaveLength(5), {
+                interval: 5,
+                timeout: 5000,
+            });
+            rmSync(lockPath, { recursive: true });
+            // Fire proper-lockfile's timer, and wait for the look at the lock
+            // on disk that it starts to finish: Node lists a file system call
+            // in progress as an `FSReqCallback`.
+            vi.advanceTimersByTime(5_000);
+            while (process.getActiveResourcesInfo().includes("FSReqCallback")) {
+                await new Promise((resolve) => setImmediate(resolve));
+            }
+            for (const release of held) release();
+
+            expect(await rejection).toEqual(
+                new Error(
+                    `lost the lock on ${dir}: ENOENT: no such file or directory, stat '${lockPath}'`,
+                ),
+            );
+            expect(downloads.sort()).toEqual([101, 102, 103, 104, 105]);
+            expect(storedWhenRejected).toEqual([101, 102, 103, 104, 105]);
+        } finally {
+            vi.useRealTimers();
+            await lib.close();
+        }
+    });
+
+    // Working on several originals at once changes only how long a backup
+    // takes. The same two runs, with a content pool of one, which backs up
+    // one original at a time, and with the default pools, end the same way:
+    // the same result, the same progress lines (in another order), the same
+    // failures.json, and the same tree. The first run stores every photo but
+    // 103, whose download is answered 404. Then two stored originals change
+    // on disk, and a run with `verify` puts them back and tries 103 again.
+    it("ends as a backup of one original at a time does", async () => {
+        const twoRuns = async (pools: RequestPools) => {
+            const dir = mkdtempSync(join(root, "backup-"));
+            const lines: string[] = [];
+            const backup = async (opts: BackupOptions) => {
+                const lib = await openLibrary(dir, failingFor(103), pools);
+                try {
+                    return await lib.backup({
+                        ...opts,
+                        onProgress: (line) => lines.push(line),
+                    });
+                } finally {
+                    await lib.close();
+                }
+            };
+            const first = await backup({});
+            writeFileSync(savedAt(dir, 101), "not the photo any more");
+            writeFileSync(savedAt(dir, 105), "not the photo any more");
+            const second = await backup({ verify: true });
+
+            // Each entry of failures.json without the time it was last
+            // tried, which differs from run to run.
+            const { files } = JSON.parse(
+                readFileSync(join(dir, "failures.json"), "utf-8"),
+            ) as { files: Record<string, { lastTriedAt?: number }> };
+            for (const entry of Object.values(files)) delete entry.lastTriedAt;
+
+            return {
+                first,
+                second,
+                lines: lines.sort(),
+                failures: files,
+                tree: listTree(dir),
+            };
+        };
+
+        const oneAtATime = await twoRuns(
+            new RequestPools({ contentConcurrency: 1 }),
+        );
+        const fiveAtATime = await twoRuns(new RequestPools());
+
+        expect(oneAtATime.first).toMatchObject({
+            totalFiles: 8,
+            downloaded: 7,
+            skipped: 0,
+            failed: 1,
+        });
+        expect(oneAtATime.second).toMatchObject({
+            totalFiles: 8,
+            downloaded: 2,
+            skipped: 5,
+            verified: 5,
+            mismatched: 2,
+            unchecked: 0,
+            failed: 1,
+        });
+        expect(oneAtATime.failures).toEqual({
+            "103": expect.objectContaining({ fileID: 103, attempts: 2 }),
+        });
+        expect(fiveAtATime).toEqual(oneAtATime);
+    });
+});

test/cli/backup-end-to-end.test.ts

@@ -9,7 +9,8 @@
  * data this file encrypts with libsodium, as Ente's clients encrypt it. So
  * quak's own code fetches each original, decrypts it, checks it against the
  * hash its uploader recorded and unpacks a live photo, and quak's own retries
- * repeat a failed request.
+ * repeat a failed request. Three tests run `lib.backup()` instead, as a program
+ * using quak as a library does.
  *
  * The account, user 42, has three albums:
  *
@@ -24,6 +25,7 @@
 import {
     existsSync,
     lstatSync,
+    mkdirSync,
     mkdtempSync,
     readdirSync,
     readFileSync,
@@ -36,7 +38,16 @@ import { join } from "node:path";
 import { Writable } from "node:stream";
 import { gzipSync } from "node:zlib";
 import sodium from "libsodium-wrappers-sumo";
-import { afterAll, beforeAll, describe, expect, it } from "vitest";
+import {
+    afterAll,
+    afterEach,
+    beforeAll,
+    beforeEach,
+    describe,
+    expect,
+    it,
+    vi,
+} from "vitest";
 
 import {
     backupCommand,
@@ -47,6 +58,7 @@ import { run } from "../../src/cli-run.js";
 import { loadSession } from "../../src/cli-session.js";
 import { Client } from "../../src/client.js";
 import { init, toBase64 } from "../../src/crypto/index.js";
+import { Library, lockBackupDirectory } from "../../src/library/index.js";
 import { UNATTENDED_RETRY_OPTIONS } from "../../src/retry.js";
 import {
     buildAccount,
@@ -262,15 +274,17 @@ afterAll(() => {
 
 // Run `quak backup <dir>` as `bin/quak.ts` runs it, sending its requests
 // through `fetch`, and return its exit code and everything it printed on
-// stderr. Its requests are retried as those of `quak backup` are, except that
-// a retry does not wait. Its cache is beside `dir`, so each run of one
-// directory finds the cache the run before it left, as cron's runs find the
-// user's.
+// stderr and on stdout. Its requests are retried as those of `quak backup`
+// are, except that a retry does not wait. A download is given up after
+// `downloadTimeoutMs` with no bytes received, 60 seconds unless given. Its
+// cache is beside `dir`, so each run of one directory finds the cache the run
+// before it left, as cron's runs find the user's.
 const backup = async (
     dir: string,
-    opts: { verify?: boolean } = {},
+    opts: { verify?: boolean; json?: boolean } = {},
     fetch = server,
-): Promise<{ code: number; stderr: string }> => {
+    downloadTimeoutMs?: number,
+): Promise<{ code: number; stderr: string; stdout: string }> => {
     let printed = "";
     const stderr = new Writable({
         write(chunk: Buffer, _encoding, done) {
@@ -278,8 +292,10 @@ const backup = async (
             done();
         },
     });
+    let output = "";
     const stdout = new Writable({
-        write(_chunk, _encoding, done) {
+        write(chunk: Buffer, _encoding, done) {
+            output += chunk.toString();
             done();
         },
     });
@@ -296,6 +312,7 @@ const backup = async (
             loadSession(path, {
                 fetch,
                 retry: { ...UNATTENDED_RETRY_OPTIONS, sleep: async () => {} },
+                downloadTimeoutMs,
             }),
         login: prompted,
         prompt: prompted,
@@ -304,7 +321,7 @@ const backup = async (
     const code = await new Promise<number>((resolve) => {
         void run(backupCommand(ctx, dir, opts), stdout, stderr, resolve);
     });
-    return { code, stderr: printed };
+    return { code, stderr: printed, stdout: output };
 };
 
 // The ID of the file whose original a request downloads, or undefined for any
@@ -317,6 +334,41 @@ const downloadOf = (input: RequestInfo | URL): number | undefined => {
         : undefined;
 };
 
+// The error Node gives for a connection to `address` that was refused.
+const refusedAt = (address: string): Error =>
+    Object.assign(new Error(`connect ECONNREFUSED ${address}`), {
+        code: "ECONNREFUSED",
+    });
+
+// What Node's `fetch` rejects with when the server refuses the connection, as
+// for every failure below HTTP: a TypeError that says only "fetch failed",
+// with the reason, and its errno, as its cause. When the server has several
+// addresses and each one refuses, the cause is an AggregateError with an empty
+// message, one error per address and the errno of the first. `reason` is the
+// error as quak prints it and records it in `failures.json`.
+const REFUSED = [
+    {
+        addresses: "its one address",
+        error: (): TypeError =>
+            new TypeError("fetch failed", { cause: refusedAt("1.2.3.4:443") }),
+        reason: "fetch failed: connect ECONNREFUSED 1.2.3.4:443",
+    },
+    {
+        addresses: "each of its two addresses",
+        error: (): TypeError =>
+            new TypeError("fetch failed", {
+                cause: Object.assign(
+                    new AggregateError([
+                        refusedAt("2001:db8::1:443"),
+                        refusedAt("1.2.3.4:443"),
+                    ]),
+                    { code: "ECONNREFUSED" },
+                ),
+            }),
+        reason: "fetch failed: connect ECONNREFUSED 2001:db8::1:443, connect ECONNREFUSED 1.2.3.4:443",
+    },
+];
+
 // The fake server, pushing onto `downloads` the ID of each file whose original
 // is downloaded.
 const recordingDownloads =
@@ -580,6 +632,40 @@ describe("quak backup", () => {
         ).toEqual([]);
     });
 
+    // The user has put a file in collections/ with the name of the Trips
+    // folder, so quak cannot make that folder. It says so on one line, writes
+    // the two albums after it, and exits 1, and --json lists the album's ID
+    // in `failedAlbums`. No file failed, so there is no failures.json.
+    it("exits 1 when an album's folder cannot be made, writes the other albums, and lists it in --json", async () => {
+        const dir = mkdtempSync(join(root, "backup-"));
+        mkdirSync(join(dir, "collections"));
+        writeFileSync(join(dir, "collections", "Trips"), "mine");
+
+        const first = await backup(dir);
+        const { code, stdout } = await backup(dir, { json: true });
+
+        expect(first.code).toBe(1);
+        expect(first.stderr).toMatch(/^FAILED album Trips: EEXIST: /m);
+        expect(code).toBe(1);
+        expect(JSON.parse(stdout)).toMatchObject({
+            failed: 0,
+            undecryptableAlbums: [],
+            failedAlbums: [10],
+        });
+        const collections = join(dir, "collections");
+        expect(readJSON(join(collections, "Family.json")).id).toBe(20);
+        expect(readdirSync(join(collections, "Family")).sort()).toEqual([
+            "IMG_0002.heic",
+            "IMG_0002.mov",
+            "beach.jpg",
+        ]);
+        expect(readJSON(join(collections, "Friend's Wedding.json")).id).toBe(
+            30,
+        );
+        expect(existsSync(join(collections, "Trips.json"))).toBe(false);
+        expect(existsSync(join(dir, "failures.json"))).toBe(false);
+    });
+
     // The server no longer accepts the saved session's token, as after
     // `quak logout` elsewhere, and answers every request with 401. The run
     // stops at the refresh it starts with, says to log in again, and exits 3.
@@ -607,6 +693,317 @@ describe("quak backup", () => {
         expect(readdirSync(dir)).toEqual([]);
     });
 
+    // The server refuses every connection, at its one address or at each of
+    // several. quak tries each request again, here without waiting, and then
+    // the run stops at the refresh it starts with. The line it prints gives
+    // the reason Node's `fetch` put on the error's cause, so a cron log tells
+    // a refused connection from a name that did not resolve, and the exit
+    // code is 1.
+    it.each(REFUSED)(
+        "prints why the refresh could not reach the server at $addresses, and exits 1",
+        async ({ error, reason }) => {
+            const dir = mkdtempSync(join(root, "backup-"));
+            const fetch: typeof globalThis.fetch = async () => {
+                throw error();
+            };
+
+            const { code, stderr } = await backup(dir, {}, fetch);
+
+            expect(code).toBe(1);
+            expect(stderr).toBe(
+                "Starting backup...\n" +
+                    "Refreshing library...\n" +
+                    `quak: ${reason}\n`,
+            );
+        },
+    );
+
+    // The files CDN refuses every connection for toast.mp4, at its one
+    // address or at each of several. Once its retries are spent, the file
+    // goes into `failures.json` with the reason, as a transient failure,
+    // since a later run may well reach the CDN, and the lines the backup
+    // prints for it give the same reason. The rest of the account is backed
+    // up, and the exit code is 1.
+    it.each(REFUSED)(
+        "records why a download could not reach the server at $addresses in failures.json, as transient, and exits 1",
+        async ({ error, reason }) => {
+            let attempts = 0;
+            const fetch: typeof globalThis.fetch = async (input, init) => {
+                if (downloadOf(input) === 200) {
+                    attempts++;
+                    throw error();
+                }
+                return server(input, init);
+            };
+            const dir = mkdtempSync(join(root, "backup-"));
+
+            const { code, stderr } = await backup(dir, {}, fetch);
+
+            expect(code).toBe(1);
+            expect(attempts).toBe(UNATTENDED_RETRY_OPTIONS.attempts);
+            const failures = readJSON(join(dir, "failures.json"));
+            expect(Object.keys(failures.files)).toEqual(["200"]);
+            expect(failures.files["200"]).toMatchObject({
+                fileID: 200,
+                title: "toast.mp4",
+                classification: "transient",
+                attempts: 1,
+                error: reason,
+            });
+            expect(stderr).toContain(`FAILED original toast.mp4: ${reason}\n`);
+            expect(stderr).toContain(
+                `[Friend's Wedding] toast.mp4 (id 200): ${reason}\n`,
+            );
+            expect(readFileSync(join(dir, DAY, "2024-03-01.100.jpg"))).toEqual(
+                Buffer.from(JPEG_WITH_EXIF),
+            );
+            expect(existsSync(join(dir, DAY, "2024-03-01.200.mp4"))).toBe(
+                false,
+            );
+        },
+    );
+
+    // The download of toast.mp4 fails as bun's `fetch` reports a failure, in
+    // the binary `make install` builds: a plain `Error` with bun's own code
+    // and message, and no cause. These are what bun gave for a refused
+    // connection, a name that did not resolve and a reset connection. The
+    // file is tried as often as under Node, and goes into `failures.json`
+    // with that message, as a transient failure.
+    it.each([
+        [
+            "ConnectionRefused",
+            "Unable to connect. Is the computer able to access the url?",
+        ],
+        ["FailedToOpenSocket", "Was there a typo in the url or port?"],
+        [
+            "ECONNRESET",
+            "The socket connection was closed unexpectedly. For more information, pass `verbose: true` in the second argument to fetch()",
+        ],
+    ])(
+        "records a download that failed with bun's %s in failures.json, as transient",
+        async (bunCode, message) => {
+            let attempts = 0;
+            const fetch: typeof globalThis.fetch = async (input, init) => {
+                if (downloadOf(input) === 200) {
+                    attempts++;
+                    throw Object.assign(new Error(message), {
+                        code: bunCode,
+                        errno: 0,
+                    });
+                }
+                return server(input, init);
+            };
+            const dir = mkdtempSync(join(root, "backup-"));
+
+            const { code } = await backup(dir, {}, fetch);
+
+            expect(code).toBe(1);
+            expect(attempts).toBe(UNATTENDED_RETRY_OPTIONS.attempts);
+            const failures = readJSON(join(dir, "failures.json"));
+            expect(failures.files["200"]).toMatchObject({
+                classification: "transient",
+                error: message,
+            });
+        },
+    );
+
+    // The download of toast.mp4 stalls: the server sends the headers and then
+    // never a byte of the body. Each attempt is given up after 50 ms, quak's
+    // own deadline for this run, with a `TimeoutError` whose message says
+    // only that the download stalled. The file goes into `failures.json` as a
+    // transient failure, as a request that ran out of time is retried.
+    it("records a stalled download in failures.json, as transient", async () => {
+        let attempts = 0;
+        const fetch: typeof globalThis.fetch = async (input, init) => {
+            if (downloadOf(input) === 200) {
+                attempts++;
+                return new Response(
+                    new ReadableStream<Uint8Array>({
+                        pull: () => new Promise<void>(() => {}),
+                    }),
+                );
+            }
+            return server(input, init);
+        };
+        const dir = mkdtempSync(join(root, "backup-"));
+
+        const { code } = await backup(dir, {}, fetch, 50);
+
+        expect(code).toBe(1);
+        expect(attempts).toBe(UNATTENDED_RETRY_OPTIONS.attempts);
+        const failures = readJSON(join(dir, "failures.json"));
+        expect(failures.files["200"]).toMatchObject({
+            classification: "transient",
+            error: "download stalled: no bytes received for 50 ms",
+        });
+    });
+
+    // The name of Ente's API server does not resolve when quak fetches the ML
+    // data, after every original has been backed up. No file has ML data
+    // cached yet, so each goes into `failures.json` with the reason, as a
+    // transient failure, as a download that could not reach the server does,
+    // and the exit code is 1.
+    it("records why the ML data fetch could not reach the server in failures.json, as transient, and exits 1", async () => {
+        const fetch: typeof globalThis.fetch = async (input, init) => {
+            if (new URL(String(input)).pathname === "/files/data/fetch") {
+                throw new TypeError("fetch failed", {
+                    cause: Object.assign(
+                        new Error("getaddrinfo ENOTFOUND api.ente.io"),
+                        { code: "ENOTFOUND" },
+                    ),
+                });
+            }
+            return server(input, init);
+        };
+        const dir = mkdtempSync(join(root, "backup-"));
+
+        const { code, stderr } = await backup(dir, {}, fetch);
+
+        expect(code).toBe(1);
+        const error =
+            "ML data: fetch failed: getaddrinfo ENOTFOUND api.ente.io";
+        expect(stderr).toContain(`FAILED ${error}\n`);
+        const failures = readJSON(join(dir, "failures.json"));
+        expect(
+            Object.values(failures.files).map((f) => [
+                f.fileID,
+                f.classification,
+                f.error,
+            ]),
+        ).toEqual([
+            [100, "transient", error],
+            [200, "transient", error],
+            [300, "transient", error],
+        ]);
+    });
+
+    // The server sends beach.jpg's ML data encrypted under a key that is not
+    // the file's, so quak cannot decrypt it. The file's JSON has the reason
+    // in place of the ML data, and the file goes into `failures.json` with
+    // it, so the exit code is 1. The other two files have no ML data, which
+    // is no failure.
+    it("records a file whose ML data does not decrypt in its JSON and in failures.json, and exits 1", async () => {
+        const otherKey = sodium.crypto_secretstream_xchacha20poly1305_keygen();
+        const mlData = encryptStreamBlob(
+            gzipSync(JSON.stringify(ML_DATA)),
+            otherKey,
+        );
+        const fetch: typeof globalThis.fetch = async (input, init) => {
+            if (new URL(String(input)).pathname !== "/files/data/fetch") {
+                return server(input, init);
+            }
+            const data = [
+                {
+                    fileID: BEACH.id,
+                    encryptedData: toBase64(mlData.ciphertext),
+                    decryptionHeader: toBase64(mlData.header),
+                },
+            ];
+            return new Response(JSON.stringify({ data }), {
+                headers: { "content-type": "application/json" },
+            });
+        };
+        const dir = mkdtempSync(join(root, "backup-"));
+
+        const { code, stderr } = await backup(dir, {}, fetch);
+
+        expect(code).toBe(1);
+        const reason = "secretstream chunk authentication failed";
+        const beachJSON = readJSON(join(dir, DAY, "2024-03-01.100.json"));
+        expect(beachJSON).not.toHaveProperty("mlData");
+        expect(beachJSON.mlDataError).toBe(reason);
+        const failures = readJSON(join(dir, "failures.json"));
+        expect(Object.keys(failures.files)).toEqual(["100"]);
+        expect(failures.files["100"]).toMatchObject({
+            fileID: 100,
+            title: "beach.jpg",
+            error: `ML data: ${reason}`,
+        });
+        expect(stderr).toContain(`beach.jpg (id 100): ML data: ${reason}\n`);
+        for (const fileID of [200, 300]) {
+            const json = readJSON(join(dir, DAY, `2024-03-01.${fileID}.json`));
+            expect(json).not.toHaveProperty("mlData");
+            expect(json).not.toHaveProperty("mlDataError");
+        }
+    });
+
+    // A program using quak as a library backs the account up, and another ML
+    // data fetch, as one the library's refresh starts, fails while the backup
+    // writes the JSON files. The backup's own fetch got beach.jpg's ML data
+    // under a key that is not the file's; the fetch that fails leaves that
+    // reason in place, so beach.jpg's JSON and `failures.json` still give it.
+    it("keeps the reason in beach.jpg's JSON and in failures.json when an ML data fetch fails while lib.backup() writes the JSON files", async () => {
+        const otherKey = sodium.crypto_secretstream_xchacha20poly1305_keygen();
+        const mlData = encryptStreamBlob(
+            gzipSync(JSON.stringify(ML_DATA)),
+            otherKey,
+        );
+        let mlDataFetchFails = false;
+        const fetch: typeof globalThis.fetch = async (input, init) => {
+            if (new URL(String(input)).pathname !== "/files/data/fetch") {
+                return server(input, init);
+            }
+            if (mlDataFetchFails) throw new Error("the connection was reset");
+            const data = [
+                {
+                    fileID: BEACH.id,
+                    encryptedData: toBase64(mlData.ciphertext),
+                    decryptionHeader: toBase64(mlData.header),
+                },
+            ];
+            return new Response(JSON.stringify({ data }), {
+                headers: { "content-type": "application/json" },
+            });
+        };
+        const dir = mkdtempSync(join(root, "backup-"));
+        const lib = await Library.open({
+            client: loadSession(join(sessionDir, "session.json"), {
+                fetch,
+                retry: { ...UNATTENDED_RETRY_OPTIONS, sleep: async () => {} },
+            })!,
+            cacheDirectory: `${dir}-cache`,
+            downloadDirectory: dir,
+            refreshIntervalSeconds: Infinity,
+            precacheThumbnails: false,
+            precacheOriginals: false,
+        });
+        try {
+            // The backup reads a file's cached ML data just before it writes
+            // that file's JSON. The first time, a refresh starts an ML data
+            // fetch whose request fails, and the backup goes on once that
+            // fetch has failed.
+            const forFile = lib.mldata.forFile.bind(lib.mldata);
+            vi.spyOn(lib.mldata, "forFile").mockImplementationOnce(
+                async (args) => {
+                    mlDataFetchFails = true;
+                    await lib.fresh();
+                    await vi.waitFor(
+                        () => expect(lib.status().lastMLError).toBeDefined(),
+                        { timeout: 2000, interval: 5 },
+                    );
+                    return forFile(args);
+                },
+            );
+
+            const result = await lib.backup();
+
+            expect(result.failed).toBe(1);
+            const reason = "secretstream chunk authentication failed";
+            const beachJSON = readJSON(join(dir, DAY, "2024-03-01.100.json"));
+            expect(beachJSON).not.toHaveProperty("mlData");
+            expect(beachJSON.mlDataError).toBe(reason);
+            const failures = readJSON(join(dir, "failures.json"));
+            expect(Object.keys(failures.files)).toEqual(["100"]);
+            expect(failures.files["100"]).toMatchObject({
+                fileID: 100,
+                title: "beach.jpg",
+                error: `ML data: ${reason}`,
+            });
+        } finally {
+            await lib.close();
+        }
+    });
+
     // Something changed beach.jpg on disk after it was backed up. A run with
     // --verify hashes every stored original, finds that one no longer matches
     // the hash Ente recorded, says so, and downloads it again; it put back the
@@ -633,3 +1030,191 @@ describe("quak backup", () => {
         expect(stderr).toContain("  Verified:    2\n  Mismatched:  1\n");
     });
 });
+
+// `backup.lock` is a directory that proper-lockfile touches every 5 seconds,
+// from a timer, while the backup runs. Each time, it first looks at the lock
+// on disk. The lock is lost when it is gone, or when another backup has made
+// its own in its place, as one does with a lock nobody touched for 10
+// seconds. Another backup may then be writing to the directory, so the backup
+// stops before it starts another file.
+//
+// The fake server holds one file's download open until the test lets it go,
+// so the lock can be lost while the backup runs: beach.jpg's, the first file
+// backed up, unless the test names another. Only `setTimeout` and
+// `clearTimeout` are faked, so a test can fire proper-lockfile's timer without
+// waiting 5 seconds.
+describe("a backup whose lock is lost", () => {
+    let dir: string;
+    let lockPath: string;
+    let downloads: number[];
+    let heldFile: number;
+    let heldRequested: Promise<void>;
+    let releaseHeld: () => void;
+    let fetch: typeof globalThis.fetch;
+
+    beforeEach(() => {
+        dir = mkdtempSync(join(root, "backup-"));
+        lockPath = join(dir, "backup.lock");
+        downloads = [];
+        heldFile = BEACH.id;
+        let requested!: () => void;
+        heldRequested = new Promise((resolve) => {
+            requested = resolve;
+        });
+        const released = new Promise<void>((resolve) => {
+            releaseHeld = resolve;
+        });
+        const recording = recordingDownloads(downloads);
+        fetch = async (input, init) => {
+            if (downloadOf(input) === heldFile) {
+                requested();
+                await released;
+            }
+            return recording(input, init);
+        };
+        vi.useFakeTimers({ toFake: ["setTimeout", "clearTimeout"] });
+    });
+
+    afterEach(() => {
+        vi.useRealTimers();
+    });
+
+    // Fire proper-lockfile's timer, and wait for the look at the lock on disk
+    // that it starts to finish: Node lists a file system call in progress as
+    // an `FSReqCallback`.
+    const fireLockTimer = async (): Promise<void> => {
+        vi.advanceTimersByTime(5_000);
+        while (process.getActiveResourcesInfo().includes("FSReqCallback")) {
+            await new Promise((resolve) => setImmediate(resolve));
+        }
+    };
+
+    // The backup downloads up to five originals at once, so it has started
+    // all three when the lock is lost, and beach.jpg is stored. The backup
+    // then stops before it fetches ML data. `quak backup` prints why as one
+    // line, with no stack trace, and exits 1.
+    it("stops quak backup before it fetches ML data when backup.lock is removed, and exits 1", async () => {
+        const result = backup(dir, {}, fetch);
+        await heldRequested;
+        rmSync(lockPath, { recursive: true });
+        await fireLockTimer();
+        releaseHeld();
+        const { code, stderr } = await result;
+
+        expect(code).toBe(1);
+        expect(stderr).toBe(
+            "Starting backup...\n" +
+                "Refreshing library...\n" +
+                "Fetching original beach.jpg (100)...\n" +
+                "Fetching original IMG_0002.HEIC (300)...\n" +
+                "Fetching original toast.mp4 (200)...\n" +
+                `quak: lost the lock on ${dir}: ENOENT: no such file or directory, stat '${lockPath}'\n`,
+        );
+        expect(downloads.sort()).toEqual([100, 200, 300]);
+        expect(readFileSync(join(dir, DAY, "2024-03-01.100.jpg"))).toEqual(
+            Buffer.from(JPEG_WITH_EXIF),
+        );
+    });
+
+    // Here the lock is lost while toast.mp4, the last original, downloads.
+    // All three originals are stored, beside what the backup wrote before it
+    // started them, but no file's JSON and no album is written.
+    it("stops quak backup before any file's JSON when the lock is lost during the last original", async () => {
+        heldFile = TOAST_VIDEO.id;
+        const result = backup(dir, {}, fetch);
+        await heldRequested;
+        rmSync(lockPath, { recursive: true });
+        await fireLockTimer();
+        releaseHeld();
+        const { code, stderr } = await result;
+
+        expect(code).toBe(1);
+        expect(stderr).toContain(`quak: lost the lock on ${dir}: `);
+        expect(listTree(dir)).toEqual([
+            "2024",
+            "2024/2024-03",
+            "2024/2024-03/2024-03-01",
+            "2024/2024-03/2024-03-01/2024-03-01.100.jpg",
+            "2024/2024-03/2024-03-01/2024-03-01.200.mp4",
+            "2024/2024-03/2024-03-01/2024-03-01.300.heic",
+            "2024/2024-03/2024-03-01/2024-03-01.300.livephoto.json",
+            "2024/2024-03/2024-03-01/2024-03-01.300.mov",
+            "account.json",
+            "collections",
+        ]);
+    });
+
+    // Nothing touches the lock but quak, which finds it as it left it each
+    // time it looks, so the backup runs to the end and removes its lock.
+    it("lets quak backup run to the end while its lock stays", async () => {
+        const result = backup(dir, {}, fetch);
+        await heldRequested;
+        await fireLockTimer();
+        await fireLockTimer();
+        releaseHeld();
+        const { code } = await result;
+
+        expect(code).toBe(0);
+        expect(downloads.sort()).toEqual([100, 200, 300]);
+        expect(existsSync(lockPath)).toBe(false);
+    });
+
+    // The library a program using quak opens to back `dir` up.
+    const openLibrary = () =>
+        Library.open({
+            client: loadSession(join(sessionDir, "session.json"), { fetch })!,
+            cacheDirectory: `${dir}-cache`,
+            downloadDirectory: dir,
+            refreshIntervalSeconds: Infinity,
+            precacheThumbnails: false,
+            precacheOriginals: false,
+        });
+
+    // A program using quak as a library sees the same: `lib.backup()` rejects
+    // with the reason. Here another backup has made its own lock in place of
+    // quak's, and that lock is left where it is.
+    it("rejects lib.backup() when another backup has taken the lock over", async () => {
+        const lib = await openLibrary();
+        try {
+            const result = lib.backup();
+            await heldRequested;
+            rmSync(lockPath, { recursive: true });
+            mkdirSync(lockPath);
+            await fireLockTimer();
+            releaseHeld();
+
+            await expect(result).rejects.toThrow(
+                `lost the lock on ${dir}: Unable to update lock within the stale threshold`,
+            );
+            expect(downloads.sort()).toEqual([100, 200, 300]);
+            expect(existsSync(lockPath)).toBe(true);
+        } finally {
+            await lib.close();
+        }
+    });
+
+    // A program can take the lock before it opens its library, as `quak
+    // backup` does: `lockBackupDirectory` returns the lock, which the program
+    // passes to `lib.backup()` and releases once the backup ends. The backup
+    // checks it as it checks a lock it took, and the lock's `lost` says why it
+    // was lost.
+    it("rejects lib.backup() given its caller's lock once that lock is lost", async () => {
+        const lock = await lockBackupDirectory(dir);
+        const lib = await openLibrary();
+        try {
+            const result = lib.backup({ lock });
+            await heldRequested;
+            rmSync(lockPath, { recursive: true });
+            await fireLockTimer();
+            releaseHeld();
+
+            const lost = `lost the lock on ${dir}: ENOENT: no such file or directory, stat '${lockPath}'`;
+            await expect(result).rejects.toThrow(lost);
+            expect(lock.lost?.message).toBe(lost);
+            expect(downloads.sort()).toEqual([100, 200, 300]);
+        } finally {
+            await lib.close();
+            await lock.release();
+        }
+    });
+});

test/cli/backup-undecryptable.test.ts

@@ -23,7 +23,13 @@
  * key replaced by bytes no key opens, or its metadata text that is not JSON.
  */
 
-import { existsSync, mkdtempSync, readFileSync, rmSync } from "node:fs";
+import {
+    existsSync,
+    mkdtempSync,
+    readdirSync,
+    readFileSync,
+    rmSync,
+} from "node:fs";
 import { join } from "node:path";
 import { tmpdir } from "node:os";
 import sodium from "libsodium-wrappers-sumo";
@@ -253,6 +259,18 @@ const brokenAccount = (): FakeEnte => {
     return ente;
 };
 
+// Album 1, "Trip", holds file 10, and album 2, "Family", holds file 20. Both
+// files decrypt, but album 2's own key does not, so the album is the only
+// record left out. All of it as of time 1.
+const brokenAlbumAccount = (): FakeEnte => {
+    const ente = new FakeEnte();
+    ente.putAlbum(1, "Trip", 1);
+    ente.putFile(1, 10, 1);
+    ente.putAlbum(2, "Family", 1, true);
+    ente.putFile(2, 20, 1);
+    return ente;
+};
+
 // Collects everything written to it.
 class Output {
     text = "";
@@ -321,6 +339,12 @@ const failures = (): Record<string, { error: string; attempts: number }> => {
 const stored = (id: number): boolean =>
     existsSync(join(backupDir(), DAY, `2026-03-01.${id}.jpg`));
 
+// The IDs of the files that album `name`'s JSON lists.
+const albumFiles = (name: string): number[] =>
+    JSON.parse(
+        readFileSync(join(backupDir(), "collections", `${name}.json`), "utf-8"),
+    ).files.map((f: { id: number }) => f.id);
+
 describe("quak backup with records it cannot decrypt", () => {
     it("backs up every other file, records each file it cannot decrypt in failures.json, names each such album, and exits 1", async () => {
         expect(await backup(brokenAccount())).toBe(1);
@@ -398,6 +422,67 @@ describe("quak backup with records it cannot decrypt", () => {
         expect(stderr.text).not.toContain("FAILED");
     });
 
+    it("exits 1 when the only record it cannot decrypt is an album, and 0 once the album decrypts", async () => {
+        const ente = brokenAlbumAccount();
+
+        // No file is in `failures.json`: the album alone keeps the exit code
+        // at 1.
+        expect(await backup(ente)).toBe(1);
+        expect(failures()).toEqual({});
+
+        ente.putAlbum(2, "Family", 2);
+        expect(await backup(ente)).toBe(0);
+        expect(stored(20)).toBe(true);
+    });
+
+    it("exits 1 when the only record it cannot decrypt is an album, and 0 once the album is deleted", async () => {
+        const ente = brokenAlbumAccount();
+
+        expect(await backup(ente)).toBe(1);
+        expect(failures()).toEqual({});
+
+        ente.deleteAlbum(2, 2);
+        expect(await backup(ente)).toBe(0);
+    });
+
+    it("removes what an earlier run linked for an album or a file that no longer decrypts, and keeps the originals", async () => {
+        // At time 1 everything decrypts: the first run links files 10 and 11
+        // in Trip and file 20 in Family.
+        const ente = new FakeEnte();
+        ente.putAlbum(1, "Trip", 1);
+        ente.putFile(1, 10, 1);
+        ente.putFile(1, 11, 1);
+        ente.putAlbum(2, "Family", 1);
+        ente.putFile(2, 20, 1);
+        expect(await backup(ente)).toBe(0);
+        const collections = join(backupDir(), "collections");
+        expect(readdirSync(join(collections, "Trip")).sort()).toEqual([
+            "photo-10.jpg",
+            "photo-11.jpg",
+        ]);
+        expect(readdirSync(join(collections, "Family"))).toEqual([
+            "photo-20.jpg",
+        ]);
+
+        // At time 2 the server's copy of file 11 in Trip, and album 2 itself,
+        // no longer decrypt. A run builds the album folders, their JSON and
+        // their links from what the library holds, so it removes them as if
+        // the album and the file had been deleted: Family's folder and JSON
+        // go, and Trip neither links nor lists file 11. The originals stay at
+        // their save paths, and the run exits 1.
+        ente.putFile(1, 11, 2, { corruptKey: true });
+        ente.putAlbum(2, "Family", 2, true);
+        expect(await backup(ente)).toBe(1);
+
+        expect(readdirSync(collections).sort()).toEqual(["Trip", "Trip.json"]);
+        expect(readdirSync(join(collections, "Trip"))).toEqual([
+            "photo-10.jpg",
+        ]);
+        expect(albumFiles("Trip")).toEqual([10]);
+        expect([10, 11, 20].map(stored)).toEqual([true, true, true]);
+        expect(Object.keys(failures())).toEqual(["11"]);
+    });
+
     it("backs up a file that does not decrypt in one album but does in another, and does not report it", async () => {
         // File 30 is in albums 1 and 3, each holding its key sealed under the
         // album's own key. In album 1 that is corrupt; in album 3 it is not,
@@ -412,6 +497,13 @@ describe("quak backup with records it cannot decrypt", () => {
 
         expect(stored(30)).toBe(true);
         expect(failures()).toEqual({});
+        // Album 1 has no copy of file 30 that decrypts, so it neither links
+        // the original nor lists the file; album 3 lists it.
+        expect(readdirSync(join(backupDir(), "collections", "Trip"))).toEqual(
+            [],
+        );
+        expect(albumFiles("Trip")).toEqual([]);
+        expect(albumFiles("Favorites")).toEqual([30]);
     });
 
     it("backs up that file in a run limited to the album where it does not decrypt", async () => {

test/cli/backup.test.ts

@@ -191,9 +191,12 @@ class HostileClient extends MockClient {
 }
 
 // A content source that writes byte buffers of the expected length and can be
-// told to fail one fileID's original, to exercise per-file resilience.
+// told to fail one fileID's original, as a server error or as a 404, or to
+// write other bytes for it, to exercise per-file resilience.
 interface StubSource extends ContentSource {
     failID?: number;
+    notFoundID?: number;
+    otherBytesID?: number;
     failThumbID?: number;
     originalCalls: number;
 }
@@ -204,8 +207,10 @@ const stubSource = (): StubSource => {
         original: async ({ file: f, destination }) => {
             s.originalCalls++;
             if (s.failID === f.id) throw new Error("HTTP 500 from server");
+            if (s.notFoundID === f.id) throw new Error("HTTP 404 from server");
             const size = SIZE_BY_ID[f.id] ?? 10;
-            writeFileSync(destination, Buffer.alloc(size));
+            const fill = s.otherBytesID === f.id ? 2 : 0;
+            writeFileSync(destination, Buffer.alloc(size, fill));
             return { bytesWritten: size };
         },
         thumbnail: async ({ file: f, destination }) => {
@@ -696,17 +701,26 @@ describe("lib.backup", () => {
         // A child that has already exited: its process ID is not running.
         const exitedPID = spawnSync(process.execPath, ["-e", ""]).pid;
         const leftover = `.quak-backup-2026-03-01.100.jpg-${exitedPID}-abc123.tmp`;
+        // A live photo's image and video downloaded to replace a stored one,
+        // before they were renamed into place.
+        const replacement = [
+            `.quak-${exitedPID}-abc789.heic`,
+            `.quak-${exitedPID}-abc789.mov`,
+        ];
         // This test's own process stands in for a backup running at the same
         // time.
         const inProgress = `.quak-backup-2026-03-01.101.jpg-${process.pid}-def456.tmp`;
-        writeFileSync(join(day, leftover), "partial");
-        writeFileSync(join(day, inProgress), "partial");
+        for (const name of [leftover, ...replacement, inProgress]) {
+            writeFileSync(join(day, name), "partial");
+        }
         const lib = await openLibrary(stubSource());
 
         await lib.backup({ downloadDirectory: outDir });
 
         const names = readdirSync(day);
         expect(names).not.toContain(leftover);
+        expect(names).not.toContain(replacement[0]);
+        expect(names).not.toContain(replacement[1]);
         expect(names).toContain(inProgress);
         lib.close();
     });
@@ -909,6 +923,27 @@ describe("the backup lock", () => {
         expect(existsSync(lockPath(outDir))).toBe(false);
         await lib.close();
     });
+
+    // `runBackup` takes a library object from its caller. With a
+    // `contentConcurrency` below 1, or one that is not a number, it would
+    // back up no original and report no error. It rejects any that is not a
+    // positive integer, as `BoundedPool` does, before it takes the lock,
+    // which would create the download directory, and before it asks the
+    // library for anything: the object here has nothing else.
+    it("rejects a library whose contentConcurrency is not a positive integer, before taking the lock", async () => {
+        const outDir = join(root, "backup");
+        for (const contentConcurrency of [undefined, 0, -1, 1.5, NaN]) {
+            const lib = { contentConcurrency } as unknown as BackupLibrary;
+
+            await expect(
+                runBackup(lib, { downloadDirectory: outDir }),
+            ).rejects.toThrow(
+                `contentConcurrency must be a positive integer, got ${contentConcurrency}`,
+            );
+
+            expect(existsSync(outDir)).toBe(false);
+        }
+    });
 });
 
 // Also serves ML data: file 100 has `ML_PAYLOAD`, the other two have none.
@@ -917,13 +952,14 @@ const ML_PAYLOAD = { face: { faces: [] }, clip: { embedding: [0.5, 0.25] } };
 
 class MLClient extends MockClient {
     mlError?: string;
-    async fetchMLData(args: {
-        fileIDs: number[];
-    }): Promise<Map<number, MLData>> {
+    async fetchMLData(args: { fileIDs: number[] }): Promise<{
+        payloads: Map<number, MLData>;
+        errors: Map<number, unknown>;
+    }> {
         if (this.mlError) throw new Error(this.mlError);
-        const result = new Map<number, MLData>();
-        if (args.fileIDs.includes(100)) result.set(100, ML_PAYLOAD);
-        return result;
+        const payloads = new Map<number, MLData>();
+        if (args.fileIDs.includes(100)) payloads.set(100, ML_PAYLOAD);
+        return { payloads, errors: new Map() };
     }
 }
 
@@ -1273,7 +1309,8 @@ describe("image metadata in each file's JSON", () => {
 });
 
 // With `verify`, each original already stored is hashed, and one that does not
-// match the content hash its metadata records is downloaded again.
+// match the content hash its metadata records is downloaded again, and replaced
+// only by a download that matches.
 describe("backup with verify", () => {
     // MockClient's files, each recording the hash of what `stubSource` writes
     // for it, except diagram.png (200), which records none.
@@ -1316,6 +1353,12 @@ describe("backup with verify", () => {
     const good = Buffer.alloc(SIZE_BY_ID[100]!);
     const corrupt = Buffer.alloc(SIZE_BY_ID[100]!, 1);
 
+    // The temporary files left in the date folder of the backup in `outDir`.
+    const tempFiles = (outDir: string): string[] =>
+        readdirSync(join(outDir, DAY)).filter((name) =>
+            name.startsWith(".quak-"),
+        );
+
     it("leaves an original that matches its hash, and one with no hash, as they are", async () => {
         const { lib, source, outDir } = await backedUp();
         const calls = source.originalCalls;
@@ -1360,13 +1403,16 @@ describe("backup with verify", () => {
         expect(log.filter((msg) => msg.startsWith("MISMATCH"))).toEqual([
             "MISMATCH original beach.jpg (100): its bytes do not match its content hash",
         ]);
+        expect(tempFiles(outDir)).toEqual([]);
         await lib.close();
     });
 
-    it("records a failed download in failures.json, with the original removed", async () => {
+    // A copy whose hash does not match is still mostly the photo, so the
+    // backup keeps it until it holds a better one.
+    it("keeps an original that does not match its hash when its download answers 404", async () => {
         const { lib, source, outDir } = await backedUp();
         writeFileSync(saved(outDir, "100.jpg"), corrupt);
-        source.failID = 100;
+        source.notFoundID = 100;
 
         const result = await lib.backup({
             downloadDirectory: outDir,
@@ -1379,18 +1425,18 @@ describe("backup with verify", () => {
             failed: 1,
         });
         expect(Object.keys(readLedger(outDir).files)).toEqual(["100"]);
-        expect(existsSync(saved(outDir, "100.jpg"))).toBe(false);
+        expect(readFileSync(saved(outDir, "100.jpg"))).toEqual(corrupt);
+        expect(
+            readlinkSync(join(outDir, "collections", "Vacation", "beach.jpg")),
+        ).toBe(linkTo("100.jpg"));
+        expect(tempFiles(outDir)).toEqual([]);
         await lib.close();
     });
 
-    it("records in failures.json an original the cache puts back with the same bad bytes", async () => {
-        const lib = await openLibrary(stubSource(), new HashedClient());
-        const outDir = join(root, "backup");
-        // The cache holds a bad copy, and a backup copies an original the
-        // cache holds to its save path.
-        const cached = await lib.photos.byID({ fileID: 100 })!.original();
-        writeFileSync(cached.path, corrupt);
-        await lib.backup({ downloadDirectory: outDir });
+    it("keeps an original that does not match its hash when its download does not match either", async () => {
+        const { lib, source, outDir } = await backedUp();
+        writeFileSync(saved(outDir, "100.jpg"), corrupt);
+        source.otherBytesID = 100;
 
         const result = await lib.backup({
             downloadDirectory: outDir,
@@ -1403,10 +1449,39 @@ describe("backup with verify", () => {
             failed: 1,
         });
         expect(result.errors.map((e) => e.error)).toEqual([
-            "the original put back does not match its content hash either",
+            "the downloaded original does not match its content hash either",
         ]);
         expect(Object.keys(readLedger(outDir).files)).toEqual(["100"]);
         expect(readFileSync(saved(outDir, "100.jpg"))).toEqual(corrupt);
+        expect(tempFiles(outDir)).toEqual([]);
+        await lib.close();
+    });
+
+    // The cache can hold the same bad bytes, as a copy from it is not checked
+    // as a download is, so the replacement is always downloaded.
+    it("downloads the replacement even when the cache holds the original", async () => {
+        const source = stubSource();
+        const lib = await openLibrary(source, new HashedClient());
+        const outDir = join(root, "backup");
+        // The cache holds a bad copy, and a backup copies an original the
+        // cache holds to its save path.
+        const cached = await lib.photos.byID({ fileID: 100 })!.original();
+        writeFileSync(cached.path, corrupt);
+        await lib.backup({ downloadDirectory: outDir });
+        const calls = source.originalCalls;
+
+        const result = await lib.backup({
+            downloadDirectory: outDir,
+            verify: true,
+        });
+
+        expect(result).toMatchObject({
+            downloaded: 1,
+            mismatched: 1,
+            failed: 0,
+        });
+        expect(source.originalCalls).toBe(calls + 1);
+        expect(readFileSync(saved(outDir, "100.jpg"))).toEqual(good);
         await lib.close();
     });
 
@@ -1477,22 +1552,29 @@ describe("backup album folders", () => {
         files: EnteFile[];
     }
 
-    const libraryOf = (albums: Album[]): BackupLibrary => ({
+    // The original of the file `notFound`, if given, always fails to download.
+    const libraryOf = (albums: Album[], notFound?: number): BackupLibrary => ({
         refresh: async () => {},
         listCollections: () => albums.map((a) => a.collection),
         listFiles: (id) =>
             albums.find((a) => a.collection.id === id)?.files ?? [],
         undecryptable: () => [],
         original: async (fileID) => {
+            if (fileID === notFound) throw new Error("HTTP 404");
             const path = join(root, `source-${fileID}`);
             writeFileSync(path, `original ${fileID}`);
             return { path };
         },
+        fetchOriginal: async () => {
+            throw new Error("no verify in this stand-in");
+        },
         thumbnail: async () => {
             throw new Error("no thumbnails in this stand-in");
         },
+        contentConcurrency: 5,
         fetchMLData: async () => {},
         mlData: async () => undefined,
+        mlDataError: () => undefined,
         whoami: () => ({ email: "backup@example.com", userID: USER_ID }),
     });
 
@@ -1578,6 +1660,252 @@ describe("backup album folders", () => {
         expect(albumID(outDir, "Trip (11) (12).json")).toBe(12);
     });
 
+    // The next three tests end with the album "ok" (14), whose one file
+    // always fails to download, so each run has a failure to record in
+    // failures.json after the album folders are written.
+    const ok: Album = {
+        collection: collection(14, "ok"),
+        files: [file(5, 14, "e.jpg")],
+    };
+
+    // The folder of an album named "x.json" would be the JSON file of the
+    // album "x", and once ended every run there. Two runs, as on two nights,
+    // write the same tree, and the failed file's attempt count goes up.
+    it("adds the ID to an album whose folder would have another album's JSON name", async () => {
+        const outDir = join(root, "backup");
+        const lib = libraryOf(
+            [
+                {
+                    collection: collection(10, "x"),
+                    files: [file(1, 10, "a.jpg")],
+                },
+                {
+                    collection: collection(11, "x.json"),
+                    files: [file(2, 11, "b.jpg")],
+                },
+                {
+                    collection: collection(12, "Trip"),
+                    files: [file(3, 12, "c.jpg")],
+                },
+                {
+                    collection: collection(13, "TRIP.json"),
+                    files: [file(4, 13, "d.jpg")],
+                },
+                ok,
+            ],
+            5,
+        );
+
+        await runBackup(lib, { downloadDirectory: outDir });
+        const result = await runBackup(lib, { downloadDirectory: outDir });
+
+        expect(result.failedAlbums).toEqual([]);
+        expect(tree(outDir)).toEqual([
+            "TRIP.json (13)/",
+            `TRIP.json (13)/d.jpg -> ${linkTo("4.jpg")}`,
+            "TRIP.json (13).json",
+            "Trip/",
+            `Trip/c.jpg -> ${linkTo("3.jpg")}`,
+            "Trip.json",
+            "ok/",
+            "ok.json",
+            "x/",
+            `x/a.jpg -> ${linkTo("1.jpg")}`,
+            "x.json",
+            "x.json (11)/",
+            `x.json (11)/b.jpg -> ${linkTo("2.jpg")}`,
+            "x.json (11).json",
+        ]);
+        expect(albumID(outDir, "x.json")).toBe(10);
+        expect(albumID(outDir, "x.json (11).json")).toBe(11);
+        expect(albumID(outDir, "TRIP.json (13).json")).toBe(13);
+        expect(albumID(outDir, "ok.json")).toBe(14);
+        expect(result.failed).toBe(1);
+        expect(readLedger(outDir).files["5"]).toMatchObject({ attempts: 2 });
+    });
+
+    // A name over 255 bytes is refused as a file name, and an album folder's
+    // name leaves room for `.json` in its JSON's. "é" takes two bytes, and a
+    // name is cut between two characters.
+    it("cuts an album name and a title too long for a file name short and adds the ID", async () => {
+        const outDir = join(root, "backup");
+        const lib = libraryOf(
+            [
+                {
+                    collection: collection(10, "L".repeat(300)),
+                    files: [file(1, 10, `${"é".repeat(150)}.jpg`)],
+                },
+                ok,
+            ],
+            5,
+        );
+
+        await runBackup(lib, { downloadDirectory: outDir });
+        const result = await runBackup(lib, { downloadDirectory: outDir });
+
+        const folder = `${"L".repeat(245)} (10)`;
+        const link = `${"é".repeat(123)} (1).jpg`;
+        expect(Buffer.byteLength(`${folder}.json`)).toBe(255);
+        expect(Buffer.byteLength(link)).toBe(254);
+        expect(result.failedAlbums).toEqual([]);
+        expect(tree(outDir)).toEqual([
+            `${folder}/`,
+            `${folder}/${link} -> ${linkTo("1.jpg")}`,
+            `${folder}.json`,
+            "ok/",
+            "ok.json",
+        ]);
+        expect(albumID(outDir, `${folder}.json`)).toBe(10);
+        expect(albumID(outDir, "ok.json")).toBe(14);
+        expect(result.failed).toBe(1);
+        expect(readLedger(outDir).files["5"]).toMatchObject({ attempts: 2 });
+    });
+
+    // The user has put a file in collections/ with the name of the album
+    // Trip's folder, so the folder cannot be made.
+    it("goes on to the next album when an album's folder cannot be made, and lists it as failed", async () => {
+        const outDir = join(root, "backup");
+        const lib = libraryOf(
+            [
+                {
+                    collection: collection(10, "Trip"),
+                    files: [file(1, 10, "a.jpg")],
+                },
+                ok,
+            ],
+            5,
+        );
+        mkdirSync(join(outDir, "collections"), { recursive: true });
+        writeFileSync(join(outDir, "collections", "Trip"), "mine");
+        const logged: string[] = [];
+
+        const result = await runBackup(lib, {
+            downloadDirectory: outDir,
+            onProgress: (message) => logged.push(message),
+        });
+
+        expect(result.failedAlbums).toEqual([10]);
+        expect(logged).toContainEqual(
+            expect.stringMatching(/^FAILED album Trip: EEXIST: /),
+        );
+        expect(tree(outDir)).toEqual(["Trip", "ok/", "ok.json"]);
+        expect(albumID(outDir, "ok.json")).toBe(14);
+        expect(existsSync(saved(outDir, "1.jpg"))).toBe(true);
+        expect(result.failed).toBe(1);
+        expect(readLedger(outDir).files["5"]).toMatchObject({ attempts: 1 });
+    });
+
+    // The extensions of these titles leave no room for the ID, so each is cut
+    // short and the name before the ID keeps only the bytes that cut leaves
+    // over. File 1's extension is 303 bytes and is cut to 251, which leaves
+    // none. File 2's is 271 bytes and is cut to 250, since "€" takes three
+    // bytes, which leaves one, for the "a". Neither extension is only letters
+    // and digits, so both originals are saved with `.bin`, a name that fits.
+    it("cuts the extension of a title short when it leaves no room for the ID", async () => {
+        const outDir = join(root, "backup");
+        const lib = libraryOf([
+            {
+                collection: collection(10, "Trip"),
+                files: [
+                    file(1, 10, `a.${"L".repeat(300)}-x`),
+                    file(2, 10, `a.${"€".repeat(90)}`),
+                ],
+            },
+        ]);
+
+        const result = await runBackup(lib, { downloadDirectory: outDir });
+
+        const link = ` (1).${"L".repeat(250)}`;
+        const link2 = `a (2).${"€".repeat(83)}`;
+        expect(Buffer.byteLength(link)).toBe(255);
+        expect(Buffer.byteLength(link2)).toBe(255);
+        expect(result.failed).toBe(0);
+        expect(tree(outDir)).toEqual([
+            "Trip/",
+            `Trip/${link} -> ${linkTo("1.bin")}`,
+            `Trip/${link2} -> ${linkTo("2.bin")}`,
+            "Trip.json",
+        ]);
+    });
+
+    // Files 1 and 3 share a title, so each gets its ID before the
+    // extension. File 2's title is too long, and its extension leaves no
+    // room for the ID. Were the ID put at the end of file 2's title cut
+    // short, its link would have file 1's name, and the link made second
+    // would replace the first.
+    it("keeps names unique when a title's extension leaves no room for the ID", async () => {
+        const outDir = join(root, "backup");
+        const shared = `x.${"Q".repeat(245)} (2)`;
+        const lib = libraryOf([
+            {
+                collection: collection(10, "Trip"),
+                files: [
+                    file(1, 10, shared),
+                    file(2, 10, `x (1).${"Q".repeat(250)}-`),
+                    file(3, 10, shared),
+                ],
+            },
+        ]);
+
+        const result = await runBackup(lib, { downloadDirectory: outDir });
+
+        expect(result.failed).toBe(0);
+        expect(tree(outDir)).toEqual([
+            "Trip/",
+            `Trip/ (2).${"Q".repeat(250)} -> ${linkTo("2.bin")}`,
+            `Trip/x (1).${"Q".repeat(245)} (2) -> ${linkTo("1.bin")}`,
+            `Trip/x (3).${"Q".repeat(245)} (2) -> ${linkTo("3.bin")}`,
+            "Trip.json",
+        ]);
+    });
+
+    // A title of 255 bytes is the longest file name allowed, and an album
+    // name of 250 bytes the longest whose JSON name fits; both are kept as
+    // they are. A name one byte longer gets the ID and is cut short. "é"
+    // takes two bytes, and in both names the room left ends inside one, so
+    // the cut falls before it, one byte short of the room.
+    it("keeps a name at the byte limit and adds the ID to one a byte over it", async () => {
+        const outDir = join(root, "backup");
+        const album = "A".repeat(250);
+        const title = `${"t".repeat(251)}.jpg`;
+        const longAlbum = `${"é".repeat(125)}B`;
+        const longTitle = `${"é".repeat(126)}.jpg`;
+        const lib = libraryOf([
+            {
+                collection: collection(10, album),
+                files: [file(1, 10, title), file(2, 10, longTitle)],
+            },
+            {
+                collection: collection(11, longAlbum),
+                files: [file(3, 11, "c.jpg")],
+            },
+        ]);
+
+        const result = await runBackup(lib, { downloadDirectory: outDir });
+
+        const folder = `${"é".repeat(122)} (11)`;
+        const link = `${"é".repeat(123)} (2).jpg`;
+        expect(Buffer.byteLength(`${album}.json`)).toBe(255);
+        expect(Buffer.byteLength(title)).toBe(255);
+        expect(Buffer.byteLength(longAlbum)).toBe(251);
+        expect(Buffer.byteLength(longTitle)).toBe(256);
+        expect(Buffer.byteLength(`${folder}.json`)).toBe(254);
+        expect(Buffer.byteLength(link)).toBe(254);
+        expect(result.failed).toBe(0);
+        expect(result.failedAlbums).toEqual([]);
+        expect(tree(outDir)).toEqual([
+            `${album}/`,
+            `${album}/${title} -> ${linkTo("1.jpg")}`,
+            `${album}/${link} -> ${linkTo("2.jpg")}`,
+            `${album}.json`,
+            `${folder}/`,
+            `${folder}/c.jpg -> ${linkTo("3.jpg")}`,
+            `${folder}.json`,
+        ]);
+        expect(albumID(outDir, `${album}.json`)).toBe(10);
+        expect(albumID(outDir, `${folder}.json`)).toBe(11);
+    });
+
     it("changes nothing on a second run over an unchanged account", async () => {
         const outDir = join(root, "backup");
         const lib = libraryOf([
@@ -1974,11 +2302,12 @@ describe("backup of live photos", () => {
             Buffer.from(IMAGE),
         );
         expect(readFileSync(video)).toEqual(Buffer.from(VIDEO));
+        expect(readdirSync(join(outDir, DAY)).sort()).toEqual(stored);
         expect(tree(outDir)).toEqual(linked);
         await lib.close();
     });
 
-    it("removes both files of a live photo that does not match its hash when downloading it again fails", async () => {
+    it("keeps both files of a live photo that does not match its hash when downloading it again fails", async () => {
         const { file: live, body } = await asLivePhoto(
             file(500, 10, "IMG_0500.HEIC"),
         );
@@ -1996,8 +2325,14 @@ describe("backup of live photos", () => {
 
         expect(result).toMatchObject({ mismatched: 1, failed: 1 });
         expect(Object.keys(readLedger(outDir).files)).toEqual(["500"]);
-        expect(existsSync(saved(outDir, "500.heic"))).toBe(false);
-        expect(existsSync(saved(outDir, "500.mov"))).toBe(false);
+        expect(readFileSync(saved(outDir, "500.heic"))).toEqual(
+            Buffer.from(IMAGE),
+        );
+        expect(readFileSync(saved(outDir, "500.mov"), "utf-8")).toBe(
+            "another few seconds of video",
+        );
+        expect(readdirSync(join(outDir, DAY)).sort()).toEqual(stored);
+        expect(tree(outDir)).toEqual(linked);
         await lib.close();
     });
 

test/cli/commands.test.ts

@@ -25,6 +25,7 @@ import { join } from "node:path";
 import { tmpdir } from "node:os";
 import { PassThrough } from "node:stream";
 import * as jpegJs from "jpeg-js";
+import sodium from "libsodium-wrappers-sumo";
 import {
     describe,
     it,
@@ -55,7 +56,7 @@ import { loadSession } from "../../src/cli-session.js";
 import type { Client, ClientSnapshot, LoginOptions } from "../../src/client.js";
 import { savePath, type ContentSource } from "../../src/library/content.js";
 import type { Collection, EnteFile } from "../../src/model/types.js";
-import { init, toBase64 } from "../../src/crypto/index.js";
+import { encryptBlob, init, toBase64 } from "../../src/crypto/index.js";
 import { defaultCacheDirectory } from "../../src/library/index.js";
 import { HEIC_WITH_EXIF } from "../exif-heic.js";
 import {
@@ -903,6 +904,146 @@ describe("backup", () => {
         // Opening the library would have made its cache directory.
         expect(existsSync(join(root, "cache"))).toBe(false);
     });
+
+    // A backup runs for as long as its downloads take, hours on a large
+    // account. It lists its files from the refresh it waits for as it starts,
+    // so it must not ask the server for changes again on a timer while it
+    // runs: each refresh is a request, and each also starts a pass that asks
+    // again for the ML data of every file Ente holds none for.
+    //
+    // A saved session is read back by `loadSession`, its `fetch` a fake of the
+    // Ente HTTP API serving one album with one photo, encrypted as the server
+    // serves them. The fake holds the photo's download open while the clock
+    // moves on by two hours, then serves it. Only `setTimeout` and
+    // `clearTimeout` are faked: the library's refresh timer runs on them.
+    it("asks the server for changes only as it starts, however long it runs", async () => {
+        const masterKey = new Uint8Array(32).fill(1);
+        const albumKey = new Uint8Array(32).fill(2);
+        const photoKey = new Uint8Array(32).fill(3);
+        saveSession(join(root, "session"), {
+            email: "cli@example.com",
+            userID: USER_ID,
+            token: "token",
+            masterKey: toBase64(masterKey),
+            secretKey: toBase64(new Uint8Array(32)),
+            publicKey: toBase64(new Uint8Array(32)),
+        });
+        // A secretbox as the server sends one: ciphertext and nonce, base64.
+        const box = (plaintext: Uint8Array, key: Uint8Array) => {
+            const nonce = sodium.randombytes_buf(
+                sodium.crypto_secretbox_NONCEBYTES,
+            );
+            return {
+                ciphertext: toBase64(
+                    sodium.crypto_secretbox_easy(plaintext, nonce, key),
+                ),
+                nonce: toBase64(nonce),
+            };
+        };
+        const albumKeyBox = box(albumKey, masterKey);
+        const albumName = box(new TextEncoder().encode("Vacation"), albumKey);
+        const photoKeyBox = box(photoKey, albumKey);
+        const metadata = encryptBlob(
+            new TextEncoder().encode(
+                JSON.stringify({
+                    title: "beach.jpg",
+                    fileType: 0,
+                    creationTime: 1_700_000_000_000_000,
+                    modificationTime: 1_700_000_000_000_000,
+                }),
+            ),
+            photoKey,
+        );
+        const content = encryptBlob(
+            new TextEncoder().encode("the photo's bytes"),
+            photoKey,
+        );
+        const album = {
+            id: 1,
+            owner: { id: USER_ID },
+            encryptedKey: albumKeyBox.ciphertext,
+            keyDecryptionNonce: albumKeyBox.nonce,
+            encryptedName: albumName.ciphertext,
+            nameDecryptionNonce: albumName.nonce,
+            type: "album",
+            updationTime: 1,
+        };
+        const photo = {
+            id: 100,
+            collectionID: 1,
+            ownerID: USER_ID,
+            encryptedKey: photoKeyBox.ciphertext,
+            keyDecryptionNonce: photoKeyBox.nonce,
+            metadata: {
+                encryptedData: toBase64(metadata.ciphertext),
+                decryptionHeader: toBase64(metadata.header),
+            },
+            file: { decryptionHeader: toBase64(content.header) },
+            // The backup fetches no thumbnail.
+            thumbnail: { decryptionHeader: "" },
+            updationTime: 1,
+        };
+
+        // Each refresh starts with `collectionsSince`, which sends one
+        // `/collections/v2` request.
+        let refreshes = 0;
+        let downloadRequested!: () => void;
+        const requested = new Promise<void>((resolve) => {
+            downloadRequested = resolve;
+        });
+        let releaseDownload!: () => void;
+        const released = new Promise<void>((resolve) => {
+            releaseDownload = resolve;
+        });
+        const json = (body: unknown): Response =>
+            new Response(JSON.stringify(body), {
+                headers: { "content-type": "application/json" },
+            });
+        const fetch = async (url: string | URL): Promise<Response> => {
+            const { hostname, pathname } = new URL(String(url));
+            if (hostname === "files.ente.io") {
+                downloadRequested();
+                await released;
+                return new Response(content.ciphertext);
+            }
+            if (pathname === "/collections/v2") {
+                refreshes++;
+                return json({ collections: [album] });
+            }
+            if (pathname === "/collections/v2/diff") {
+                return json({ diff: [photo], hasMore: false });
+            }
+            // Ente holds no ML data for the photo.
+            if (pathname === "/files/data/fetch") return json({ data: [] });
+            return new Response("not found", { status: 404 });
+        };
+        const ctx = {
+            ...context(),
+            // The download's idle deadline is set past the two hours, or quak
+            // would give up the held download as stalled.
+            loadSession: (path: string) =>
+                loadSession(path, {
+                    fetch: fetch as typeof globalThis.fetch,
+                    downloadTimeoutMs: 3 * 3_600_000,
+                }),
+        };
+
+        vi.useFakeTimers({ toFake: ["setTimeout", "clearTimeout"] });
+        try {
+            const backup = backupCommand(ctx, join(root, "backup"), {});
+            await requested;
+            vi.advanceTimersByTime(2 * 3_600_000);
+            releaseDownload();
+            expect(await backup).toBe(0);
+        } finally {
+            vi.useRealTimers();
+        }
+
+        // Two refreshes, both before the download: the library opens onto the
+        // empty cache with one, and the backup waits for the other before it
+        // lists its files.
+        expect(refreshes).toBe(2);
+    });
 });
 
 describe("helper list-missing-thumbnails", () => {

test/cli/metadata-backup.test.ts

@@ -489,15 +489,15 @@ describe("quak backup-metadata", () => {
     });
 });
 
-describe("quak backup-metadata when an ML data request fails", () => {
-    // Run the CLI command against the mock and return its exit code, stderr
-    // and output directory.
-    const runCommand = async (failMLDataFor?: number) => {
+describe("quak backup-metadata when a file's ML data cannot be read", () => {
+    // Run the CLI command against `fetch`, the mock unless given, and return
+    // its exit code, stderr and output directory.
+    const runCommand = async (fetch = buildMetaFetch(mock)) => {
         const client = await Client.login({
             email: TEST_EMAIL,
             password: TEST_PASSWORD,
             apiOptions: {
-                fetch: buildMetaFetch(mock, failMLDataFor),
+                fetch,
                 retry: { sleep: async () => {} },
             },
         });
@@ -518,7 +518,9 @@ describe("quak backup-metadata when an ML data request fails", () => {
     };
 
     it("writes every file, marks the failed batch's files, and exits 1", async () => {
-        const { code, stderr, outDir } = await runCommand(200);
+        const { code, stderr, outDir } = await runCommand(
+            buildMetaFetch(mock, 200),
+        );
 
         expect(code).toBe(1);
         expect(stderr).toContain("ML data request for 1 file(s) failed");
@@ -543,6 +545,49 @@ describe("quak backup-metadata when an ML data request fails", () => {
         expect(failed.mlDataError).toContain("500");
     });
 
+    // The server sends ML data for diagram.png encrypted under a key that is
+    // not the file's, so quak cannot decrypt it.
+    it("writes the reason for ML data that does not decrypt, and exits 1", async () => {
+        const otherKey = sodium.crypto_secretstream_xchacha20poly1305_keygen();
+        const { header, ciphertext } = encryptBlob(
+            new Uint8Array(gzipSync(Buffer.from("{}"))),
+            otherKey,
+        );
+        const { code, stderr, outDir } = await runCommand(
+            buildMetaFetch({
+                ...mock,
+                encryptedMLData: {
+                    ...mock.encryptedMLData,
+                    200: {
+                        encryptedData: toBase64(ciphertext),
+                        decryptionHeader: toBase64(header),
+                    },
+                },
+            }),
+        );
+
+        expect(code).toBe(1);
+        const reason = "secretstream chunk authentication failed";
+        expect(stderr).toContain(
+            `ML data of file 200 could not be read: ${reason}\n`,
+        );
+        const failed = JSON.parse(
+            readFileSync(
+                join(outDir, "collections", "20-__Work", "200.json"),
+                "utf-8",
+            ),
+        );
+        expect(failed.mlData).toBeUndefined();
+        expect(failed.mlDataError).toBe(reason);
+        const ok = JSON.parse(
+            readFileSync(
+                join(outDir, "collections", "10-Vacation", "100.json"),
+                "utf-8",
+            ),
+        );
+        expect(ok.mlData.clip.embedding).toEqual([0.5, 0.6, 0.7]);
+    });
+
     it("exits 0 when every ML data request succeeds", async () => {
         const { code } = await runCommand();
         expect(code).toBe(0);

test/cli/run.test.ts

@@ -48,6 +48,78 @@ describe("run", () => {
         );
     });
 
+    // Node's `fetch` rejects every failure below HTTP with a TypeError that
+    // says only "fetch failed", and hangs the reason, with its errno, off
+    // `cause`. The printed line carries the reason, so a cron log tells a
+    // refused connection from a name that did not resolve.
+    it("prints the reason a fetch failed, from the error's cause", async () => {
+        const refused = new TypeError("fetch failed", {
+            cause: Object.assign(
+                new Error("connect ECONNREFUSED 1.2.3.4:443"),
+                { code: "ECONNREFUSED" },
+            ),
+        });
+        const result = await runToExit(Promise.reject(refused));
+        expect(result).toEqual({
+            code: 1,
+            stdout: "",
+            stderr: "quak: fetch failed: connect ECONNREFUSED 1.2.3.4:443\n",
+        });
+    });
+
+    // When the host has several addresses and every one refuses, Node's
+    // `fetch` gives as the cause an AggregateError with an empty message and
+    // one error per address. The printed line gives each address's reason.
+    it("prints the reason for each address a fetch could not reach", async () => {
+        const refusedAt = (address: string): Error =>
+            Object.assign(new Error(`connect ECONNREFUSED ${address}`), {
+                code: "ECONNREFUSED",
+            });
+        const refused = new TypeError("fetch failed", {
+            cause: Object.assign(
+                new AggregateError([
+                    refusedAt("2001:db8::1:443"),
+                    refusedAt("1.2.3.4:443"),
+                ]),
+                { code: "ECONNREFUSED" },
+            ),
+        });
+        const result = await runToExit(Promise.reject(refused));
+        expect(result).toEqual({
+            code: 1,
+            stdout: "",
+            stderr: "quak: fetch failed: connect ECONNREFUSED 2001:db8::1:443, connect ECONNREFUSED 1.2.3.4:443\n",
+        });
+    });
+
+    // An error that already quotes its cause's message is printed as it is.
+    it("does not print a cause's message twice", async () => {
+        const reset = new Error("read ECONNRESET");
+        const wrapped = new Error(`download: file 7: ${reset.message}`, {
+            cause: reset,
+        });
+        const result = await runToExit(Promise.reject(wrapped));
+        expect(result.stderr).toBe("quak: download: file 7: read ECONNRESET\n");
+    });
+
+    // OpenSSL ends its message for a failed TLS connection, such as one to a
+    // server that answers in plain HTTP, with a newline. The printed line is
+    // still one line.
+    it("prints a cause's message without the white space around it", async () => {
+        const wrongVersion = new TypeError("fetch failed", {
+            cause: Object.assign(
+                new Error(
+                    "C0B73A55A47F0000:error:0A00010B:SSL routines:ssl3_get_record:wrong version number:../deps/openssl/openssl/ssl/record/ssl3_record.c:354:\n",
+                ),
+                { code: "ERR_SSL_WRONG_VERSION_NUMBER" },
+            ),
+        });
+        const result = await runToExit(Promise.reject(wrongVersion));
+        expect(result.stderr).toBe(
+            "quak: fetch failed: C0B73A55A47F0000:error:0A00010B:SSL routines:ssl3_get_record:wrong version number:../deps/openssl/openssl/ssl/record/ssl3_record.c:354:\n",
+        );
+    });
+
     it("prints a thrown value that is not an Error", async () => {
         const result = await runToExit(Promise.reject("offline"));
         expect(result).toEqual({
@@ -57,6 +129,14 @@ describe("run", () => {
         });
     });
 
+    // A cause that is not an Error, such as a thrown string, is printed as a
+    // thrown value that is not an Error is.
+    it("prints a cause that is not an Error", async () => {
+        const wrapped = new Error("ML data", { cause: "offline" });
+        const result = await runToExit(Promise.reject(wrapped));
+        expect(result.stderr).toBe("quak: ML data: offline\n");
+    });
+
     it("on a 401 from the server says to run quak login, on one line, and exits 3", async () => {
         const result = await runToExit(
             Promise.reject(new ApiError("unauthorized", 401)),

test/download/download.test.ts

@@ -74,7 +74,11 @@ import {
 } from "vitest";
 import { init, toBase64, STREAM_CHUNK_SIZE } from "../../src/crypto/index.js";
 import { ApiClient } from "../../src/api/client.js";
-import { ApiError, TruncatedStreamError } from "../../src/errors.js";
+import {
+    ApiError,
+    describeError,
+    TruncatedStreamError,
+} from "../../src/errors.js";
 import type { RetryOptions } from "../../src/retry.js";
 import {
     downloadFile,
@@ -834,6 +838,11 @@ describe.each(entryPoints)(
             expect(((err as Error).cause as Error).message).toMatch(
                 /authentication failed/i,
             );
+            // As quak prints it and records it in `failures.json`, the error
+            // states the authentication failure once, from its cause.
+            expect(describeError(err)).toMatch(
+                /^download: stream truncated: response body ended with \d+ trailing bytes \(transfer stopped mid-chunk, or the data is corrupt\): secretstream chunk authentication failed$/,
+            );
 
             // And, as with every other failure, the destination is untouched
             // and no staged temp file survives.

test/filename/filename.test.ts

@@ -14,7 +14,11 @@
 
 import { describe, expect, it } from "vitest";
 
-import { safeExtension, sanitizeFileName } from "../../src/filename.js";
+import {
+    safeExtension,
+    sanitizeFileName,
+    shortenToBytes,
+} from "../../src/filename.js";
 
 const FALLBACK = "file-42";
 
@@ -91,3 +95,26 @@ describe("safeExtension", () => {
         expect(safeExtension("x.jpg\0")).toBe(".bin");
     });
 });
+
+describe("shortenToBytes", () => {
+    it("returns a name that fits unchanged", () => {
+        expect(shortenToBytes("beach", 5)).toBe("beach");
+        expect(shortenToBytes("café", 5)).toBe("café");
+        expect(shortenToBytes("", 0)).toBe("");
+    });
+
+    it("cuts a name to the number of bytes given", () => {
+        expect(shortenToBytes("L".repeat(300), 255)).toBe("L".repeat(255));
+    });
+
+    // In UTF-8 "é" takes two bytes and an emoji four, which a JavaScript
+    // string holds as two halves.
+    it("cuts between two characters, never inside one", () => {
+        expect(shortenToBytes("ééé", 5)).toBe("éé");
+        expect(shortenToBytes("a\u{1F600}b", 4)).toBe("a");
+        expect(shortenToBytes("a\u{1F600}b", 5)).toBe("a\u{1F600}");
+        expect(Buffer.byteLength(shortenToBytes("é".repeat(200), 255))).toBe(
+            254,
+        );
+    });
+});

test/library/mldata.test.ts

@@ -12,9 +12,11 @@
  *
  *   2. `Library` wiring: after each refresh the library fetches ML data through
  *      the metadata pool for every known file not yet cached, is incremental on
- *      later refreshes, and refetches a file whose `updationTime` advanced.
- *      Opening a cache directory written by another account starts empty,
- *      its ML data included (issue #104).
+ *      later refreshes, and refetches a file whose `updationTime` advanced. A
+ *      file whose ML data did not decode is reported with the reason, which
+ *      stays listed until the next pass ends, and asked for again. Opening a
+ *      cache directory written by another account starts empty, its ML data
+ *      included (issue #104).
  *
  * Embedding values are chosen to be exactly representable as float32 so the
  * round-trip through `clip.f32` compares equal.
@@ -26,7 +28,7 @@ import { tmpdir } from "node:os";
 import { join } from "node:path";
 
 import { MLDataStore } from "../../src/library/mldata.js";
-import { Library } from "../../src/library/index.js";
+import { Library, type RefreshEvent } from "../../src/library/index.js";
 import type { CollectionsPage, FilesPage } from "../../src/client.js";
 import type { MLData } from "../../src/mldata-fetch.js";
 import type { Collection, EnteFile } from "../../src/model/types.js";
@@ -256,12 +258,14 @@ const file = (
 
 // A mock client that serves scripted collection/file pages and per-file ML
 // payloads, recording every ML fetch request so incremental behaviour is
-// provable.
+// provable. A file in `mlErrorByFile` has ML data that did not decode, for
+// the reason given there.
 class MLMockClient {
     userID = USER_ID;
     collectionsQueue: CollectionsPage[] = [];
     filesByCollection = new Map<number, FilesPage[]>();
     mlByFile = new Map<number, MLData>();
+    mlErrorByFile = new Map<number, Error>();
     mlFetchCalls: number[][] = [];
 
     whoami(): { email: string; userID: number } {
@@ -298,14 +302,20 @@ class MLMockClient {
     async fetchMLData(args: {
         fileIDs: number[];
         fileKeys: Map<number, Uint8Array>;
-    }): Promise<Map<number, MLData>> {
+    }): Promise<{
+        payloads: Map<number, MLData>;
+        errors: Map<number, unknown>;
+    }> {
         this.mlFetchCalls.push([...args.fileIDs]);
-        const result = new Map<number, MLData>();
+        const payloads = new Map<number, MLData>();
+        const errors = new Map<number, unknown>();
         for (const id of args.fileIDs) {
             const p = this.mlByFile.get(id);
-            if (p) result.set(id, p);
+            if (p) payloads.set(id, p);
+            const error = this.mlErrorByFile.get(id);
+            if (error) errors.set(id, error);
         }
-        return result;
+        return { payloads, errors };
     }
 
     filesFor(collectionID: number, ...pages: FilesPage[]): void {
@@ -449,6 +459,137 @@ describe("Library ML-data fetch on refresh", () => {
         }
     });
 
+    it("reports a file whose ML data did not decode through status() and onProgress, and asks for it again", async () => {
+        const client = new MLMockClient();
+        client.collectionsQueue.push({
+            collections: [collection(1, 100)],
+            deleted: [],
+            cursor: 100,
+        });
+        client.filesFor(1, {
+            files: [file(1001, 1, 90), file(1002, 1, 95)],
+            deleted: [],
+            cursor: 95,
+        });
+        client.mlByFile.set(1001, payload([0.5, 0.25, 0.75]));
+        client.mlErrorByFile.set(
+            1002,
+            new Error("secretstream chunk authentication failed"),
+        );
+        const events: RefreshEvent[] = [];
+
+        const lib = await Library.open({
+            client,
+            cacheDirectory,
+            refreshIntervalSeconds: Infinity,
+            onProgress: (event) => events.push(event),
+        });
+        try {
+            await vi.waitFor(
+                () => expect(lib.status().lastMLFetchAt).toBeGreaterThan(0),
+                { timeout: 2000, interval: 5 },
+            );
+            const reasons = [
+                {
+                    fileID: 1002,
+                    error: "secretstream chunk authentication failed",
+                },
+            ];
+            expect(lib.status().mlDataErrors).toEqual(reasons);
+            expect(lib.status().mlStored).toBe(1);
+            expect(events.filter((e) => e.operation === "fetchMLData")).toEqual(
+                [
+                    { operation: "fetchMLData", status: "started" },
+                    {
+                        operation: "fetchMLData",
+                        status: "done",
+                        fetched: 1,
+                        mlDataErrors: reasons,
+                    },
+                ],
+            );
+
+            // Nothing was cached for 1002, so the pass a refresh starts asks
+            // for it again, and once its ML data decodes the reason is gone.
+            client.mlErrorByFile.clear();
+            client.mlByFile.set(1002, payload([1, -2, 0.5]));
+            await lib.fresh();
+            await vi.waitFor(
+                () => {
+                    expect(lib.status().mlStored).toBe(2);
+                    expect(lib.status().mlDataErrors).toEqual([]);
+                },
+                { timeout: 2000, interval: 5 },
+            );
+            expect(client.mlFetchCalls.at(-1)).toEqual([1002]);
+        } finally {
+            await lib.close();
+        }
+    });
+
+    it("still lists the last pass's reasons while the next pass runs", async () => {
+        const client = new MLMockClient();
+        client.collectionsQueue.push({
+            collections: [collection(1, 100)],
+            deleted: [],
+            cursor: 100,
+        });
+        client.filesFor(1, {
+            files: [file(1002, 1, 95)],
+            deleted: [],
+            cursor: 95,
+        });
+        client.mlErrorByFile.set(
+            1002,
+            new Error("secretstream chunk authentication failed"),
+        );
+        const reasons = [
+            { fileID: 1002, error: "secretstream chunk authentication failed" },
+        ];
+        let release!: () => void;
+        const held = new Promise<void>((r) => (release = r));
+
+        const lib = await Library.open({
+            client,
+            cacheDirectory,
+            refreshIntervalSeconds: Infinity,
+        });
+        try {
+            await vi.waitFor(
+                () => expect(lib.status().lastMLFetchAt).toBeGreaterThan(0),
+                { timeout: 2000, interval: 5 },
+            );
+            expect(lib.status().mlDataErrors).toEqual(reasons);
+
+            // Hold the next ML data fetch open until the test releases it.
+            let fetchStarted!: () => void;
+            const started = new Promise<void>((r) => (fetchStarted = r));
+            const realFetch = client.fetchMLData.bind(client);
+            client.fetchMLData = async (args) => {
+                fetchStarted();
+                await held;
+                return realFetch(args);
+            };
+            client.mlErrorByFile.clear();
+            client.mlByFile.set(1002, payload([1, -2, 0.5]));
+
+            // The pass a refresh starts asks for 1002 again. While it is
+            // held, the reason from the pass before is still listed; once it
+            // ends, its own empty list takes the place of that one.
+            await lib.fresh();
+            await started;
+            expect(lib.status().mlDataErrors).toEqual(reasons);
+            release();
+            await vi.waitFor(
+                () => expect(lib.status().mlDataErrors).toEqual([]),
+                { timeout: 2000, interval: 5 },
+            );
+        } finally {
+            release();
+            await lib.close();
+        }
+    });
+
     it("close() resolves only after a running ML data fetch has stored its payloads", async () => {
         const client = new MLMockClient();
         client.collectionsQueue.push({

test/library/store.test.ts

@@ -172,6 +172,29 @@ describe("MetadataStore", () => {
         expect(store.listCollections()).toEqual([]);
     });
 
+    it("loads an empty store when the list of what the refresh could not decrypt is damaged", async () => {
+        // A backup reads that list on every run, so a damaged one that loaded
+        // would fail every backup until the cache was deleted. A string is
+        // damaged too, though JavaScript can loop over its characters.
+        mkdirSync(join(dir, "cache"), { recursive: true });
+        for (const undecryptable of [{}, "not a list", [1]]) {
+            writeFileSync(
+                path,
+                JSON.stringify({
+                    schemaVersion: METADATA_SCHEMA_VERSION,
+                    userID: 42,
+                    collectionsSinceTime: 0,
+                    collections: [],
+                    files: [],
+                    undecryptable,
+                }),
+            );
+            const store = await MetadataStore.load(path);
+            expect(store.userID).toBe(0);
+            expect(store.listUndecryptable()).toEqual([]);
+        }
+    });
+
     it("creates the directory 0700 and the file 0600", async () => {
         const store = await MetadataStore.load(path);
         store.putCollection(sampleCollection());

test/mldata-fetch/mldata-fetch.test.ts

@@ -0,0 +1,54 @@
+// `fetchMLDataBatch`, against a server that sends one file's ML data.
+//
+// ML data that decrypts and gunzips but is not a JSON object is reported with
+// the reason, as ML data that does not parse is, and is never returned as a
+// payload: the library would fail to store it.
+
+import { gzipSync } from "node:zlib";
+import sodium from "libsodium-wrappers-sumo";
+import { beforeAll, describe, expect, it } from "vitest";
+
+import { ApiClient } from "../../src/api/client.js";
+import { encryptBlob, init, toBase64 } from "../../src/crypto/index.js";
+import { fetchMLDataBatch } from "../../src/mldata-fetch.js";
+
+beforeAll(async () => {
+    await init();
+    await sodium.ready;
+});
+
+// Fetch the ML data of file 1 from a server that sends `json`, gzipped and
+// encrypted with the file's key.
+const fetchServing = async (json: string) => {
+    const key = sodium.crypto_secretstream_xchacha20poly1305_keygen();
+    const { header, ciphertext } = encryptBlob(
+        new Uint8Array(gzipSync(json)),
+        key,
+    );
+    const data = [
+        {
+            fileID: 1,
+            encryptedData: toBase64(ciphertext),
+            decryptionHeader: toBase64(header),
+        },
+    ];
+    const api = new ApiClient({
+        fetch: async () =>
+            new Response(JSON.stringify({ data }), {
+                headers: { "content-type": "application/json" },
+            }),
+    });
+    return fetchMLDataBatch(api, [1], new Map([[1, key]]));
+};
+
+describe("fetchMLDataBatch", () => {
+    it.each(["null", "5", "[]"])(
+        "gives a reason for ML data that is %s",
+        async (json) => {
+            const { payloads, errors } = await fetchServing(json);
+
+            expect(payloads.size).toBe(0);
+            expect(errors.get(1)).toEqual(new Error("not a JSON object"));
+        },
+    );
+});

test/retry/retry.test.ts

@@ -77,6 +77,14 @@ const recordingSleep = (): {
 const errnoError = (code: string, message = code): Error =>
     Object.assign(new Error(message), { code });
 
+/**
+ * An error shaped like what bun's `fetch` rejects with for a failure below
+ * HTTP: a plain `Error` with bun's own message, a `code`, an `errno` of 0, and
+ * no `cause`.
+ */
+const bunError = (code: string, message: string): Error =>
+    Object.assign(new Error(message), { code, errno: 0 });
+
 // ---------------------------------------------------------------------------
 // Classification
 // ---------------------------------------------------------------------------
@@ -244,6 +252,82 @@ describe("isRetryable: transport failures", () => {
     });
 });
 
+describe("isRetryable and isSafeToReplay: bun's fetch errors", () => {
+    // The binary `make install` builds runs quak under bun, whose `fetch`
+    // does not reject as Node's does. Each shape here was found by running
+    // that binary, built with bun 1.3.13, against listeners on 127.0.0.1 and
+    // a name under `.invalid`.
+
+    it("retries a refused connection, and does not replay it", () => {
+        // Bun gives the same error when the TLS handshake fails, and when a
+        // new connection it opened to resend a request is refused. That
+        // request may already have reached the server on a connection bun
+        // kept open from an earlier request.
+        const refused = bunError(
+            "ConnectionRefused",
+            "Unable to connect. Is the computer able to access the url?",
+        );
+        expect(isRetryable(refused)).toBe(true);
+        expect(isSafeToReplay(refused)).toBe(false);
+    });
+
+    it("retries a name that does not resolve, with either code bun gives, and does not replay it", () => {
+        // For the same name bun gave `ConnectionRefused` on some requests and
+        // `FailedToOpenSocket` on others. Either can also come from a new
+        // connection bun opened to resend a request, as above.
+        for (const unresolved of [
+            bunError(
+                "ConnectionRefused",
+                "Unable to connect. Is the computer able to access the url?",
+            ),
+            bunError(
+                "FailedToOpenSocket",
+                "Was there a typo in the url or port?",
+            ),
+        ]) {
+            expect(isRetryable(unresolved)).toBe(true);
+            expect(isSafeToReplay(unresolved)).toBe(false);
+        }
+    });
+
+    it("retries a connection closed before the response ended, and does not replay it", () => {
+        // Bun gives Node's errno, `ECONNRESET`, for a reset and for a close,
+        // before the response or part-way through the body. The request may
+        // already have reached the server.
+        const reset = bunError(
+            "ECONNRESET",
+            "The socket connection was closed unexpectedly. For more information, pass `verbose: true` in the second argument to fetch()",
+        );
+        expect(isRetryable(reset)).toBe(true);
+        expect(isSafeToReplay(reset)).toBe(false);
+    });
+
+    it("retries a request that ran out of time, and does not replay it", () => {
+        // What bun's fetch rejects with when `AbortSignal.timeout()`, which
+        // bounds each JSON request, fires.
+        const timedOut = new DOMException(
+            "The operation timed out.",
+            "TimeoutError",
+        );
+        expect(isRetryable(timedOut)).toBe(true);
+        expect(isSafeToReplay(timedOut)).toBe(false);
+    });
+
+    it("does not retry a certificate bun rejects", () => {
+        // A certificate that fails verification fails the same way however
+        // often it is asked. Under Node it is retried only because every
+        // `TypeError` is.
+        expect(
+            isRetryable(
+                bunError(
+                    "DEPTH_ZERO_SELF_SIGNED_CERT",
+                    "self signed certificate",
+                ),
+            ),
+        ).toBe(false);
+    });
+});
+
 describe("isRetryable: stream truncation versus corruption", () => {
     it("retries a truncated stream", () => {
         // Truncation is a transfer that stopped early. The bytes that did