Build the binary without cgo (closes #545)

make build now builds with CGO_ENABLED=0, and the image builds through
it. The binary's only cgo code was mattn/go-sqlite3, which
gorm.io/driver/sqlite imports and webhooker never opens: every SQLite
file is opened with modernc.org/sqlite and the connection handed to
gorm. Without cgo, mattn's package compiles to a stub nothing calls.

The Dockerfile builds the binary once. The static relink and its
netgo, osusergo and sqlite_omit_load_extension tags are gone, and with
them the Makefile's GO_LDFLAGS and GO_TAGS, which only the relink used.
The test phase keeps cgo for -race. No README or Dockerfile sentence
says the binary needs cgo any more.

Model: opus-5-5

Dockerfile

@@ -76,8 +76,6 @@ RUN --network=none node_modules/.bin/prettier --check '**/*.md'
 # depends on it.
 #
 # golangci/golangci-lint:v2.14.0 (Debian-based), 2026-09-24
-# Using Debian-based image because mattn/go-sqlite3 (CGO) does not
-# compile on Alpine musl (off64_t is a glibc type).
 FROM golangci/golangci-lint:v2.14.0@sha256:ad862ba6b3798cbe0fd9fd7408d498fd74fbd2623a92406b2fd3898faf0bf98f AS lint
 
 WORKDIR /src
@@ -167,9 +165,7 @@ RUN --mount=type=tmpfs,target=/tmp/tests,size=512m \
 
 # Build stage
 # golang:1.26.1-bookworm (Debian-based), 2026-03-17
-# Using Debian-based image because gorm.io/driver/sqlite pulls in
+# The image ships git and make, which the version step below uses.
-# mattn/go-sqlite3 (CGO), which does not compile on Alpine musl. The image
-# ships git and make, which the version step below uses.
 FROM golang:1.26.1-bookworm@sha256:4465644228bc2857a954b092167e12aa59c006a3492282a6c820bf4755fd64a4 AS builder
 
 # Nothing is wanted from the lint and test phases or from the stylesheet and
@@ -211,20 +207,11 @@ RUN version="$(make version VERSION="$VERSION")"; \
     fi
 
 # Builds through the Makefile's build target, which runs script/assets
-# (Alpine.js, extracted from its tarball in 3p/) first.
+# (Alpine.js, extracted from its tarball in 3p/) first. The target builds
+# with cgo off, so the binary is statically linked and runs on the Alpine
+# runtime image below as it is.
 RUN make build VERSION="$VERSION"
 
-# Rebuild with static linking for Alpine runtime.
-# make build already verified compilation.
-# The CGO binary from `make build` is dynamically linked against glibc,
-# which doesn't exist on Alpine (musl). Rebuild with static linking so
-# the binary runs on Alpine without glibc.
-#
-# The static flags go in through GO_LDFLAGS rather than a -ldflags of
-# their own: the build target composes them with the -X that stamps the
-# version, so this relink cannot silently drop the stamp.
-RUN CGO_ENABLED=1 make build VERSION="$VERSION" GO_LDFLAGS='-extldflags "-static"'
-
 # Runtime stage
 # alpine:3.21, 2026-03-17
 FROM alpine:3.21@sha256:c3f8e73fdb79deaebaa2037150150191b9dcbfba68b4a46d70103204c53f4709
@@ -268,8 +255,17 @@ EXPOSE 8080
 # port.
 ENV BIND_ADDRESS=0.0.0.0
 
+# The probe requests the container's own address, not loopback, so a
+# listener that answers only on loopback (an empty or loopback
+# BIND_ADDRESS) fails it. On a network with IPv4, hostname -i prints the
+# container's IPv4 address, which 0.0.0.0 and :: both answer. On an
+# IPv6-only network it prints none, and the probe requests the IPv6
+# address /etc/hosts lists under the container's hostname instead, which
+# only :: answers.
 HEALTHCHECK --interval=30s --timeout=3s --start-period=5s --retries=3 \
-    CMD wget --no-verbose --tries=1 --spider http://localhost:8080/.well-known/healthcheck || exit 1
+    CMD addr="$(hostname -i 2>/dev/null | cut -d ' ' -f 1)"; \
+        [ -n "$addr" ] || addr="[$(getent hosts "$(hostname)" | cut -d ' ' -f 1)]"; \
+        wget --no-verbose --tries=1 --spider "http://$addr:8080/.well-known/healthcheck" || exit 1
 
 ENTRYPOINT ["/usr/local/bin/docker-entrypoint.sh"]
 CMD ["/app/webhooker"]

Makefile

@@ -16,15 +16,6 @@ VERSION ?= $(shell script/version)
 # exists to correct.
 override VERSION := $(or $(strip $(VERSION)),$(shell script/version))
 
-# Extra linker flags for the build target. The static relink in the
-# Dockerfile adds -extldflags here rather than passing its own -ldflags,
-# so composing flags cannot drop the version stamp.
-#
-# The build target itself always passes -trimpath and -s -w, as the Go
-# Dockerfile in REPO_POLICIES.md does: no build paths, symbol table or
-# debug information in the binary.
-GO_LDFLAGS ?=
-
 bootstrap:
 	@script/bootstrap
 
@@ -52,8 +43,12 @@ fmt-check:
 check:
 	@script/check
 
+# Built with cgo off, -trimpath and -s -w, as the Go Dockerfile in
+# REPO_POLICIES.md does: the binary carries no build paths, symbol table or
+# debug information, and on Linux, the image's included, it is statically
+# linked.
 build: assets
-	go build -trimpath -ldflags '$(strip -s -w -X main.version=$(VERSION) $(GO_LDFLAGS))' -o bin/webhooker ./cmd/webhooker
+	CGO_ENABLED=0 go build -trimpath -ldflags '-s -w -X main.version=$(VERSION)' -o bin/webhooker ./cmd/webhooker
 
 run: build
 	./bin/webhooker

README.md

@@ -310,16 +310,17 @@ the namespace, whatever the process bound. Exposure is decided at the publish
 flag instead — `-p 127.0.0.1:8080:8080` rather than `-p 8080:8080` — which is
 the operator's to choose and is what [Running with Docker](#running-with-docker)
 shows. A loopback bind inside a container buys nothing and makes the process
-unreachable through its own published port.
+unreachable through its own published port. A container on an IPv6-only Docker
+network has no IPv4 address for `0.0.0.0` to answer on, so there it needs
+`BIND_ADDRESS=::`.
 
 The value must be an IP address literal:
 
 - `127.0.0.1` — loopback only (the binary's default). Use this with a reverse
   proxy on the same host.
-- `0.0.0.0` — every IPv4 address. The image's default; on a bare host, only
+- `0.0.0.0` — every IPv4 address and no IPv6 address. The image's default; on a
-  behind a firewall on the port.
+  bare host, only behind a firewall on the port.
-- `::` — every address, IPv6 and (on Linux, with the default
+- `::` — every IPv6 address and every IPv4 address.
-  `net.ipv6.bindv6only=0`) IPv4 as well.
 - A specific address such as `10.0.0.5` — that interface only.
 
 An **empty** value is treated as unset, as everywhere else here, and takes the
@@ -340,7 +341,8 @@ machine that has no such interface — is a valid literal, so it reaches the
 listener and fails there. The process logs the bind error and exits non-zero
 rather than staying up with nothing listening. The effective value is in the
 `bindAddress` field of the startup log line, which is the way to check what a
-running deployment actually bound.
+running deployment actually bound. On a bare host, `ss -ltn` lists the socket
+itself as `0.0.0.0:8080` for a `0.0.0.0` bind and as `*:8080` for `::`.
 
 #### Metrics credentials
 
@@ -465,10 +467,7 @@ idle timestamp is rewritten at most once per tenth of the idle window rather
 than on every request, which means a session may expire up to 10% early relative
 to the user's true last request, but never late.
 
-Both clocks are anchored by timestamps stored in the session cookie. Sessions
+Both clocks are anchored by timestamps stored in the session cookie.
-issued before this feature existed carry neither, so they are treated as
-expired: upgrading to a build that has it logs every existing session out once,
-and those users sign in again.
 
 #### Invalid values abort startup
 
@@ -671,9 +670,9 @@ Either set a literal or leave the variable out entirely.
 Overriding `BIND_ADDRESS` to a loopback address in a container — by that route
 or deliberately — makes the container unreachable from outside its namespace
 even with `-p`. The published port answers nothing, and the health check fails
-too: it requests `http://localhost:8080`, `localhost` resolves to `::1` first,
+too: it requests the container's own address (its IPv4 address, or its IPv6
-and a `127.0.0.1` bind is not listening there. The container then goes
+address on an IPv6-only network), where a loopback bind is not listening. The
-`unhealthy` about **65 seconds** after start — from
+container then goes `unhealthy` about **65 seconds** after start — from
 `HEALTHCHECK --start-period=5s --interval=30s --retries=3`, so failing probes at
 5s, 35s and 65s, and `unhealthy` on the third. (Docker's probe cadence during
 the start period has changed between versions; re-derive from those three values
@@ -702,11 +701,11 @@ app as that user.
 `webhooker.db` holds target configuration in plaintext — bearer tokens, API
 keys, Slack webhook URLs — along with the session encryption key, so webhooker
 creates every SQLite file it owns `0600`: each database and both of its `-wal`
-and `-shm` sidecars, across all three tiers. Files an earlier build left `0644`
+and `-shm` sidecars, across all three tiers. Any of them found at another mode
-are tightened when they are opened. The directory's `0750` is defence in depth —
+is set to `0600` when the database is opened. The directory's `0750` is defence
-it stops other local users listing the directory and learning your webhook UUIDs
+in depth — it stops other local users listing the directory and learning your
-from the `events-{uuid}.db` filenames — not the barrier protecting the
+webhook UUIDs from the `events-{uuid}.db` filenames — not the barrier protecting
-credentials.
+the credentials.
 
 ### Running under upaas
 
@@ -729,8 +728,8 @@ credentials.
       address (directly, or through anything that can rewrite source addresses,
       such as NAT or a published container port). The `remoteIP` field of the
       `http request` log line for a request that came through the proxy shows
-      that address; the health check's own lines show `::1`. See
+      that address; the health check's own lines show the container's own
-      [Trusted proxies](#trusted-proxies).
+      address. See [Trusted proxies](#trusted-proxies).
     - Leave `BIND_ADDRESS` and `DATA_DIR` unset: the image sets `BIND_ADDRESS`
       to `0.0.0.0`, and `DATA_DIR` defaults to `/var/lib/webhooker`.
     - Everything else is optional; see [Configuration](#configuration).
@@ -805,7 +804,12 @@ comes up, serves pages, and is broken in a way nothing reports.
    proxy's log names it regardless. nginx's default `combined` format already
    logs `$remote_addr`; do not replace it with one that drops the client
    address, and retain those logs as long as you would want to answer a question
-   about traffic.
+   about traffic. That log also records the full request path, so it holds the
+   entrypoint UUID of every event a sender posts. webhooker's own log does not,
+   apart from the line a `log` target writes. Protect the proxy's log as you
+   would the entrypoint URLs themselves: anyone who can read it can post events
+   to every entrypoint in it (see
+   [The entrypoint URL is the authentication secret](#the-entrypoint-url-is-the-authentication-secret)).
 
 ### nginx
 
@@ -831,7 +835,9 @@ server {
     client_max_body_size 1m;
 
     # $remote_addr is the client. webhooker's own log names it, as
-    # clientIP, only while TRUSTED_PROXIES covers this proxy.
+    # clientIP, only while TRUSTED_PROXIES covers this proxy. The
+    # request path in this log holds the entrypoint UUID, so protect
+    # the file like the entrypoint URL itself.
     access_log /var/log/nginx/webhooker.access.log combined;
 
     location / {
@@ -911,13 +917,14 @@ as `example.com`, not `hooks.example.com`. A listed domain is HTTPS-only in
 every browser that ships the list, even for users who have never visited it, and
 removal takes months to reach those browsers, so treat a listing as permanent.
 
-Setting `HSTS_MAX_AGE=0` later stops the header being sent, but a browser that
+Setting `HSTS_MAX_AGE=0` stops the header being sent, but a browser that already
-already received it over HTTPS does not forget it. That browser keeps refusing
+received it over HTTPS does not forget it: it keeps refusing plaintext HTTP
-plaintext HTTP until the `max-age` it saw runs out. Only a `max-age=0` header
+until the `max-age` it saw runs out. Each header a browser receives over HTTPS
-received over HTTPS clears it sooner. webhooker never sends one; a reverse proxy
+replaces the `max-age` it stored, so to end the policy sooner, set a short
-in front of it can.
+`HSTS_MAX_AGE` such as `1s` and keep it until returning browsers have picked it
+up, or have a reverse proxy send `max-age=0`.
 
-## Backup, Restore, and Upgrades
+## Backup and Restore
 
 ### What to back up
 
@@ -1070,59 +1077,14 @@ beside it, or wait until there are none.
 
 4. Start the service. The container gives the directory and the restored files
    to the `webhooker` user before the app starts, whoever restored them (see
-   [Running with Docker](#running-with-docker)). `AutoMigrate` runs against each
+   [Running with Docker](#running-with-docker)).
-   restored database as it is opened.
-
-### Upgrades
-
-**Back up before every upgrade.** Every start runs GORM `AutoMigrate`
-unconditionally against whatever files it finds:
-
-- the main database on connect — `Setting`, `User`, `APIKey`, `Webhook`,
-  `Entrypoint`, `Target`
-- each event database when it is lazily opened — `Event`, `Delivery`,
-  `DeliveryResult`, `EventTotals`, `TargetTotals`, `EntrypointTotals`
-- each archive database on every open and reopen
-
-There is no schema version table, no migration ledger, and no down migrations.
-Nothing in the files records which version wrote them, and no code path undoes a
-migration.
-
-Upgrade procedure:
-
-1. Stop the service.
-2. Back up `DATA_DIR` using one of the procedures above.
-3. Pull the new image and start it.
-4. Confirm `database migrations completed` in the logs before putting traffic
-   back on it.
-5. Confirm the new build is the one running:
-   `curl -s http://host:8080/.well-known/healthcheck` reports the version it was
-   stamped with (see [Version stamping](#version-stamping)).
-
-**Upgrading past the introduction of `BIND_ADDRESS`:** earlier versions always
-bound every interface. **Container deployments are unaffected** — the image
-ships `ENV BIND_ADDRESS=0.0.0.0`, so a `docker run` or Compose service that
-worked before still works with nothing changed.
-
-A **bare binary** is the case that changes: the listener now binds `127.0.0.1`
-unless `BIND_ADDRESS` says otherwise, so a deployment that relied on reaching it
-from another host becomes unreachable until it sets the address the proxy
-connects to. Check the `bindAddress` field of the startup log to see what a
-running process bound. See [Bind address](#bind-address).
-
-**Downgrade is unsupported.** Once a newer binary has migrated the files there
-is no way to move them back. `AutoMigrate` is additive — it adds tables, columns
-and indexes and never drops or rewrites them — so an older binary will generally
-open migrated files and appear to work while writing against a schema it does
-not know about. The failure mode is silent divergence, not a startup error. The
-only supported way back to an older version is restoring the pre-upgrade backup,
-which discards everything received since that backup was taken.
 
 ### Version stamping
 
 The binary reports its version at `/.well-known/healthcheck` (the `version`
 field), in the UI footer, and in the startup log line (`msg=starting`,
 `version=...`). It is also the Sentry release name, as `webhooker-{version}`.
+`curl -s http://host:8080/.well-known/healthcheck` shows which build is running.
 The value is stamped in at build time by the linker; it is not read from a file
 at runtime, so it identifies the build itself.
 
@@ -1181,26 +1143,6 @@ at rest and restrict who can read them.
 - `events-{uuid}.db` and `archive-….db` hold the **full payload body and
   headers** of every event as received, including whatever the sending service
   put in them — tokens, signatures, personal data.
-- Event databases written before
-  [issue #206](https://git.eeqj.de/sneak/webhooker/issues/206) was fixed **also
-  contain target credentials**: a GORM association upsert on the delivery and
-  retry write path copied `targets` rows, `config` included, into the
-  per-webhook database. For a Slack target the `webhookUrl` _is_ the bearer
-  credential, and an `http` target's URL can embed userinfo. This version never
-  writes those rows; the first time it opens such a file it deletes them and
-  vacuums the file, which removes the credential bytes rather than only
-  unlinking the rows. Deleting alone would not: the bytes stay readable in the
-  file's free pages until it is rewritten. The sweep is recorded in the file's
-  `user_version` only once the vacuum returns, so a sweep that fails or is
-  interrupted fails the open and is retried on the next one, and a file this
-  version has opened without error holds no leaked rows and no recoverable bytes
-  from them. On upgrade this rewrites each existing `events-{uuid}.db` once, on
-  its first open. Two cases still hand over live delivery destinations: a backup
-  taken from an older build, and a backup of a file this version has not yet
-  opened successfully. Copies already made stay affected — the sweep only
-  rewrites the file it opens, and freed blocks may persist in filesystem
-  snapshots and on the underlying storage. Rotate any target credential that was
-  in a backup you cannot account for.
 - `webhooker.db` stores target config **unencrypted**, tracked at
   [issue #212](https://git.eeqj.de/sneak/webhooker/issues/212), next to the
   session encryption key and the Argon2id password hashes.
@@ -1227,7 +1169,16 @@ What that means for an operator:
 
 - **The URL is a capability, so treat it as a secret.** Keep it out of logs,
   ticket bodies, chat messages and screenshots. Anyone who reads it anywhere can
-  post events as that sender.
+  post events as that sender. Apart from a `log` target's line, webhooker's own
+  log lines never carry the UUID: they name a request to the receiver by the
+  route pattern `/h/{uuid}`, the `webhook request received` line names the
+  entrypoint by its `entrypoint_id`, and the access log keeps only the scheme
+  and host of a `Referer`. A `log` target writes the request's headers as
+  received, so a `Referer` naming the entrypoint URL appears in its line, as it
+  does when a sender has followed a redirect from `http://` to `https://`. The
+  reverse proxy's access log does carry the UUID, in the request path, so it
+  must be protected like the URL itself (see
+  [Deployment behind a reverse proxy](#deployment-behind-a-reverse-proxy)).
 - **Rotating means minting a new entrypoint, not changing a key.** There is no
   way to rotate the UUID in place. To retire one, delete the entrypoint (or
   deactivate it, which answers `410`) and create a new one, then point the
@@ -1302,34 +1253,37 @@ markup. The CSP build runs no expressions, so every Alpine directive in
 `static/js/app.js`: `x-data="collapsible"` and `@click="toggle"`, never
 `x-data="{ open: false }"` or `@click="open = !open"`.
 
-A browser test in `internal/server` loads the webhook page, its edit pages and
+A browser test in `internal/server` loads the sign-in page, the webhook page,
-the event log under the real policy and checks that: the add entrypoint form
+its edit pages, the event log, the webhook list, the settings page, the profile
-stays hidden until Add is clicked; for every target type, the targets section's
+page and the `404` error page shown for a path that does not exist, under the
-Add shows only a choice of type with Next and Cancel, Next shows only that
+real policy and checks that: the add entrypoint form stays hidden until Add is
-type's fields (no url field for `database` or `log`), Cancel at either step
+clicked; for every target type, the targets section's Add shows only a choice of
-closes the form, and saving adds the target; a refused target comes back with
+type with Next and Cancel, Next shows only that type's fields (no url field for
-its form open, the values entered and the reason, and after Cancel the next Add
+`database` or `log`), Cancel at either step closes the form, and saving adds the
-starts with an empty form and no reason; a refused save on the target edit page
+target; a refused target comes back with its form open, the values entered and
-and on the webhook edit page comes back with the reason and every value entered;
+the reason, and after Cancel the next Add starts with an empty form and no
-the Copy button beside an entrypoint URL reads "Copied" once clicked; an
+reason; a refused save on the target edit page and on the webhook edit page
-entrypoint's Edit button shows its edit form in place of its description and
+comes back with the reason and every value entered; the Copy button beside an
-hides until the form closes, Cancel hides the form and drops what was typed, as
+entrypoint URL reads "Copied" once clicked; an entrypoint's Edit button shows
-does leaving the page and going back to it, and Save changes the description; of
+its edit form in place of its description and hides until the form closes,
-the recent events on the webhook page only the newest starts expanded, each
+Cancel hides the form and drops what was typed, as does leaving the page and
-expands and collapses, and Open leads to the event's own page; of the events in
+going back to it, and Save changes the description; of the recent events on the
-the event log only the newest starts expanded, and an event there expands and
+webhook page only the newest starts expanded, each expands and collapses, and
-collapses when its row's caret or its ID is clicked, and from the keyboard, but
+Open leads to the event's own page; of the events in the event log only the
-not when its ID is selected with the mouse, and a delivery's attempts inside it
+newest starts expanded, and an event there expands and collapses when its row's
-expand and collapse; and at phone width the menu button opens and closes the
+caret or its ID is clicked, and from the keyboard, but not when its ID is
-mobile menu, and neither the webhook page nor the event log, with a delivery's
+selected with the mouse, and a delivery's attempts inside it expand and
-attempts open, scrolls sideways or cuts anything off at the page's or a card's
+collapse; at phone width the menu button opens and closes the mobile menu, and
-edge. It also fails if the browser reports a console warning or error, an
+neither the webhook page nor the event log, with a delivery's attempts open,
-uncaught exception, or anything the policy refused. `make check` and the image
+scrolls sideways or cuts anything off at the page's or a card's edge; and a
-build lint it but do not run it, and `make test` leaves it out (its file is
+target's Delete asks first, naming the target, sends nothing when the answer is
-built only with the `browser` build tag). Run it with `make test-browser` after
+no and deletes the target when it is yes. It also fails if the browser reports a
-changing `templates/` or `static/js/`: that builds `Dockerfile.browser`, which
+console warning or error, an uncaught exception, or anything the policy refused.
-runs the test in a digest-pinned headless browser image, so the host needs no
+`make check` and the image build lint it but do not run it, and `make test`
-browser.
+leaves it out (its file is built only with the `browser` build tag). Run it with
+`make test-browser` after changing `templates/` or `static/js/`: that builds
+`Dockerfile.browser`, which runs the test in a digest-pinned headless browser
+image, so the host needs no browser.
 
 The package's tarball is committed as `3p/alpinejs-csp-3.14.9.tgz`, byte for
 byte as the npm registry publishes it. It is a dependency, not this repo's build
@@ -1362,7 +1316,9 @@ and commit the regenerated file with the change. Tailwind takes classes only
 from the files that `input.css` names in its `@source` lines; a class written in
 any other file is not generated until that file is named there too. `make check`
 and the image build fail when the committed file differs from what `make css`
-generates. `static/css/style.css` is hand-written and is not generated.
+generates. `static/css/style.css` is hand-written and is not generated. The
+pages' Content-Security-Policy refuses inline styles, so a template styles an
+element with a class, never a `style` attribute.
 
 `make css` runs the Tailwind standalone CLI in Docker, at the version and sha256
 pinned in the Dockerfile's stylesheet stages; it is never installed on the host.
@@ -1384,20 +1340,22 @@ webhooker solves this by acting as a durable intermediary:
    nothing is lost if downstream targets are unavailable.
 
 2. **Guaranteed delivery** — Events are queued for delivery to each configured
-   target. Failed deliveries are retried with configurable backoff. Every
+   target. Failed deliveries are retried with configurable backoff. Delivery
-   delivery attempt is logged with status codes, response bodies, and timing.
+   attempts are logged with status codes, response bodies, and timing.
 
     **That guarantee is at-least-once, not exactly-once.** When a send reaches
     its target but the write recording that outcome fails, the delivery is
     deliberately left in a recoverable state rather than marked done — losing a
     delivery is the worse failure — so the pending sweep picks it up about
     fifteen minutes later, or the next restart does, and the target receives a
-    payload it already got. webhooker adds no delivery identifier of its own to
+    payload it already got. A send still in flight when the service stops is
-    an outbound request, so **make your receiver idempotent** against whatever
+    left the same way: its attempt is not recorded, and the next start sends it
-    the payload itself carries.
+    again (see Shutdown). webhooker adds no delivery identifier of its own to an
+    outbound request, so **make your receiver idempotent** against whatever the
+    payload itself carries.
 
 3. **Observability** — Full request/response logging for every webhook received
-   and every delivery attempted. Prometheus metrics expose volume, latency, and
+   and for delivery attempts. Prometheus metrics expose volume, latency, and
    error rates. The web UI provides real-time visibility into event flow.
 
 4. **Fan-out** — A single incoming webhook can be delivered to multiple targets
@@ -1435,10 +1393,10 @@ It uses:
   management
 - **[go-chi](https://github.com/go-chi/chi)** for HTTP routing
 - **[GORM](https://gorm.io)** for database access with
-  **[modernc.org/sqlite](https://pkg.go.dev/modernc.org/sqlite)** as the runtime
+  **[modernc.org/sqlite](https://pkg.go.dev/modernc.org/sqlite)**, which is pure
-  SQLite driver. Note: `gorm.io/driver/sqlite` transitively depends on
+  Go, as the SQLite driver. `gorm.io/driver/sqlite` also imports
-  `mattn/go-sqlite3`, which requires CGO at build time (see [Docker](#docker)
+  `mattn/go-sqlite3`, whose stub for builds without cgo is linked into the
-  section)
+  binary but never used
 - **[slog](https://pkg.go.dev/log/slog)** (stdlib) for structured logging with
   TTY detection (text for dev, JSON for prod)
 - **[gorilla/sessions](https://github.com/gorilla/sessions)** for encrypted
@@ -1601,8 +1559,7 @@ The cap is not arbitrary: the reaper computes its cutoff as a `time.Duration`,
 an int64 nanosecond count, and a longer period overflows it. An overflowed
 cutoff lands in the future, where it matches every row, so the sweep would
 delete every event the webhook has instead of none. The reaper also clamps the
-value it is given, so a row written by an older version cannot trigger that
+value it reads to the cap, so no stored value can trigger that either.
-either.
 
 #### Entrypoint
 
@@ -1836,7 +1793,7 @@ is what bounds it.
 #### DeliveryResult
 
 The result of a single delivery attempt. Every attempt (including retries) is
-individually logged for full observability.
+individually logged, except one cut off by a stop (see Shutdown).
 
 | Field           | Type    | Description                               |
 | --------------- | ------- | ----------------------------------------- |
@@ -1918,8 +1875,9 @@ Deliveries still `pending` or `retrying` count in neither.
 
 #### Event-tier indexes
 
-These indexes on the per-webhook event databases are declared in the model tags,
+These indexes on the per-webhook event databases are created by
-so `AutoMigrate` creates them on a fresh database:
+`internal/db/migrations/events/001_schema.sql`, and declared in the model tags
+to match:
 
 | Table              | Columns                                                            | Serves                                                                                                                                                                                                                                                                                                                                                                                             |
 | ------------------ | ------------------------------------------------------------------ | -------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- |
@@ -1983,8 +1941,9 @@ and application state:
 - **Targets** — delivery destination configurations
 - **APIKeys** — programmatic access credentials
 
-On first startup the main database is auto-migrated, a session encryption key is
+On first startup the main database is created from its SQL files (see
-generated and stored, and an `admin` user is created.
+[Schema](#schema)), a session encryption key is generated and stored, and an
+`admin` user is created.
 
 **Per-Webhook Event Databases** (`{DATA_DIR}/events-{webhook_uuid}.db`) — each
 webhook gets its own dedicated SQLite file containing:
@@ -1996,9 +1955,10 @@ webhook gets its own dedicated SQLite file containing:
   the above, the deliveries per target, and each entrypoint's last event, kept
   through retention
 
-Per-webhook databases are created automatically when a webhook is created. They
+Per-webhook databases are created automatically when a webhook is created, from
-are managed by the `WebhookDBManager` component, which handles connection
+the SQL files in `internal/db/migrations/events/`. They hold no copy of the main
-pooling, lazy opening, migrations, and cleanup.
+database's tables. They are managed by the `WebhookDBManager` component, which
+handles connection pooling, lazy opening, migrations, and cleanup.
 
 A per-webhook database that is missing or zero-length later means its webhook's
 events and pending deliveries are gone. The next time it is opened, an empty one
@@ -2082,14 +2042,15 @@ the service again.
 After each write the archive handle is closed and reopened, debounced to at most
 once per second, so an operator can move the archive file away for offline
 archiving without stopping the service; a moved or removed archive file is
-recreated automatically on the next write. A zero-length archive file is written
+recreated automatically on the next write, from the SQL files in
-to as a new archive: SQLite opens it as an empty database, so it holds nothing
+`internal/db/migrations/archive/`. A zero-length archive file is written to as a
-to lose. An optional `expiry` in the target's config JSON (e.g.
+new archive: SQLite opens it as an empty database, so it holds nothing to lose.
-`{"expiry":"720h"}`) is validated when the target is created — the default
+An optional `expiry` in the target's config JSON (e.g. `{"expiry":"720h"}`) is
-(unset or the literal `never`) keeps rows forever — and rows older than the
+validated when the target is created — the default (unset or the literal
-expiry are pruned each time the archive is (re)opened. An archive write failure
+`never`) keeps rows forever — and rows older than the expiry are pruned each
-is never silent success: the delivery records a failed attempt with the error
+time the archive is (re)opened. An archive write failure is never silent
-and is marked failed.
+success: the delivery records a failed attempt with the error and is marked
+failed.
 
 Because reopens only happen on writes, an archive whose target has stopped
 receiving events would never be pruned. A background **archive sweeper** closes
@@ -2174,9 +2135,22 @@ endpoint. That is the JSON key; the error text for a missing one reads
 `webhook_url is required`, which is the message, not the key.
 
 The database uses the
-[modernc.org/sqlite](https://pkg.go.dev/modernc.org/sqlite) driver at runtime,
+[modernc.org/sqlite](https://pkg.go.dev/modernc.org/sqlite) driver.
-though CGO is required at build time due to the transitive `mattn/go-sqlite3`
+
-dependency from `gorm.io/driver/sqlite`.
+#### Schema
+
+Each kind of database file has its schema in SQL files built into the binary,
+one directory per kind under `internal/db/migrations/`. In each,
+`000_migration.sql` creates the `schema_migrations` table, which records the
+number of every file the database has had, and `001_schema.sql` holds the whole
+schema. Opening a database runs, in one transaction, every file in its directory
+that `schema_migrations` does not list:
+
+- `main/` for the main database, on connect — settings, users, API keys,
+  webhooks, entrypoints and targets
+- `events/` for each event database, when it is lazily opened — events,
+  deliveries, delivery results and their running totals
+- `archive/` for each archive database, on every open and reopen
 
 ### Request Flow
 
@@ -2255,12 +2229,16 @@ This means:
   buffer rather than spawning more goroutines.
 - **Independent results** — each worker records its own delivery result in the
   per-webhook database without coordination.
-- **Graceful shutdown** — cancel the context, workers finish their current task
+- **Graceful shutdown** — the stop hook cancels the workers' context. A worker
-  and exit. The stop hook waits for the pool via `lifecycle.WaitForShutdown`,
+  may still take tasks already queued before it exits. An `http` or `slack`
-  which bounds that wait by fx's stop timeout rather than blocking forever on a
+  request in flight is cut off, and one started after the cancel fails at once.
-  wedged worker. On timeout it logs at `ERROR` and returns an error, and the
+  Neither is recorded or counted in the metrics, and the next start sends the
-  goroutines that did not finish are still running — an unclean shutdown is
+  delivery again (see Shutdown). A `database` or `log` task runs as normal. The
-  reported rather than hidden.
+  stop hook waits for the pool via `lifecycle.WaitForShutdown`, which bounds
+  that wait by fx's stop timeout rather than blocking forever on a wedged
+  worker. On timeout it logs at `ERROR` and returns an error, and the goroutines
+  that did not finish are still running — an unclean shutdown is reported rather
+  than hidden.
 
 **Recovery paths:**
 
@@ -2366,7 +2344,7 @@ getting anywhere.
 | Metric                                 | Type      | Meaning                                                                                                                                                                                                                                                                                                   |
 | -------------------------------------- | --------- | --------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- |
 | `webhooker_events_received_total`      | counter   | Events received and durably stored. Compare against the delivery counters on one dashboard                                                                                                                                                                                                                |
-| `webhooker_delivery_attempts_total`    | counter   | Delivery attempts actually dispatched to a target. A delivery a circuit breaker refused is not one: it is counted as a retry instead, but only when the refusal moves it into `retrying`                                                                                                                  |
+| `webhooker_delivery_attempts_total`    | counter   | Delivery attempts actually dispatched to a target. A delivery a circuit breaker refused is not one: it is counted as a retry instead, but only when the refusal moves it into `retrying`. An attempt cut off by a stop is not counted (see Shutdown)                                                      |
 | `webhooker_deliveries_succeeded_total` | counter   | Deliveries that reached `delivered`                                                                                                                                                                                                                                                                       |
 | `webhooker_deliveries_failed_total`    | counter   | Deliveries that failed terminally and will not be retried                                                                                                                                                                                                                                                 |
 | `webhooker_delivery_retries_total`     | counter   | Deliveries put back into `retrying`                                                                                                                                                                                                                                                                       |
@@ -2465,27 +2443,32 @@ log volume it caps rather than eliminates. A path that names no entrypoint is
 recorded by the handler at `DEBUG`, and the aggregate limiter logs its own
 rejections at `DEBUG` and without the path, so neither appears at all under the
 default level. The per-entrypoint limiter is the loud one: it logs every
-rejection at `WARN` with the request path, which on this route is
+rejection at `WARN`, naming the route pattern `/h/{uuid}` rather than the
-attacker-controlled text. A client hammering a single invented path is served
+request path. A client hammering a single invented path is served
 `RECEIVER_RATE_LIMIT` requests and has the rest of its aggregate budget rejected
 there, so the aggregate limit is what bounds the _number_ of those `WARN` lines
 — to under ten times `RECEIVER_RATE_LIMIT` per minute per client IP, 1080 at the
-defaults, where before it there was no bound at all. Their _width_ is bounded by
+defaults, where before it there was no bound at all. The access log is bounded
-the field budgets below, the same ones the access log spends. The access log is
+by neither limit: every request is recorded once at `INFO`, served or rejected
-bounded by neither limit: every request is recorded once at `INFO`, served or
+alike.
-rejected alike.
 
-What the access log does bound is the _content_ of those lines. A 3xx or 4xx
+What the access log does bound is the _content_ of those lines. On the receiver
-response logs the chi route pattern — `/h/{uuid}`, `/user/{username}//`, or the
+route it logs the route pattern `/h/{uuid}` whatever the status, because the
-literal `(unmatched)` when the request hit no route at all — in place of the
+concrete path carries the entrypoint UUID, which is the entrypoint's credential
-concrete URL. Those are the outcomes an unauthenticated client can drive for
+(see
-free: 404 and 429 on any invented receiver path, a login redirect on any
+[The entrypoint URL is the authentication secret](#the-entrypoint-url-is-the-authentication-secret)).
-invented profile path. Logging the URL there would let a flood write text of its
+A request chi never routed, because it matched no route or because a middleware
-own choosing, at a length of its own choosing, into the log. 2xx and 5xx
+answered it first, logs the literal `(unmatched)` whatever the status. In the
-responses keep the concrete path — a success resolved against a static route or
+`dev` environment the CORS middleware answers a preflight that way, so a
-against the operator's own data (on the receiver, against a stored entrypoint
+preflight to the receiver does not log its UUID either. On every other route, a
-UUID), and a 5xx is a bug in this service, where the exact path is the evidence
+3xx or 4xx response logs the chi route pattern, such as `/user/{username}//`, in
-and no client can provoke one at will.
+place of the concrete URL. Those are the outcomes an unauthenticated client can
+drive for free: a 404 on any invented path, a login redirect on any invented
+profile path. Logging the URL there would let a flood write text of its own
+choosing, at a length of its own choosing, into the log. 2xx and 5xx responses
+keep the concrete path — a success resolved against a static route or against
+the operator's own data, and a 5xx is a bug in this service, where the exact
+path is the evidence and no client can provoke one at will.
 
 The query string is never logged; it is replaced by the fixed marker
 `?(redacted)`. It is client-chosen on every route, and
@@ -2497,6 +2480,11 @@ to, `notice`, which names the line a page shows after an action, and the event
 log's `show`, which picks the events it lists. A query string sent to an
 entrypoint is not lost either: the event stores it, and the event log shows it.
 
+The `referer` field keeps only the scheme and host of the request's `Referer`,
+and logs `(redacted)` for one without them, because its path can be an
+entrypoint URL: a sender that follows the `http://` to `https://` redirect in
+the nginx example sends the URL it was redirected from.
+
 Client-supplied request content does not leave the host by the other route
 either. The Sentry SDK attaches the request to every event it captures,
 independently of the access log, and `SendDefaultPII=false` does not cover all
@@ -2511,9 +2499,10 @@ The same hook rewrites the request URL. The SDK builds it as
 `scheme://host/path` from the concrete path, which on the receiver route is
 `/h/<uuid>` in full — and that UUID is a write capability, not an identifier:
 anyone holding it can post events this service accepts and its targets then
-deliver. A tracker has its own retention, access control and deletion policy, so
+deliver. What is sent is the chi route pattern instead: `http://host/h/{uuid}`.
-the rule the access log follows above does not carry across that boundary. What
+A tracker has its own retention, access control and deletion policy, so the
-is sent is the chi route pattern instead: `http://host/h/{uuid}`.
+pattern is sent on every route, including the 2xx and 5xx responses outside the
+receiver for which the access log keeps the concrete path.
 
 The scheme and the host are kept, and everything else in the URL is discarded
 rather than edited, so a future SDK version that starts appending a query string
@@ -2550,10 +2539,23 @@ Where the pattern is out of reach — the transaction dispatch, an event capture
 outside the router, or a request that matched no route — the fallback is never
 the concrete path. The path becomes the literal `/(redacted)`, so the URL reads
 `http://host/(redacted)`; a URL the rewrite cannot parse into a scheme is
-withheld whole. A transaction event additionally carries the SDK's own
+withheld whole.
-`METHOD /path` name, built from the concrete path as well; it is rewritten on
+
-the same terms, to `POST /h/{uuid}` where the pattern is known and
+A transaction event also carries a name, which the SDK builds as `METHOD /path`
-`POST /(redacted)` where it is not.
+from the concrete path and gives the transaction source `url`. Sentry groups
+transactions by name, so withholding it would put every request in the service
+into one group per method, such as `POST /(redacted)`. Inside each Sentry
+handler, a second middleware renames the transaction once the request has been
+routed, to the method and the route pattern (`POST /h/{uuid}`), and sets its
+source to `route`. The hook keeps a name whose source is `route` and rewrites
+any other to `METHOD /(redacted)`. A path that matches nothing at the top level
+has no pattern, so its transaction keeps the SDK's name and the hook withholds
+it. Inside a route group, a path that matches no route in the group, and a
+request the group's own middleware turns away before routing (a CSRF refusal,
+the redirect to `/pages/login`), are named after the group's pattern, such as
+`GET /pages/*` or `POST /hook/{sourceID}/*`, and the hook keeps that name. The
+URL on a transaction event still falls back to `http://host/(redacted)`, because
+the hook has no request there.
 
 The headers are an allowlist for the same reason the rules above are
 unconditional: the SDK's own filter removes four names and passes everything
@@ -2629,9 +2631,7 @@ than it:
 | ------------------------------------------ | ------- | ------------------- | ------------------------------------------ |
 | `request body exceeds limit` (413)         | `WARN`  | path, method        | yes — `MaxBodySize` precedes `RequireAuth` |
 | `csrf: token validation failed` (403)      | `WARN`  | path, method        | yes — `CSRF` precedes `RequireAuth`        |
-| `... rate limit exceeded` (429)            | `WARN`  | path                | yes, on the receiver                       |
 | `auth middleware: unauthenticated request` | `DEBUG` | path, method        | yes, by definition                         |
-| `entrypoint not found`                     | `DEBUG` | entrypoint UUID     | yes, on the receiver                       |
 | `user not found` / `invalid password`      | `DEBUG` | username            | yes, on the sign-in form                   |
 | `login failure limit exceeded` (429)       | `WARN`  | path                | yes, on the sign-in form                   |
 | `password verification capacity exhausted` | `WARN`  | path                | yes, on the sign-in form                   |
@@ -2659,13 +2659,12 @@ wider of the two handlers spends more on than the client spent sending them.
 Every case holds each line to the 2,560-byte ceiling. That per-line ceiling is
 what the figure above states, and every row establishes it.
 
-Three of the sites go further and bound the whole flood's output — the total
+Two of the sites go further and bound the whole flood's output — the total bytes
-bytes a run of distinct invented values wrote, which is the shape an operator
+a run of distinct invented values wrote, which is the shape an operator sizing
-sizing storage cares about. They are `request body exceeds limit`
+storage cares about. They are `request body exceeds limit`
-(`TestMaxBodySize_FloodOfOversizePathsDoesNotGrowTheLog`),
+(`TestMaxBodySize_FloodOfOversizePathsDoesNotGrowTheLog`) and `user not found`
-`entrypoint not found` and `user not found` (the last two through
+(through `assertBoundedFlood`). The other rows carry no aggregate assertion; the
-`assertBoundedFlood`). The other rows carry no aggregate assertion; the per-line
+per-line ceiling is what they establish.
-ceiling is what they establish.
 
 `internal/logfield/logfield_test.go` measures the per-rune charge against what
 the handlers really emit, over roughly 3,000 code points on each, so an
@@ -2683,7 +2682,7 @@ everything else, so its lines take the level the operator set and the handler
 `internal/logger` selected, and every value it emits is spent through the same
 `internal/logfield` budget. A record-not-found is not logged as an error: it is
 the expected outcome on both of those paths, and each handler already records
-its own miss at `DEBUG` — bounded, per the table above — without the SQL. Slow
+its own miss at `DEBUG`, without the SQL and within the ceiling above. Slow
 statements are kept, at `WARN`, above the same 200 ms threshold GORM used and
 with the statement bounded, because that report is the one thing GORM's logger
 gave an operator that nothing else here does. The adapter orders its cases
@@ -3018,7 +3017,6 @@ webhooker/
 │   ├── database/
 │   │   ├── base_model.go      # BaseModel with UUID primary keys
 │   │   ├── database.go        # GORM connection, migrations, admin seed
-│   │   ├── models.go          # AutoMigrate for config-tier models
 │   │   ├── model_setting.go   # Setting entity (key-value app config)
 │   │   ├── model_user.go      # User entity
 │   │   ├── model_webhook.go    # Webhook entity
@@ -3035,6 +3033,11 @@ webhooker/
 │   │   └── databasetest/      # Test support: a WebhookDBManager for tests in other packages
 │   ├── datadir/
 │   │   └── lock.go            # Exclusive advisory lock on DATA_DIR (one instance)
+│   ├── db/migrations/
+│   │   ├── migrations.go      # Applies a database file's SQL files when it is opened
+│   │   ├── main/              # 000_migration.sql and 001_schema.sql for webhooker.db
+│   │   ├── events/            # The same for each events-{uuid}.db
+│   │   └── archive/           # The same for each archive file
 │   ├── globals/
 │   │   └── globals.go         # Build-time variables (appname, version, arch)
 │   ├── gormlog/
@@ -3108,8 +3111,8 @@ webhooker/
 │   ├── static.go              # //go:embed directive
 │   ├── css/input.css          # Tailwind input, source for tailwind.css (make css)
 │   ├── css/tailwind.css       # Generated stylesheet the pages load
-│   ├── css/style.css          # Hand-written, loaded after tailwind.css: btn-small, the pointer cursor for input.css's buttons, the webhook list cards' focus outline
+│   ├── css/style.css          # Hand-written, loaded after tailwind.css: btn-small, the pointer cursor for input.css's buttons, the webhook list cards' focus outline, the rule hiding x-cloak elements
-│   ├── js/app.js              # Copy-to-clipboard, and the Alpine.js components
+│   ├── js/app.js              # Copy-to-clipboard, the delete prompts, and the Alpine.js components
 │   └── js/alpine.min.js       # Alpine.js CSP build, extracted from 3p/ by make assets, not committed
 ├── templates/                 # Go HTML templates (base, login, sources, etc.)
 ├── script/                    # Scripts to Rule Them All entrypoints
@@ -3163,8 +3166,9 @@ Applied to all routes in this order:
 
 1. **RequestID** — Generate unique request IDs (chi built-in)
 2. **SecurityHeaders** — Production security headers on every response (HSTS
-   unless `HSTS_MAX_AGE` is `0`, X-Content-Type-Options, X-Frame-Options, CSP,
+   unless `HSTS_MAX_AGE` is `0`, X-Content-Type-Options, X-Frame-Options, CSP
-   Referrer-Policy, Permissions-Policy)
+   `default-src 'self'; script-src 'self'; style-src 'self'`, Referrer-Policy,
+   Permissions-Policy)
 3. **Logging** — Structured request logging (method, URL, status, latency,
    remote IP, client IP, user agent, request ID)
 4. **Metrics** — Prometheus HTTP metrics (if `METRICS_USERNAME` and
@@ -3174,7 +3178,8 @@ Applied to all routes in this order:
 7. **Recoverer** — Panic recovery: one `ERROR` record through `internal/logger`
    and a `500`
 8. **Sentry** — Error reporting to Sentry (if `SENTRY_DSN` is set; configured
-   with `Repanic: true` so panics still reach Recoverer)
+   with `Repanic: true` so panics still reach Recoverer), and inside it the
+   naming of the request's Sentry transaction after its route pattern
 
 Recoverer sits seventh rather than first, and both neighbours are the reason. It
 runs **inside** everything that observes the response, so the `500` it writes
@@ -3189,9 +3194,9 @@ set a header or start a timer.
 
 Each admin page route group (`/pages`, `/user/*`, `/settings`, `/hooks`,
 `/hook/*`) starts with its own **Recoverer** and, if `SENTRY_DSN` is set, its
-own **Sentry** error reporting. That Recoverer answers a panic with the `500`
+own **Sentry** error reporting and transaction naming. That Recoverer answers a
-error page in the normal layout; the global one keeps the plain-text `500` for
+panic with the `500` error page in the normal layout; the global one keeps the
-every other route.
+plain-text `500` for every other route.
 
 Additionally, form endpoints (`/pages`, `/user/*`, `/settings`, `/hooks`,
 `/hook/*`) apply a **MaxBodySize** middleware that limits POST/PUT/PATCH request
@@ -3248,6 +3253,12 @@ credential check, see [The login endpoint](#the-login-endpoint).
 - Production security headers on all responses: HSTS (unless `HSTS_MAX_AGE` is
   `0`), X-Content-Type-Options (`nosniff`), X-Frame-Options (`DENY`),
   Content-Security-Policy, Referrer-Policy, and Permissions-Policy
+- The Content-Security-Policy is
+  `default-src 'self'; script-src 'self'; style-src 'self'`: a page takes
+  scripts and styles only from files webhooker serves, and the browser runs no
+  inline script or event handler and ignores any `<style>` element or `style`
+  attribute in a page. The pages need none: their scripts are in `static/js/`
+  and their styles in `static/css/`
 - Request body size limits (1 MB) on all form POST endpoints, enforced by
   middleware that runs before CSRF parses the form
 - **CSRF protection** via [gorilla/csrf](https://github.com/gorilla/csrf) on all
@@ -3315,7 +3326,8 @@ stop-hook log:
 3. `server` — the HTTP drain, bounded by `server.ShutdownTimeout` (**3
    seconds**) and by what the hooks before it left, then a Sentry flush if
    `SENTRY_DSN` is set
-4. `delivery.Engine` — waits for its workers, then closes the archive databases
+4. `delivery.Engine` — cuts off any `http` or `slack` request in flight, waits
+   for its workers, then closes the archive databases
 5. `healthcheck`
 6. `WebhookDBManager`
 7. the database close
@@ -3326,6 +3338,12 @@ by the stop context, so a wedge there consumes the 5 seconds before the HTTP
 server hook is ever entered. The hooks after the server are microsecond-scale in
 normal operation.
 
+The delivery engine cancels a delivery request in flight. That attempt is not
+recorded or counted in the metrics, and the delivery keeps its `pending` or
+`retrying` status, so startup recovery sends it again on the next start (see
+Recovery paths under Bounded Worker Pool). If the receiver already had it, it
+receives it twice: the at-least-once duplicate described under Rationale.
+
 The HTTP drain budget is deliberately **shorter** than the sequence budget. Were
 the two equal, a drain that used its whole budget would exhaust the sequence
 budget at the instant it finished, and every later hook — the delivery engine,
@@ -3448,14 +3466,12 @@ fixed independently of the compiler's:
    a step's log, and then fails. `make test` builds this stage alone.
 6. **Builder stage** (`golang:1.26.1-bookworm`) — depends on the lint and test
    phases and the `css-check` and `markdown-check` stages passing (it copies a
-   file from each), runs `make build` (which extracts Alpine.js from `3p/`
+   file from each), and runs `make build` (which extracts Alpine.js from `3p/`
-   first), and then rebuilds the binary with `CGO_ENABLED=1` and static linking
+   first). `make build` compiles with cgo off, so the binary is statically
-   so it runs on musl. Both builds go through `make build`, the relink adding
+   linked and runs on the Alpine runtime image. The version is the `VERSION`
-   its `-extldflags` via `GO_LDFLAGS`, so neither can drop the `-X` that stamps
+   build arg if one is given, otherwise derived from the `.git` in the context,
-   the version. The version is the `VERSION` build arg if one is given,
+   and the stage fails if a context with `.git` would stamp an empty version,
-   otherwise derived from the `.git` in the context, and the stage fails if a
+   `dev` or `unknown` (see [Version stamping](#version-stamping)).
-   context with `.git` would stamp an empty version, `dev` or `unknown` (see
-   [Version stamping](#version-stamping)).
 7. **Runtime stage** (`alpine:3.21`) — copies the static binary and
    `deploy/docker-entrypoint.sh`, creates the `/var/lib/webhooker` directory for
    all SQLite databases, exposes port 8080, and includes a health check against
@@ -3467,11 +3483,6 @@ The lint and test phases invoke `gofmt`, `golangci-lint` and `go test` directly
 rather than `make fmt-check`, `make lint` and `make test`: those targets build
 docker stages, which would need a docker daemon inside this build.
 
-The lint phase and the Go stages use Debian rather than Alpine because
-`gorm.io/driver/sqlite` pulls in `mattn/go-sqlite3`, which needs CGO and does
-not compile against musl. Only the final binary is statically linked, which is
-what lets it run on the Alpine runtime image.
-
 `script/cibuild` is the CI gate: it runs `script/bootstrap`, then
 `script/check`, then builds the image, whose build runs the lint and test phases
 and the stylesheet and Markdown checks again. A build that succeeds is a repo

TODO.md

@@ -12,24 +12,14 @@ One issue per unit of work, one branch and one PR per issue:
 `main` without notice. One `next` -> `main` PR accumulates the milestone;
 releases are cut from `main` separately.
 
-Issue branches do NOT touch this file — the manager maintains it on `next`.
+The tracker (https://git.eeqj.de/sneak/webhooker/issues) is the record of what
-Every branch editing `TODO.md` conflicts with every other (#112).
+is done and what is under way. This file keeps no list of either.
 
 # Status
 
-The milestone (https://git.eeqj.de/sneak/webhooker/milestone/9) is the
+What is still open for 1.0.0 is in its milestone
-authoritative list, and the only place to read a count or a state of play from.
+(https://git.eeqj.de/sneak/webhooker/milestone/9). `v1.0.0` is not tagged, and
-This file records where the project is, not what is in flight: a sentence whose
+only @sneak tags it.
-truth depends on a branch being unmerged is wrong the moment it merges, and this
-file has been wrong that way before.
-
-The durability defect that held the tag has landed
-(https://git.eeqj.de/sneak/webhooker/issues/256, commit `8d64259`). Every SQLite
-handle opens with WAL journaling and a busy timeout, a bookkeeping write that
-fails leaves its delivery in a recoverable state rather than a lying one, and
-recovery skips a delivery that already has a successful result row. Final
-pre-tag verification exercised it and confirmed it holds. Whatever the milestone
-still shows open is what remains before `v1.0.0`.
 
 Delivery is at-least-once by design, not by accident: a send whose result row
 does not land is attempted again, so a receiver can see a duplicate. That is
@@ -38,338 +28,31 @@ under Rationale. It is not a defect to re-file.
 
 # Next Step
 
-Clear the rest of the open 1.0.0 milestone
+@sneak runs webhooker in limited production, then tags `v1.0.0`. Merging `next`
-(https://git.eeqj.de/sneak/webhooker/milestone/9) and tag `v1.0.0`. Merging
+into `main` is a separate act from tagging and waits on neither of those: `next`
-`next` into `main` is a separate act from tagging and waits on neither of those:
+is kept mergeable at all times, which is the point of the branch.
-`next` is kept mergeable at all times, which is the point of the branch.
-
-# Completed Steps
-
-- 2026-08-24 Bind the plaintext HTTP listener deliberately, via `BIND_ADDRESS`
-  defaulting to `127.0.0.1`, and document the reverse-proxy deployment. A
-  hostname, an empty value or a value carrying a port is a startup error, and
-  the `Dockerfile` sets `0.0.0.0` because a loopback bind inside a container is
-  unreachable (https://git.eeqj.de/sneak/webhooker/issues/268). The same commit
-  removed the shutdown race: `httpServer` is built in the constructor rather
-  than assigned from the serving goroutine, which orders the write before every
-  fx hook and rules out the nil dereference a SIGTERM arriving first would have
-  caused, and `sentryEnabled` is an `atomic.Bool`
-  (https://git.eeqj.de/sneak/webhooker/issues/226)
-- 2026-08-24 Remove inbound request signature verification. The entrypoint UUID
-  is the authentication secret, so the per-entrypoint shared secret, the
-  `internal/signature` package, the receiver check, the model fields and the
-  forms are all gone. This reverses the feature that landed earlier in the same
-  milestone (https://git.eeqj.de/sneak/webhooker/issues/67,
-  https://git.eeqj.de/sneak/webhooker/issues/279)
-- 2026-08-24 Stamp the build version into the binary and render it in the UI
-  footer. `script/version` is the single source — `$VERSION`, else
-  `git describe --tags --always --dirty`, else `unknown` — so a `make build`
-  binary and a `make docker` image from one checkout report the same thing, and
-  nothing in it varies between two builds of the same commit, which the release
-  gate's byte-identical assertion would catch
-  (https://git.eeqj.de/sneak/webhooker/issues/253)
-- 2026-08-24 Derive cookie `Secure` and CSRF strictness from the request
-  transport rather than from `WEBHOOKER_ENVIRONMENT`. Behind a real TLS proxy
-  with the environment left at its `dev` default, the session cookie silently
-  lost `Secure` while the CSRF cookie on the same response kept it.
-  `X-Forwarded-Proto` is now matched case-insensitively on its first
-  comma-separated element, so `HTTPS` and `https, http` no longer fall to the
-  relaxed CSRF path (https://git.eeqj.de/sneak/webhooker/issues/269)
-- 2026-08-24 Roll back a failed webhook deletion instead of committing it. A
-  failing delete committed whatever had already succeeded, hard-deleted the
-  per-webhook event database anyway, and redirected as though it had worked —
-  orphaned config plus permanently destroyed history, reported as success. All
-  three delete positions now roll back with the event database intact
-  (https://git.eeqj.de/sneak/webhooker/issues/262)
-- 2026-08-24 Name a deleted target on its historical deliveries, marked
-  `(deleted)`, rather than leaving the event log unable to say where a delivery
-  went. A deleted target's credentials stay masked exactly as a live one's, and
-  it cannot become deliverable again through the receiver, resubmit, replay, the
-  edit form or the toggle (https://git.eeqj.de/sneak/webhooker/issues/211)
-- 2026-08-24 Bound both request-controlled `/metrics` label dimensions, so the
-  unauthenticated receiver is no longer a memory-exhaustion vector: `handler`
-  carries the chi route pattern, and `method` folds anything chi cannot route
-  onto a single `(unmatched)` sentinel. Both were reproduced before the fix —
-  300 random method tokens took the series count from 106 to 7,631, and path
-  flooding reached 62,532 — and a label audit across a live scrape found no
-  third unbounded dimension (https://git.eeqj.de/sneak/webhooker/issues/254,
-  https://git.eeqj.de/sneak/webhooker/issues/261)
-- 2026-08-24 Validate `max_retries` on both target forms. `abc`, `2.7` and `-5`
-  silently became 0 — fire-and-forget — including on the edit path, where it
-  destroyed a working value, and `999999999` stored verbatim. The ceiling of 20
-  is the `max` both templates already declared
-  (https://git.eeqj.de/sneak/webhooker/issues/221)
-- 2026-08-24 Resubmit a stored event as a new undelivered event, so a backend
-  under development can be tested against real captured traffic. Per-delivery
-  replay cannot serve that: it re-sends one finished delivery to its own
-  original target, and a target created for a dev backend has no prior delivery
-  to replay. Resubmit re-injects the stored event at the top of the receiver
-  path and fans it out to whatever targets are active now
-  (https://git.eeqj.de/sneak/webhooker/issues/250)
-- 2026-08-20 Take an exclusive lock on `DATA_DIR` at startup, so two instances
-  on one directory cannot both deliver
-  (https://git.eeqj.de/sneak/webhooker/issues/201)
-- 2026-08-20 Shut down the app when the HTTP listener fails. The `OnStart` hook
-  returned as soon as the serving goroutine was spawned, so a failed listen left
-  fx reporting RUNNING and a live process with nothing bound — invisible to
-  systemd and Docker restart policies
-  (https://git.eeqj.de/sneak/webhooker/issues/200)
-- 2026-08-20 Stop target credentials leaking into the per-webhook event
-  databases (https://git.eeqj.de/sneak/webhooker/issues/206), log SQL with
-  placeholders rather than bound values
-  (https://git.eeqj.de/sneak/webhooker/issues/207), and fail loudly on half-set
-  metrics auth credentials (https://git.eeqj.de/sneak/webhooker/issues/205)
-- 2026-08-20 Read queue depths with `Find`, not `Scan`. `Scan` swaps GORM's own
-  trace recorder in for the logging adapter, and that recorder does not
-  implement `gorm.ParamsFilter`, so those statements logged their bound values
-  interpolated and bypassed the suppression above. The two units gated green
-  against a `next` that lacked the other, and `next` went red when both landed
-  (https://git.eeqj.de/sneak/webhooker/issues/234)
-- 2026-08-20 Render per-attempt delivery detail in the event log
-  (https://git.eeqj.de/sneak/webhooker/issues/202) and add replay of a
-  terminally failed delivery (https://git.eeqj.de/sneak/webhooker/issues/203)
-- 2026-08-20 Expose delivery metrics on `/metrics`
-  (https://git.eeqj.de/sneak/webhooker/issues/209) and document the backup,
-  restore and upgrade procedures
-  (https://git.eeqj.de/sneak/webhooker/issues/210)
-- 2026-08-20 Add a `webhooker resetpw` subcommand and a bootstrap banner. The
-  admin bootstrap password was printed once among roughly 45 fx lines, and under
-  `docker run -d` went to container logs subject to rotation; there was no reset
-  path at all, so recovery meant hand-deleting the users row, documented
-  nowhere. The password is read from stdin or generated, never from argv where
-  `/proc` would publish it (https://git.eeqj.de/sneak/webhooker/issues/208)
-- 2026-08-20 Add `ALLOWED_EGRESS_CIDRS`, an allowlist-only escape hatch for the
-  SSRF guard, so a self-hosted proxy can forward into the operator's own
-  network. The guard's always-blocked set cannot be reopened by configuration
-  (https://git.eeqj.de/sneak/webhooker/issues/204)
-- 2026-08-20 Harden operator-set target headers, which were carried unsafely
-  across a redirect (https://git.eeqj.de/sneak/webhooker/issues/233)
-- 2026-08-20 Add a target edit form with headers and timeout fields
-  (https://git.eeqj.de/sneak/webhooker/issues/127)
-- 2026-08-18 Raise `script/test`'s per-package timeout from 30s to 90s, matching
-  the org-wide backstop. `go test` applies `-timeout` per package, and
-  `internal/handlers` had grown past the old budget: a cache-defeated build
-  failed outright at `GOMAXPROCS=4`, and every run under deliberate host load
-  breached 30s. The measurement table lives in the script (#194)
-- 2026-08-18 Re-sync `REPO_POLICIES.md` from `prompts`. The local copy was stale
-  and still mandated a 20s test target with a 30s timeout, which the org
-  replaced with a 60s cap and a 90s backstop. A synced copy is not a source;
-  reading it as one nearly produced a PR against `prompts` proposing a change
-  already merged there (#196)
-- 2026-08-18 Report handler panics through the logger and answer 500. chi
-  v1.5.5's `Recoverer` scans for a `panic(0x` frame the runtime no longer emits,
-  then indexes `pkg[-1:]`, so it panicked inside its own stack printer before
-  writing a byte: the recovery never ran, the client got a dropped connection
-  instead of a 500, and the original panic was lost. A local middleware replaces
-  it, bounded by `MaxPanicLogLineBytes` (#187)
-- 2026-08-18 Route GORM's logger through `slog` and bound it. Every `gorm.Open`
-  left `logger.Default` in place at `Warn` with `IgnoreRecordNotFoundError`
-  false, so **every record-not-found printed the fully interpolated SQL to
-  stdout** — including the client-chosen path on `/webhook/{uuid}` and the
-  submitted username on the login form, at no level the operator set and outside
-  `internal/logger` entirely. Three call sites, not the two the issue named
-  (#178)
-- 2026-08-18 Bound every `slog` line against client-chosen text. Eight sites
-  reachable unauthenticated, found by reading every `slog` call in the tree
-  rather than only the one reported; the budget moved to a shared
-  `internal/logfield` so no second truncation exists. `DEBUG` being off by
-  default is not a bound and is not treated as one (#176)
-- 2026-08-18 Stop a slow host turning a login-guard test into a segfault. A
-  non-fatal `assert` on an acquire result was dereferenced on the next line, so
-  one timing miss killed the whole `internal/middleware` binary and reddened CI
-  for unrelated PRs. The fix also removed a real production race — `acquire`
-  could shed a request with a slot standing free, because Go picks uniformly
-  among ready `select` cases (#186)
-- 2026-08-18 Send the chi route pattern to Sentry rather than the concrete path.
-  The receiver's path carries the entrypoint capability token, so every Sentry
-  event from `/webhook/{uuid}` shipped a live credential to a third party.
-  Request `Data`, `QueryString`, `Cookies` and `Env` are dropped and headers
-  reduced to an allowlist (#179)
-- 2026-08-18 Read form fields from the POST body only. `r.FormValue` merges the
-  query string, so a login could be driven by URL parameters — putting the
-  password somewhere that lands in access logs, proxy logs and browser history
-  (#160)
-- 2026-08-18 Verify login credentials before spending rate-limit budget, so a
-  flood of wrong passwords cannot lock out the account it is guessing at. The
-  manager took this decision rather than stall the queue; it is flagged on the
-  issue for reversal (#150)
-- 2026-08-18 Run all linting in Docker via `Dockerfile.lint`. Host lint was
-  wrong in both directions from version skew and shared caches. `script/lint`
-  asserts the summary line, because `--no-cache-filter` silently ignores a stage
-  name it does not match — the flag that makes the gate meaningful fails open
-  (#109)
-- 2026-08-18 Serve an event's full stored body over HTTP. The list query
-  truncates for rendering, and that truncated value was the only way to read a
-  body, so the full payload was unreachable (#157)
-- 2026-08-18 Bound the access log line against client-chosen text.
-  `internal/logfield` budgets by _encoded_ bytes, not runes, so a handler's JSON
-  escaping cannot multiply a field past its allowance (#146)
-- 2026-08-18 Mark superseded CI commits `failure` rather than `skipped`. A
-  skipped run rolls up green, so a commit that was never tested reported success
-  (#152)
-- 2026-08-18 Set `fx.StopTimeout` inside the container stop grace, so shutdown
-  hooks are bounded by a deadline the orchestrator will actually honour rather
-  than being killed mid-flush (#134)
-- 2026-08-17 Bucket IPv6 rate-limit keys by `/64`. A single allocation hands out
-  2^64 addresses, so per-address keying let one client mint unlimited buckets.
-  Manager decision, recorded on the issue (#125)
-- 2026-08-17 Correct release-blocking README and startup-warning inaccuracies,
-  including claims about behaviour the code does not have (#151)
-- 2026-08-17 Fetch and verify Alpine.js at build time against
-  `static/vendor.sha256` instead of committing the minified blob, so the
-  dependency is pinned by hash rather than by trust (#145)
-- 2026-08-17 Bound the event log's rendered bodies in the query itself, so a
-  large stored payload cannot be read into memory just to be truncated for
-  display (#135)
-- 2026-08-17 Mask the `http` target's destination URL in the UI: it can carry a
-  bearer credential in its path or query, and was rendered verbatim. Manager
-  decision to mask unconditionally (#115)
-- 2026-08-14 Bound shutdown hooks by their stop context, so a hook that hangs
-  cannot hold the process past its grace period (#102)
-- 2026-08-14 Render templates via a buffer rather than the `ResponseWriter`, so
-  a template error part-way through cannot commit a 200 and then fail — the
-  response is written only once it is whole (#123)
-- 2026-08-14 Align the session codec's max-age with the 7-day absolute cap. The
-  codec accepted cookies the session layer considered expired, so the cap was
-  enforced in one place and not the other (#108)
-- 2026-08-12 Warn when `TRUSTED_PROXIES` is empty in production, where the safe
-  default silently discards forwarded headers and every client rate-limits as
-  the proxy's address (#149)
-- 2026-08-12 Bound the receiver rate limit per client IP across the whole
-  `/webhook/*` route. The existing limiter keyed on the request path and
-  `/webhook/{uuid}` matches any single segment, so a client that invented a
-  fresh path per request minted a fresh bucket per request: the limit on the
-  only unauthenticated endpoint bounded nothing in aggregate, and every request
-  still cost an entrypoint lookup before it 404ed. An outer limiter keyed on the
-  client address alone now bounds that, chained in front of the unchanged
-  per-entrypoint limiter (#139)
-- 2026-08-12 Correct release-blocking documentation inaccuracies: the README
-  promised manual redelivery in the present tense in three places when nothing
-  implements it (the same false claim also sat in the doc comment that was its
-  source text), the env table omitted `RETENTION_SWEEP_INTERVAL`, and `TODO.md`
-  itself omitted five landed units (#141)
-- 2026-08-12 Make the CI gate execute the checks it reports on. The workflow now
-  writes a build-context fingerprint before calling `script/cibuild`, so a code
-  commit invalidates the `COPY` layer of the lint and builder stages while a
-  docs-only commit still replays from cache; a superseding run also rewrites the
-  `failure` status Gitea leaves on commits it cancelled and never tested.
-  Verified by pushing a deliberately broken test and watching CI go red (#119)
-- 2026-08-12 Require a positive `RETENTION_SWEEP_INTERVAL`: a non-positive value
-  reached `time.NewTicker` in both the retention reaper and the archive sweeper,
-  panicking two goroutines with no recover after startup had already reported
-  success (#140)
-- 2026-08-12 Bound the `X-Forwarded-For` scan's allocation to the hop cap: the
-  reverse walk cuts entries with `strings.LastIndexByte` instead of joining and
-  splitting, so a 1 MB header allocates 16 bytes rather than 1.6 MB per request
-  on the unauthenticated receiver. Semantics proven unchanged by differential
-  testing against the previous implementation (#133)
-- 2026-08-12 Cap the `X-Forwarded-For` hop walk at 64 entries, so an
-  attacker-supplied chain cannot burn unbounded CPU in the rate-limit key
-  function; running off the end falls back to the peer address (#124)
-- 2026-08-12 Gate forwarded-header trust behind a `TRUSTED_PROXIES` CIDR list:
-  all three rate limiters key on the connection's own address unless the direct
-  peer is a configured proxy, in which case `X-Forwarded-For` is walked right to
-  left for the first non-proxy hop. Default trusts nothing, and a
-  set-but-unparseable value aborts startup. Before this, any client could mint a
-  fresh bucket or drain another's by rotating a spoofed header (#88)
-- 2026-08-11 Web UI cleanup: nav terminology unified on Webhooks, the Profile
-  settings placeholder removed, a progressive-enhancement copy button for the
-  entrypoint URL, and retention form copy that states the actual policy
-  (deletion by the reaper, 0 retains forever) (#57)
-- 2026-08-11 Mask the webhook credential in delivery errors and logs: Go embeds
-  the request URL in `*url.Error`, so every transport failure persisted the full
-  Slack webhook URL into the per-webhook event database via
-  `DeliveryResult.Error`, a field a future REST API would have served.
-  `maskURLError` drops path, query and userinfo while preserving the wrapped
-  cause, so `errors.Is`/`As` and `Timeout()` still work and DNS, TLS and timeout
-  failures still read differently (#118)
-- 2026-08-11 Rate-limit the public webhook receiver endpoint
-  (`RECEIVER_RATE_LIMIT`, default 120/min), keyed on client IP plus entrypoint
-  path so one entrypoint cannot exhaust another's budget; over-limit requests
-  get 429 with `Retry-After`. It was the one unauthenticated, internet-facing
-  endpoint with no limit at all (#64)
-- 2026-08-11 Enforce the body size limit before CSRF parses the form:
-  `MaxBodySize` is now first in all four form-parsing route groups, so an
-  oversized request is rejected with 413 instead of being read in full by the
-  CSRF middleware before any cap applied (#90)
-- 2026-08-11 Mask target config on the source detail page, which rendered the
-  stored blob verbatim and so exposed the Slack incoming-webhook URL — a bearer
-  credential that cannot be revoked per-holder. Config reaches the template only
-  as a `TargetView` of labelled fields, and header values are rendered as a
-  count (#113)
-- 2026-08-11 Allow `retention_days` of 0 to mean retain forever, via a sentinel
-  written in `BeforeSave` so the GORM column default cannot win the race. Also
-  bounds the reaper's cutoff arithmetic: day counts above 106751 overflowed
-  `time.Duration` and wrapped the cutoff into the future, where every row
-  matched and the sweep deleted everything (#79)
-- 2026-08-09 Inactivity-based session timeout: sliding idle expiry
-  (`SESSION_IDLE_TIMEOUT`, default `24h`) refreshed on authenticated requests,
-  with the 7-day absolute cap kept as an independent backstop that activity
-  never extends (#66)
-- 2026-08-09 Restart recovery and the 60s retry sweep terminally fail an
-  orphaned `retrying` delivery whose target type no longer supports retries,
-  recording a `DeliveryResult` with the reason instead of leaving the delivery
-  stuck forever (#82)
-- 2026-08-09 Root the delivery engine's worker pool and the retention reaper's
-  sweep loop at `context.Background()` rather than the fx `OnStart` hook context
-  (#97), which carries fx's 15s start timeout and killed both roughly fifteen
-  seconds after boot: the proxy silently stopped delivering webhooks entirely,
-  and the reaper never ran a single sweep under its default one-hour interval
-- 2026-08-09 Archive writer lifecycle (#89): deleting a webhook (or its last
-  `database` target) evicts the cached archive writer and closes its handle
-  while deliberately leaving `archive-{webhookID}.db` on disk, and a new
-  `ArchiveSweeper` prunes idle archives on the existing
-  `RETENTION_SWEEP_INTERVAL` without ever creating an archive file
-- 2026-08-09 Configuration parsing fails loudly on set-but-unparseable
-  environment values: `envInt` removed in favour of `envPositiveInt` plus a
-  `PORT` range check, `envBool` now parses with `strconv.ParseBool`, and
-  defaults apply only to unset variables (#80)
-- 2026-08-07 Automatic event retention cleanup based on `retention_days`,
-  deleting expired events, deliveries, and delivery results from each
-  per-webhook event database (#63)
-- 2026-08-07 Update golangci-lint to v2.12.2 (Docker image digest in
-  `Dockerfile`, release-archive sha256 pins in `script/bootstrap`), adopt the
-  canonical `.golangci.yml` (v2 `linters.settings` layout so
-  `lll`/`funlen`/`cyclop`/`dupl` thresholds actually apply), and fix all newly
-  surfaced lint findings
-- 2026-07-07 Adopted scripts-to-rule-them-all: `script/` entrypoints, Makefile
-  shims, README Entrypoints section
-- 2026-03-25 pin golangci-lint Docker image for linting (#55)
-- 2026-03-18 CSRF middleware detects TLS per-request, fixing login over plain
-  HTTP and behind reverse proxies (#54)
-- 2026-03-17 root path redirects based on auth state (#52)
-- 2026-03-17 CSRF protection, SSRF prevention for HTTP delivery targets with DNS
-  rebinding defense, and per-IP login rate limiting (#42)
-- 2026-03-17 Slack target type for incoming webhook notifications (#47)
-- 2026-03-17 Dockerfile absolute paths and static linking (#49); absolute dev
-  DATA_DIR default and clarified env docs (#46)
-- 2026-03-05 security headers middleware, session regeneration on login, request
-  body size limits (#41)
-- 2026-03-04 tests for delivery, middleware, and session packages (#32); removed
-  the build-architecture global (#31)
-- 2026-03-04 1.0 MVP merge: Webhook/Entrypoint/Target rename, core delivery
-  engine with bounded worker pool and circuit breaker, parallel fan-out,
-  per-webhook event databases, management UI (#16)
-- 2026-03-01 repo brought to REPO_POLICIES standards; TODO.md folded into README
-  (#6)
 
 # Future Steps
 
-- Delivery status and retry management UI. Replay of a terminally failed
+- Per-webhook rate limits, set in the web UI and enforced in the receiver
-  delivery and per-attempt detail already landed
+  handler, on top of the receiver's existing limits, which `RECEIVER_RATE_LIMIT`
-  (https://git.eeqj.de/sneak/webhooker/issues/203,
+  sets for the whole server (see Rate Limiting in the README)
-  https://git.eeqj.de/sneak/webhooker/issues/202)
+- API key authentication for programmatic access (the `APIKey` model exists, but
-- Per-webhook rate limiting in the receiver handler (per-webhook config plus
+  nothing creates a key and no middleware checks one)
-  handler enforcement; global limits must not apply to receiver endpoints)
-- API key authentication for programmatic access (APIKey model exists; Bearer
-  token middleware does not)
 - REST API v1
     - CRUD for webhooks, entrypoints, targets
     - event viewing and filtering endpoints
     - event redelivery endpoint
     - OpenAPI specification
-- Analytics dashboard: success rates, response times, volume
+- Analytics in the web UI: response times summarised per webhook or per target,
-- A remember-me option at login
+  and figures for all webhooks together. The webhook page already shows event,
+  delivery and failure counts and failure percentages, the target list counts
+  each target's deliveries, the event log and each event's page show every
+  attempt's duration, and `/metrics` carries server-wide counts and delivery
+  durations
+- A remember-me option at login. Every session already survives a browser
+  restart: it lasts up to 7 days, and ends sooner after `SESSION_IDLE_TIMEOUT`
+  (default 24 hours) without activity
 - Password reset flow for a forgotten password over the web. The authenticated
   password _change_ flow already landed, and a lost password is recoverable from
   the console with `webhooker resetpw`
@@ -377,7 +60,13 @@ Clear the rest of the open 1.0.0 milestone
 - Later, nice to have
     - email delivery target type
     - SNS and S3 delivery targets
-    - data transformations (e.g. webhook to Slack message formatting)
+    - data transformations (e.g. a configurable Slack message format; the Slack
-    - JSONL file delivery with periodic S3 upload
+      target sends one fixed format today)
-    - webhook event search and filtering
+    - JSONL file delivery with periodic S3 upload (the `database` target already
-    - multi-user with role-based access control
+      archives events to SQLite files, rotated hourly, daily or monthly if set,
+      and the webhook page offers them as one gzipped JSON download)
+    - webhook event search, and filtering beyond the event log's failed and
+      pending views
+    - multi-user with role-based access control (each webhook already belongs to
+      one account, but there is only the account made at first boot and no way
+      to add another)

internal/database/database.go

@@ -18,6 +18,7 @@ import (
 	"sneak.berlin/go/webhooker/internal/banner"
 	"sneak.berlin/go/webhooker/internal/config"
 	"sneak.berlin/go/webhooker/internal/datadir"
+	"sneak.berlin/go/webhooker/internal/db/migrations"
 	"sneak.berlin/go/webhooker/internal/gormlog"
 	"sneak.berlin/go/webhooker/internal/logger"
 )
@@ -249,12 +250,21 @@ func (d *Database) connectTo(dataDir string) error {
 		return fmt.Errorf("migrating %s: %w", dbPath, err)
 	}
 
+	err = parseModels(
+		db, &Setting{}, &User{}, &APIKey{},
+		&Webhook{}, &Entrypoint{}, &Target{},
+	)
+	if err != nil {
+		return fmt.Errorf("connecting to %s: %w", dbPath, err)
+	}
+
 	return nil
 }
 
+// migrate applies the main database's SQL files. See
+// internal/db/migrations.
 func (d *Database) migrate() error {
-	// Run GORM auto-migrations
+	err := migrations.Apply(d.db, migrations.Main)
-	err := d.Migrate()
 	if err != nil {
 		d.log.Error(
 			"failed to run database migrations",
@@ -269,6 +279,21 @@ func (d *Database) migrate() error {
 	return nil
 }
 
+// parseModels has GORM parse each model for db before the handle is
+// shared. GORM parses a model the first time a handle uses it, and two
+// goroutines doing that at once for models that refer to each other,
+// as Webhook and Target do, race inside GORM.
+func parseModels(db *gorm.DB, models ...any) error {
+	for _, model := range models {
+		err := (&gorm.Statement{DB: db}).Parse(model)
+		if err != nil {
+			return fmt.Errorf("parsing model %T: %w", model, err)
+		}
+	}
+
+	return nil
+}
+
 // ensureAdminUser creates the bootstrap admin account when the
 // deployment has no users at all.
 func (d *Database) ensureAdminUser() error {

internal/database/event_db_isolation.go

@@ -2,7 +2,6 @@ package database
 
 import (
 	"fmt"
-	"log/slog"
 
 	"gorm.io/gorm"
 	"gorm.io/gorm/clause"
@@ -18,14 +17,14 @@ const omitAssociationsCallback = "webhooker:omit_associations"
 //
 // A per-webhook database holds the event tier only, but Delivery
 // declares belongs-to Event and Target and the delivery engine fills
-// both in memory before writing. Without this guard GORM upserts
+// both in memory before writing. Without this guard GORM upserts those
-// those parent rows here on the delivery and retry write paths,
+// parent rows here on the delivery and retry write paths, and the
-// copying targets.config, which holds destination URLs and bearer
+// target, whose config holds destination URLs and bearer credentials,
-// credentials, into the file most likely to be backed up or handed
+// goes to a targets table this file does not have, which fails the
-// to someone else. Registering the guard on the connection covers
+// write. Registering the guard on the connection
-// every write path, including writes inside a transaction and write
+// covers every write path, including writes inside a transaction and
-// paths added later. Every event-tier row this file holds is written
+// write paths added later. Every event-tier row this file holds is
-// explicitly, so nothing depends on the automatic save.
+// written explicitly, so nothing depends on the automatic save.
 func omitAssociations(db *gorm.DB) error {
 	omit := func(tx *gorm.DB) {
 		tx.Statement.Omits = append(
@@ -53,107 +52,3 @@ func omitAssociations(db *gorm.DB) error {
 
 	return nil
 }
-
-// eventDBSweptVersion is the PRAGMA user_version purgeTargetRows
-// stamps into a per-webhook database once it has removed any leaked
-// target rows *and* the VACUUM that removes their bytes has returned.
-// Nothing else in the tree uses user_version, so 0 means "not swept
-// by this build".
-//
-// The stamp, not the DELETE, is what records that a file is done. A
-// DELETE commits on its own, so a sweep that is interrupted or whose
-// VACUUM fails leaves a file whose rows are gone but whose credential
-// bytes are still in the free pages -- indistinguishable, by row
-// count, from a file that never leaked. Both leave the stamp unset,
-// so the next open sweeps again.
-const eventDBSweptVersion = 1
-
-// purgeTargetRows deletes target rows that an earlier build's
-// association upsert wrote into a per-webhook database, and rewrites
-// the file so their bytes are gone with them. AutoMigrate creates a
-// targets table in every one of these files because Delivery declares
-// a belongs-to Target, but nothing in the event tier may put rows in
-// it. The rows it did put there are junk, not history: they carry an
-// empty webhook_id, and delivery rows resolve their target against
-// the main database, so nothing here refers to them.
-//
-// The DELETE only unlinks the rows: modernc.org/sqlite leaves
-// secure_delete at SQLite's default of off, so the credential bytes
-// stay readable in the file's free pages and a backup of a swept file
-// would still hand them over. VACUUM rewrites the file without them.
-//
-// This runs before every migration and is gated on
-// eventDBSweptVersion, so a file pays for the rewrite once, on the
-// first open that finds it unstamped, and every open after that is a
-// PRAGMA read. A file this build created is stamped before its
-// targets table exists, so it never vacuums at all. A failure here
-// fails the open with the stamp left unset, so the sweep is retried
-// rather than skipped -- a webhook whose file cannot be swept stays
-// unusable instead of quietly serving from a file that still holds
-// recoverable credentials.
-func purgeTargetRows(
-	db *gorm.DB, log *slog.Logger, webhookID string,
-) error {
-	var version int
-
-	// Row().Scan, not (*gorm.DB).Scan: see internal/gormlog.
-	err := db.Raw("PRAGMA user_version").Row().Scan(&version)
-	if err != nil {
-		return fmt.Errorf(
-			"reading sweep marker of webhook database %s: %w",
-			webhookID, err,
-		)
-	}
-
-	if version >= eventDBSweptVersion {
-		return nil
-	}
-
-	var purged int64
-
-	if db.Migrator().HasTable("targets") {
-		res := db.Exec("DELETE FROM targets")
-		if res.Error != nil {
-			return fmt.Errorf(
-				"purging target rows from webhook database %s: %w",
-				webhookID, res.Error,
-			)
-		}
-
-		purged = res.RowsAffected
-
-		// Unconditional: a zero row count here does not mean there is
-		// nothing to remove, only that no *live* row is left. See
-		// eventDBSweptVersion.
-		err = db.Exec("VACUUM").Error
-		if err != nil {
-			return fmt.Errorf(
-				"purged %d leaked target rows from webhook database "+
-					"%s but vacuuming it failed, so the deleted "+
-					"target credentials are still recoverable from "+
-					"the file; it stays marked unswept and the next "+
-					"open retries: %w",
-				purged, webhookID, err,
-			)
-		}
-	}
-
-	err = db.Exec(fmt.Sprintf(
-		"PRAGMA user_version = %d", eventDBSweptVersion,
-	)).Error
-	if err != nil {
-		return fmt.Errorf(
-			"marking webhook database %s swept: %w", webhookID, err,
-		)
-	}
-
-	if purged > 0 {
-		log.Warn(
-			"purged leaked target rows from per-webhook database",
-			"webhook_id", webhookID,
-			"rows", purged,
-		)
-	}
-
-	return nil
-}

internal/database/event_db_isolation_test.go

@@ -1,392 +1,15 @@
 package database_test
 
 import (
-	"bytes"
-	"database/sql"
-	"fmt"
-	"os"
-	"path/filepath"
 	"testing"
 
 	"github.com/google/uuid"
 	"github.com/stretchr/testify/assert"
 	"github.com/stretchr/testify/require"
-	_ "modernc.org/sqlite"
 	"sneak.berlin/go/webhooker/internal/database"
 	"sneak.berlin/go/webhooker/internal/database/databasetest"
 )
 
-// testDataDirPerm is the mode the test data directory is created
-// with.
-const testDataDirPerm = 0o750
-
-// eventDBDataDir returns a data directory that a WebhookDBManager
-// can be pointed at.
-func eventDBDataDir(t *testing.T) string {
-	t.Helper()
-
-	dir := filepath.Join(t.TempDir(), "events")
-	require.NoError(t, os.MkdirAll(dir, testDataDirPerm))
-
-	return dir
-}
-
-// openRawEventDB opens the per-webhook database file directly,
-// without the manager, so a test can put a file on disk in a state
-// the manager has to cope with, or inspect one afterwards.
-func openRawEventDB(
-	t *testing.T, dataDir, webhookID string,
-) *sql.DB {
-	t.Helper()
-
-	path := filepath.Join(
-		dataDir, fmt.Sprintf("events-%s.db", webhookID),
-	)
-
-	sqlDB, err := sql.Open(
-		"sqlite",
-		fmt.Sprintf("file:%s?mode=rwc", path),
-	)
-	require.NoError(t, err)
-
-	t.Cleanup(func() { _ = sqlDB.Close() })
-
-	return sqlDB
-}
-
-// eventDBFileBytes reads a per-webhook database file off disk, so a
-// test can assert on what the file itself still holds rather than on
-// what a query returns.
-func eventDBFileBytes(t *testing.T, dataDir, webhookID string) []byte {
-	t.Helper()
-
-	//nolint:gosec // reads a file the test just created under t.TempDir()
-	raw, err := os.ReadFile(filepath.Join(
-		dataDir, fmt.Sprintf("events-%s.db", webhookID),
-	))
-	require.NoError(t, err)
-
-	return raw
-}
-
-// eventDBUserVersion returns the PRAGMA user_version of a per-webhook
-// database file, which is the marker purgeTargetRows stamps once it
-// has swept and vacuumed.
-func eventDBUserVersion(t *testing.T, sqlDB *sql.DB) int {
-	t.Helper()
-
-	var version int
-
-	require.NoError(t, sqlDB.QueryRowContext(
-		t.Context(), "PRAGMA user_version",
-	).Scan(&version))
-
-	return version
-}
-
-// clearEventDBSweptMarker resets the sweep marker to 0, which is what
-// a file written by a build without the sweep looks like. Tests that
-// seed a leaked row have to create the file through the manager to
-// get the real targets table shape, and that stamps it.
-func clearEventDBSweptMarker(t *testing.T, sqlDB *sql.DB) {
-	t.Helper()
-
-	_, err := sqlDB.ExecContext(t.Context(), "PRAGMA user_version = 0")
-	require.NoError(t, err)
-}
-
-// countTargetRows returns the number of rows in the targets table of
-// a per-webhook database file, or -1 if the table does not exist.
-func countTargetRows(t *testing.T, sqlDB *sql.DB) int {
-	t.Helper()
-
-	var tables int
-
-	require.NoError(t, sqlDB.QueryRowContext(
-		t.Context(),
-		"SELECT count(*) FROM sqlite_master "+
-			"WHERE type = 'table' AND name = 'targets'",
-	).Scan(&tables))
-
-	if tables == 0 {
-		return -1
-	}
-
-	var rows int
-
-	require.NoError(t, sqlDB.QueryRowContext(
-		t.Context(), "SELECT count(*) FROM targets",
-	).Scan(&rows))
-
-	return rows
-}
-
-// TestOpenPurgesLeakedTargetRows covers the sweep for event
-// databases written by a build that let GORM upsert target rows
-// into them: opening the database clears them, and opening it again
-// is a no-op.
-func TestOpenPurgesLeakedTargetRows(t *testing.T) {
-	t.Parallel()
-
-	dataDir := eventDBDataDir(t)
-	webhookID := uuid.New().String()
-
-	// Create the file the way the application does, so the targets
-	// table has exactly the shape AutoMigrate gives it, then write
-	// a leaked row into it the way the association upsert did.
-	initial := databasetest.NewWebhookDBManager(t, dataDir)
-
-	_, err := initial.GetDB(webhookID)
-	require.NoError(t, err)
-	require.NoError(t, initial.CloseAll())
-
-	seed := openRawEventDB(t, dataDir, webhookID)
-
-	_, err = seed.ExecContext(
-		t.Context(),
-		"INSERT INTO targets "+
-			"(id, webhook_id, name, type, config) "+
-			"VALUES (?, '', ?, ?, ?)",
-		uuid.New().String(),
-		"leaked-target",
-		"slack",
-		`{"webhookUrl":"https://hooks.example/T000/B000/secret"}`,
-	)
-	require.NoError(t, err)
-	require.Equal(t, 1, countTargetRows(t, seed))
-	clearEventDBSweptMarker(t, seed)
-	require.NoError(t, seed.Close())
-
-	mgr := databasetest.NewWebhookDBManager(t, dataDir)
-
-	_, err = mgr.GetDB(webhookID)
-	require.NoError(t, err)
-	require.NoError(t, mgr.CloseAll())
-
-	check := openRawEventDB(t, dataDir, webhookID)
-	assert.Zero(t, countTargetRows(t, check))
-	assert.Equal(
-		t, 1, eventDBUserVersion(t, check),
-		"a completed sweep must mark the file so later opens skip it",
-	)
-	require.NoError(t, check.Close())
-
-	// Idempotent: a second open leaves it at zero and does not
-	// error.
-	again := databasetest.NewWebhookDBManager(t, dataDir)
-
-	_, err = again.GetDB(webhookID)
-	require.NoError(t, err)
-	require.NoError(t, again.CloseAll())
-
-	recheck := openRawEventDB(t, dataDir, webhookID)
-	assert.Zero(t, countTargetRows(t, recheck))
-}
-
-// TestOpenPurgeRemovesCredentialBytes covers the sweep at the level
-// that matters for a backup handed to someone else: the leaked
-// credential must be gone from the raw bytes of the file, not merely
-// unreachable by query. A bare DELETE unlinks the row and leaves the
-// bytes readable in the free pages, so this fails without the VACUUM
-// in purgeTargetRows.
-func TestOpenPurgeRemovesCredentialBytes(t *testing.T) {
-	t.Parallel()
-
-	dataDir := eventDBDataDir(t)
-	webhookID := uuid.New().String()
-	credential := "T00000000/B00000000/" + uuid.New().String()
-
-	initial := databasetest.NewWebhookDBManager(t, dataDir)
-
-	_, err := initial.GetDB(webhookID)
-	require.NoError(t, err)
-	require.NoError(t, initial.CloseAll())
-
-	seed := openRawEventDB(t, dataDir, webhookID)
-
-	_, err = seed.ExecContext(
-		t.Context(),
-		"INSERT INTO targets "+
-			"(id, webhook_id, name, type, config) "+
-			"VALUES (?, '', ?, ?, ?)",
-		uuid.New().String(),
-		"leaked-target",
-		"slack",
-		fmt.Sprintf(
-			`{"webhookUrl":"https://hooks.example/%s"}`, credential,
-		),
-	)
-	require.NoError(t, err)
-	clearEventDBSweptMarker(t, seed)
-	require.NoError(t, seed.Close())
-
-	// The seed has to be in the file for its absence later to mean
-	// anything.
-	require.True(
-		t,
-		bytes.Contains(
-			eventDBFileBytes(t, dataDir, webhookID),
-			[]byte(credential),
-		),
-		"seeded credential is not in the file, so this test proves nothing",
-	)
-
-	mgr := databasetest.NewWebhookDBManager(t, dataDir)
-
-	_, err = mgr.GetDB(webhookID)
-	require.NoError(t, err)
-	require.NoError(t, mgr.CloseAll())
-
-	assert.NotContains(
-		t,
-		string(eventDBFileBytes(t, dataDir, webhookID)),
-		credential,
-		"leaked credential is still recoverable from the raw file",
-	)
-}
-
-// TestOpenRevacuumsAfterIncompleteSweep covers the case a row count
-// cannot see: the rows are already deleted but the file was never
-// vacuumed, because an earlier sweep died between the two or its
-// VACUUM failed. The credential bytes are still recoverable, and the
-// unset marker is the only thing that says so, so the next open must
-// vacuum rather than conclude from the empty table that there is
-// nothing to do.
-func TestOpenRevacuumsAfterIncompleteSweep(t *testing.T) {
-	t.Parallel()
-
-	dataDir := eventDBDataDir(t)
-	webhookID := uuid.New().String()
-	credential := "T00000000/B00000000/" + uuid.New().String()
-
-	initial := databasetest.NewWebhookDBManager(t, dataDir)
-
-	_, err := initial.GetDB(webhookID)
-	require.NoError(t, err)
-	require.NoError(t, initial.CloseAll())
-
-	seed := openRawEventDB(t, dataDir, webhookID)
-
-	_, err = seed.ExecContext(
-		t.Context(),
-		"INSERT INTO targets "+
-			"(id, webhook_id, name, type, config) "+
-			"VALUES (?, '', ?, ?, ?)",
-		uuid.New().String(),
-		"leaked-target",
-		"slack",
-		fmt.Sprintf(
-			`{"webhookUrl":"https://hooks.example/%s"}`, credential,
-		),
-	)
-	require.NoError(t, err)
-
-	// Exactly the state an interrupted sweep leaves: rows gone,
-	// marker unset, bytes still in the free pages.
-	_, err = seed.ExecContext(t.Context(), "DELETE FROM targets")
-	require.NoError(t, err)
-	require.Zero(t, countTargetRows(t, seed))
-	clearEventDBSweptMarker(t, seed)
-	require.NoError(t, seed.Close())
-
-	require.True(
-		t,
-		bytes.Contains(
-			eventDBFileBytes(t, dataDir, webhookID),
-			[]byte(credential),
-		),
-		"the deleted row's bytes must still be in the file, or this "+
-			"test proves nothing",
-	)
-
-	mgr := databasetest.NewWebhookDBManager(t, dataDir)
-
-	_, err = mgr.GetDB(webhookID)
-	require.NoError(t, err)
-	require.NoError(t, mgr.CloseAll())
-
-	assert.NotContains(
-		t,
-		string(eventDBFileBytes(t, dataDir, webhookID)),
-		credential,
-		"an interrupted sweep was not retried, so the credential is "+
-			"still recoverable from the raw file",
-	)
-
-	check := openRawEventDB(t, dataDir, webhookID)
-	assert.Equal(t, 1, eventDBUserVersion(t, check))
-}
-
-// TestOpenSkipsSweptDatabase covers the other half of the marker: a
-// file this build created is marked without ever being vacuumed, and
-// a marked file is not swept again.
-func TestOpenSkipsSweptDatabase(t *testing.T) {
-	t.Parallel()
-
-	dataDir := eventDBDataDir(t)
-	webhookID := uuid.New().String()
-
-	mgr := databasetest.NewWebhookDBManager(t, dataDir)
-
-	_, err := mgr.GetDB(webhookID)
-	require.NoError(t, err)
-	require.NoError(t, mgr.CloseAll())
-
-	marked := openRawEventDB(t, dataDir, webhookID)
-	assert.Equal(t, 1, eventDBUserVersion(t, marked))
-
-	// A marked file is left alone, so a row written into it survives
-	// a reopen. Nothing writes target rows any more; this stands in
-	// for the sweep having run.
-	_, err = marked.ExecContext(
-		t.Context(),
-		"INSERT INTO targets "+
-			"(id, webhook_id, name, type, config) "+
-			"VALUES (?, '', ?, ?, ?)",
-		uuid.New().String(), "sentinel", "slack", `{}`,
-	)
-	require.NoError(t, err)
-	require.NoError(t, marked.Close())
-
-	again := databasetest.NewWebhookDBManager(t, dataDir)
-
-	_, err = again.GetDB(webhookID)
-	require.NoError(t, err)
-	require.NoError(t, again.CloseAll())
-
-	check := openRawEventDB(t, dataDir, webhookID)
-	assert.Equal(
-		t, 1, countTargetRows(t, check),
-		"a marked file must not be swept again",
-	)
-}
-
-// TestOpenSucceedsWithoutTargetsTable covers an existing event
-// database that never grew a targets table. The sweep must not fail
-// startup on it.
-func TestOpenSucceedsWithoutTargetsTable(t *testing.T) {
-	t.Parallel()
-
-	dataDir := eventDBDataDir(t)
-	webhookID := uuid.New().String()
-
-	seed := openRawEventDB(t, dataDir, webhookID)
-
-	_, err := seed.ExecContext(
-		t.Context(),
-		"CREATE TABLE events (id text PRIMARY KEY)",
-	)
-	require.NoError(t, err)
-	require.NoError(t, seed.Close())
-
-	mgr := databasetest.NewWebhookDBManager(t, dataDir)
-
-	db, err := mgr.GetDB(webhookID)
-	require.NoError(t, err)
-	assert.NotNil(t, db)
-	require.NoError(t, mgr.CloseAll())
-}
-
 // TestEventDBCreateOmitsAssociations covers the connection-level
 // guard directly: a Delivery carrying its Event and Target in
 // memory, written through the manager's handle, must store only the
@@ -394,10 +17,9 @@ func TestOpenSucceedsWithoutTargetsTable(t *testing.T) {
 func TestEventDBCreateOmitsAssociations(t *testing.T) {
 	t.Parallel()
 
-	dataDir := eventDBDataDir(t)
 	webhookID := uuid.New().String()
 
-	mgr := databasetest.NewWebhookDBManager(t, dataDir)
+	mgr := databasetest.NewWebhookDBManager(t, t.TempDir())
 
 	db, err := mgr.GetDB(webhookID)
 	require.NoError(t, err)
@@ -432,8 +54,9 @@ func TestEventDBCreateOmitsAssociations(t *testing.T) {
 	require.NoError(t, db.Model(d).
 		Update("status", database.DeliveryStatusDelivered).
 		Error)
-	require.NoError(t, mgr.CloseAll())
 
-	check := openRawEventDB(t, dataDir, webhookID)
+	// The file has no targets table, so the create and update above
-	assert.Zero(t, countTargetRows(t, check))
+	// would have failed had GORM tried to save the target.
+	assert.False(t, db.Migrator().HasTable(&database.Target{}))
+	require.NoError(t, mgr.CloseAll())
 }

internal/database/event_tier_indexes_test.go

@@ -13,10 +13,8 @@ import (
 	"sneak.berlin/go/webhooker/internal/database"
 )
 
-// TestWebhookDBManager_OpenAddsEventTierIndexes verifies that opening a
+// TestWebhookDBManager_OpenAddsEventTierIndexes verifies that a new
-// per-webhook database that predates these indexes creates them. It
+// per-webhook database has these indexes.
-// stands in for an older database file by dropping the indexes
-// AutoMigrate just created, then reopening the same file.
 func TestWebhookDBManager_OpenAddsEventTierIndexes(t *testing.T) {
 	t.Parallel()
 
@@ -42,27 +40,9 @@ func TestWebhookDBManager_OpenAddsEventTierIndexes(t *testing.T) {
 	db, err := mgr.GetDB(webhookID)
 	require.NoError(t, err)
 
-	// A fresh database has them.
-	for _, ix := range indexes {
-		require.True(t, db.Migrator().HasIndex(ix.model, ix.name))
-	}
-
-	// Stand in for a database file created before the indexes existed.
-	for _, ix := range indexes {
-		require.NoError(t, db.Migrator().DropIndex(ix.model, ix.name))
-		require.False(t, db.Migrator().HasIndex(ix.model, ix.name))
-	}
-
-	// Drop the cached connection so the next open reopens the file and
-	// runs AutoMigrate against it, as a restart would.
-	require.NoError(t, mgr.CloseAll())
-
-	db, err = mgr.GetDB(webhookID)
-	require.NoError(t, err)
-
 	for _, ix := range indexes {
 		assert.True(t, db.Migrator().HasIndex(ix.model, ix.name),
-			"opening the existing database should create %s", ix.name)
+			"a new database should have %s", ix.name)
 	}
 }
 

internal/database/model_event.go

@@ -45,8 +45,7 @@ type Event struct {
 
 	// ResubmittedFromID names the event this one was copied from by
 	// an operator resubmit. It is nil for an event that arrived on
-	// the receiver, which is every event created before the column
+	// the receiver. It is not a foreign key: the source event can be
-	// existed. It is not a foreign key: the source event can be
 	// reaped by retention while its copies remain, and the id is
 	// kept as the record of where the copy came from either way.
 	ResubmittedFromID *string `gorm:"type:uuid;index:idx_events_resubmitted_from_id,priority:1;index:idx_events_entrypoint_id,priority:3" json:"resubmittedFromId,omitempty"`

internal/database/model_webhook.go

@@ -82,7 +82,7 @@ type Webhook struct {
 // value while building the insert statement, which runs after
 // BeforeSave; rewriting any later than this loses that race and the
 // row lands at 30 days. Living on the model also means a future call
-// site — a REST API, a fixture, a migration — cannot bypass it.
+// site, such as a REST API or a fixture, cannot bypass it.
 func (w *Webhook) BeforeSave(_ *gorm.DB) error {
 	if w.RetentionDays <= 0 {
 		w.RetentionDays = RetentionForeverDays
@@ -98,8 +98,8 @@ func (w *Webhook) BeforeSave(_ *gorm.DB) error {
 // are exempt from reaping.
 //
 // It accepts the RetentionForeverDays sentinel written by BeforeSave
-// and, defensively, the non-positive values that rows written before
+// and, defensively, any non-positive value, which BeforeSave treats
-// the sentinel existed may still carry.
+// as a request to retain forever.
 func retainsForever(retentionDays int) bool {
 	return retentionDays <= 0 ||
 		retentionDays >= RetentionForeverDays

internal/database/models.go

@@ -1,17 +0,0 @@
-package database
-
-// Migrate runs database migrations for the main application database.
-// Only configuration-tier models are stored in the main database.
-// Event-tier models (Event, Delivery, DeliveryResult, EventTotals,
-// TargetTotals, EntrypointTotals) live in
-// per-webhook dedicated databases managed by WebhookDBManager.
-func (d *Database) Migrate() error {
-	return d.db.AutoMigrate(
-		&Setting{},
-		&User{},
-		&APIKey{},
-		&Webhook{},
-		&Entrypoint{},
-		&Target{},
-	)
-}

internal/database/retention.go

@@ -253,7 +253,7 @@ func (r *RetentionReaper) reapWebhook(
 // row in the database: the sweep then deletes every event, delivery,
 // and delivery result, including ones created seconds ago. Rejecting
 // out-of-range input at the form is the primary guard; saturating here
-// means an old row, a migration, or a future call site cannot turn a
+// means a value that reaches the row by any other path cannot turn a
 // too-large retention into total data loss.
 func retentionCutoff(
 	now time.Time,

internal/database/schema_test.go

@@ -0,0 +1,168 @@
+package database_test
+
+import (
+	"log/slog"
+	"net/http"
+	"testing"
+	"time"
+
+	"github.com/google/uuid"
+	"github.com/stretchr/testify/assert"
+	"github.com/stretchr/testify/require"
+	"gorm.io/gorm"
+	"sneak.berlin/go/webhooker/internal/database"
+	"sneak.berlin/go/webhooker/internal/database/databasetest"
+)
+
+// withFixedNow returns db with GORM's clock stopped at a time in UTC,
+// so that the created and updated times it writes come back from
+// SQLite equal to the row written, which holds them in UTC.
+func withFixedNow(db *gorm.DB) (*gorm.DB, time.Time) {
+	at := time.Date(2026, 10, 13, 12, 30, 45, 123456789, time.UTC)
+
+	return db.Session(&gorm.Session{
+		NowFunc: func() time.Time { return at },
+	}), at
+}
+
+// newBaseModel returns a BaseModel with a new ID.
+func newBaseModel() database.BaseModel {
+	return database.BaseModel{ID: uuid.New().String()}
+}
+
+// assertRoundTrip writes row, which must be the only row of its
+// table, and reads back the same row.
+func assertRoundTrip[T any](t *testing.T, db *gorm.DB, row *T) {
+	t.Helper()
+
+	require.NoError(t, db.Create(row).Error)
+
+	var got T
+
+	require.NoError(t, db.First(&got).Error)
+	assert.Equal(t, *row, got)
+}
+
+// TestMainSchemaHoldsEveryModel writes a row of each main database
+// model, every field set, into a database made from its SQL files.
+func TestMainSchemaHoldsEveryModel(t *testing.T) {
+	t.Parallel()
+
+	mainDB, err := database.Open(
+		t.TempDir(), slog.New(slog.DiscardHandler),
+	)
+	require.NoError(t, err)
+
+	t.Cleanup(func() { _ = mainDB.Close() })
+
+	db, at := withFixedNow(mainDB.DB())
+
+	assertRoundTrip(t, db, &database.Setting{
+		Key: "session_key", Value: "value",
+	})
+	assertRoundTrip(t, db, &database.User{
+		BaseModel: newBaseModel(),
+		Username:  "user",
+		Password:  "password hash",
+	})
+	assertRoundTrip(t, db, &database.APIKey{
+		BaseModel:   newBaseModel(),
+		UserID:      uuid.New().String(),
+		Key:         "key",
+		Description: "API key description",
+		LastUsedAt:  &at,
+	})
+	assertRoundTrip(t, db, &database.Webhook{
+		BaseModel:     newBaseModel(),
+		UserID:        uuid.New().String(),
+		Name:          "webhook",
+		Description:   "webhook description",
+		RetentionDays: 7,
+	})
+	assertRoundTrip(t, db, &database.Entrypoint{
+		BaseModel:   newBaseModel(),
+		WebhookID:   uuid.New().String(),
+		Path:        uuid.New().String(),
+		Description: "entrypoint description",
+		Active:      true,
+	})
+	assertRoundTrip(t, db, &database.Target{
+		BaseModel:  newBaseModel(),
+		WebhookID:  uuid.New().String(),
+		Name:       "target",
+		Type:       database.TargetTypeHTTP,
+		Active:     true,
+		Config:     `{"url":"https://example.com/"}`,
+		MaxRetries: 3,
+	})
+}
+
+// TestEventSchemaHoldsEveryModel does the same for a webhook's event
+// database, which the manager makes from its SQL files.
+func TestEventSchemaHoldsEveryModel(t *testing.T) {
+	t.Parallel()
+
+	mgr := databasetest.NewWebhookDBManager(t, t.TempDir())
+
+	t.Cleanup(func() { _ = mgr.CloseAll() })
+
+	events, err := mgr.GetDB(uuid.New().String())
+	require.NoError(t, err)
+
+	db, at := withFixedNow(events)
+	source := uuid.New().String()
+
+	assertRoundTrip(t, db, &database.Event{
+		BaseModel:         newBaseModel(),
+		WebhookID:         uuid.New().String(),
+		EntrypointID:      uuid.New().String(),
+		Method:            http.MethodPost,
+		RawQuery:          "a=1",
+		Headers:           `{"X":"Y"}`,
+		Body:              "body",
+		ContentType:       "text/plain",
+		BodyBytes:         4,
+		ResubmittedFromID: &source,
+	})
+	assertRoundTrip(t, db, &database.Delivery{
+		BaseModel:  newBaseModel(),
+		EventID:    uuid.New().String(),
+		TargetID:   uuid.New().String(),
+		Status:     database.DeliveryStatusDelivered,
+		FinishedAt: &at,
+		Replay:     true,
+	})
+	assertRoundTrip(t, db, &database.DeliveryResult{
+		BaseModel:    newBaseModel(),
+		DeliveryID:   uuid.New().String(),
+		AttemptNum:   2,
+		Success:      true,
+		StatusCode:   200,
+		ResponseBody: "ok",
+		Error:        "error",
+		Duration:     15,
+	})
+	assertRoundTrip(t, db, &database.TargetTotals{
+		TargetID:          uuid.New().String(),
+		Deliveries:        5,
+		Delivered:         3,
+		Failed:            1,
+		DeliveriesRemoved: 2,
+		FailedRemoved:     1,
+	})
+	assertRoundTrip(t, db, &database.EntrypointTotals{
+		EntrypointID: uuid.New().String(), LastEventAt: at,
+	})
+
+	// A new event database has its one row of event totals already.
+	require.NoError(t, database.AddEventTotals(db, database.EventTotals{
+		Events: 3, EventsRemoved: 1, LastEventAt: &at,
+	}))
+
+	var totals []database.EventTotals
+
+	require.NoError(t, db.Find(&totals).Error)
+	assert.Equal(t, []database.EventTotals{{
+		ID: 1, Events: 3, EventsRemoved: 1, LastEventAt: &at,
+	}}, totals)
+}

internal/database/sqlite_open.go

@@ -118,9 +118,8 @@ const SQLiteFilePerm fs.FileMode = 0o600
 // SQLite still reports the absence rather than this function
 // materializing an empty database the caller asked not to create.
 //
-// Chmod of a file that already exists is what tightens a data
+// The mode is set on every open, whoever created the file, so a
-// directory an earlier build left at 0644 — including a developer's
+// database or sidecar found at a wider mode is tightened too.
-// own scratch directory — without any migration machinery.
 func reserveSQLiteFile(path string, create bool) error {
 	if create {
 		// gosec G304: the path is the database file the caller asked
@@ -242,7 +241,19 @@ func SQLiteDSN(path, mode string) string {
 	)
 	q.Add("_pragma", "journal_mode(WAL)")
 
-	return "file:" + path + "?" + q.Encode()
+	// SQLite reads a file: URI's path only up to the first ? or #, and
+	// decodes %XX escapes in it, so url.URL escapes the path. OmitHost
+	// keeps a relative path as "file:data/x.db"; written
+	// "file://data/x.db", SQLite would reject "data" as the URI's
+	// authority.
+	dsn := url.URL{
+		Scheme:   "file",
+		Path:     path,
+		OmitHost: true,
+		RawQuery: q.Encode(),
+	}
+
+	return dsn.String()
 }
 
 // OpenSQLite opens the SQLite file at path with the service's

internal/database/sqlite_open_test.go

@@ -2,6 +2,8 @@ package database_test
 
 import (
 	"context"
+	"database/sql"
+	"os"
 	"path/filepath"
 	"strings"
 	"testing"
@@ -112,6 +114,79 @@ func TestMainDBAppliesPragmasOnALiveHandle(t *testing.T) {
 	assert.Equal(t, "10000", busy)
 }
 
+// TestOpenSQLiteOpensThePathAskedFor uses a data directory whose name
+// holds ?, # and %. SQLite reads a file: URI's path only up to a ? or
+// #, and decodes %41 to A, so an unescaped path sent every database in
+// that directory to one file outside it.
+func TestOpenSQLiteOpensThePathAskedFor(t *testing.T) {
+	t.Parallel()
+
+	dir := filepath.Join(t.TempDir(), "hooks#1?x%41")
+	require.NoError(t, os.Mkdir(dir, 0o700))
+
+	for _, name := range []string{
+		database.MainDBFileName,
+		"events-" + uuid.New().String() + ".db",
+	} {
+		path := filepath.Join(dir, name)
+
+		sqlDB, err := database.OpenSQLite(
+			path, database.SQLiteModeCreate,
+		)
+		require.NoError(t, err)
+
+		t.Cleanup(func() { assert.NoError(t, sqlDB.Close()) })
+
+		assertOpenedFile(t, sqlDB, path)
+	}
+}
+
+// TestOpenSQLiteOpensARelativePath opens a database at a path relative
+// to the working directory, which is what a relative DATA_DIR such as
+// the README's DATA_DIR=../webhooker-data gives OpenSQLite.
+//
+//nolint:paralleltest // t.Chdir moves the whole process.
+func TestOpenSQLiteOpensARelativePath(t *testing.T) {
+	t.Chdir(t.TempDir())
+	require.NoError(t, os.Mkdir("data", 0o700))
+
+	path := filepath.Join("data", database.MainDBFileName)
+
+	sqlDB, err := database.OpenSQLite(path, database.SQLiteModeCreate)
+	require.NoError(t, err)
+
+	defer func() { require.NoError(t, sqlDB.Close()) }()
+
+	assertOpenedFile(t, sqlDB, path)
+}
+
+// assertOpenedFile checks that sqlDB is the SQLite file at path.
+// database_list names the file SQLite actually opened, as an absolute
+// path with symlinks resolved, so the files are compared, not the names.
+func assertOpenedFile(t *testing.T, sqlDB *sql.DB, path string) {
+	t.Helper()
+
+	var (
+		seq          int
+		schema, file string
+	)
+
+	require.NoError(t, sqlDB.
+		QueryRowContext(t.Context(), "pragma database_list").
+		Scan(&seq, &schema, &file))
+
+	asked, err := os.Stat(path)
+	require.NoError(t, err)
+
+	opened, err := os.Stat(file)
+	require.NoError(t, err)
+
+	assert.True(
+		t, os.SameFile(asked, opened),
+		"asked for %s, SQLite opened %s", path, file,
+	)
+}
+
 // TestConcurrentReaderDoesNotBlockWrites is the unit-scale form of the
 // reproduction in
 // https://git.eeqj.de/sneak/webhooker/issues/256: an operator's

internal/database/webhook_db_manager.go

@@ -14,6 +14,7 @@ import (
 	"gorm.io/gorm"
 	"sneak.berlin/go/webhooker/internal/config"
 	"sneak.berlin/go/webhooker/internal/datadir"
+	"sneak.berlin/go/webhooker/internal/db/migrations"
 	"sneak.berlin/go/webhooker/internal/gormlog"
 )
 
@@ -370,18 +371,7 @@ func (m *WebhookDBManager) openDB(
 		)
 	}
 
-	err = purgeTargetRows(db, m.log, webhookID)
+	err = migrations.Apply(db, migrations.Events)
-	if err != nil {
-		_ = sqlDB.Close()
-
-		return nil, err
-	}
-
-	// Run migrations for event-tier models only
-	err = db.AutoMigrate(
-		&Event{}, &Delivery{}, &DeliveryResult{},
-		&EventTotals{}, &TargetTotals{}, &EntrypointTotals{},
-	)
 	if err != nil {
 		_ = sqlDB.Close()
 
@@ -391,14 +381,15 @@ func (m *WebhookDBManager) openDB(
 		)
 	}
 
-	// A new database gets its row of event totals, all zero. Target
+	err = parseModels(
-	// totals rows are created by the first delivery to each target.
+		db, &Event{}, &Delivery{}, &DeliveryResult{},
-	err = db.FirstOrCreate(&EventTotals{}).Error
+		&EventTotals{}, &TargetTotals{}, &EntrypointTotals{},
+	)
 	if err != nil {
 		_ = sqlDB.Close()
 
 		return nil, fmt.Errorf(
-			"creating event totals for webhook database %s: %w",
+			"connecting to webhook database %s: %w",
 			webhookID, err,
 		)
 	}

internal/db/migrations/archive/000_migration.sql

@@ -0,0 +1,9 @@
+-- Migration 000: the table recording which migrations this file has
+-- had. Applied before the others; it records itself.
+
+CREATE TABLE IF NOT EXISTS schema_migrations (
+    version INTEGER PRIMARY KEY,
+    applied_at DATETIME DEFAULT CURRENT_TIMESTAMP
+);
+
+INSERT OR IGNORE INTO schema_migrations (version) VALUES (0);

internal/db/migrations/archive/001_schema.sql

@@ -0,0 +1,18 @@
+-- Migration 001: the schema of a database target's archive file. The
+-- archivedEvent model in internal/delivery must agree with it.
+
+CREATE TABLE archived_events (
+    id integer PRIMARY KEY AUTOINCREMENT,
+    event_id text,
+    webhook_id text,
+    entrypoint_id text,
+    method text,
+    raw_query text,
+    headers text,
+    body text,
+    content_type text,
+    archived_at datetime
+);
+
+CREATE INDEX idx_archived_events_event_id ON archived_events (event_id);
+CREATE INDEX idx_archived_events_archived_at ON archived_events (archived_at);

internal/db/migrations/events/000_migration.sql

@@ -0,0 +1,9 @@
+-- Migration 000: the table recording which migrations this file has
+-- had. Applied before the others; it records itself.
+
+CREATE TABLE IF NOT EXISTS schema_migrations (
+    version INTEGER PRIMARY KEY,
+    applied_at DATETIME DEFAULT CURRENT_TIMESTAMP
+);
+
+INSERT OR IGNORE INTO schema_migrations (version) VALUES (0);

internal/db/migrations/events/001_schema.sql

@@ -0,0 +1,98 @@
+-- Migration 001: the schema of a webhook's event database. The models
+-- in internal/database must agree with it. The webhook, entrypoint and
+-- target ids here name rows in the main database, which has those
+-- tables; this file does not.
+
+CREATE TABLE events (
+    id uuid,
+    created_at datetime,
+    updated_at datetime,
+    deleted_at datetime,
+    webhook_id uuid NOT NULL,
+    entrypoint_id uuid NOT NULL,
+    method text NOT NULL,
+    raw_query text,
+    headers text,
+    body text,
+    content_type text,
+    body_bytes integer NOT NULL,
+    resubmitted_from_id uuid,
+    PRIMARY KEY (id)
+);
+
+CREATE INDEX idx_events_created_at ON events (created_at);
+CREATE INDEX idx_events_deleted_at ON events (deleted_at);
+CREATE INDEX idx_events_deleted_at_created_at
+    ON events (deleted_at, created_at);
+CREATE INDEX idx_events_resubmitted_from_id
+    ON events (resubmitted_from_id, deleted_at);
+CREATE INDEX idx_events_entrypoint_id
+    ON events (entrypoint_id, deleted_at, resubmitted_from_id, created_at);
+
+CREATE TABLE deliveries (
+    id uuid,
+    created_at datetime,
+    updated_at datetime,
+    deleted_at datetime,
+    event_id uuid NOT NULL,
+    target_id uuid NOT NULL,
+    status text NOT NULL DEFAULT 'pending',
+    finished_at datetime,
+    replay numeric NOT NULL DEFAULT false,
+    PRIMARY KEY (id),
+    CONSTRAINT fk_events_deliveries
+        FOREIGN KEY (event_id) REFERENCES events (id)
+);
+
+CREATE INDEX idx_deliveries_deleted_at ON deliveries (deleted_at);
+CREATE INDEX idx_deliveries_event_id ON deliveries (event_id, deleted_at);
+CREATE INDEX idx_deliveries_status
+    ON deliveries (status, deleted_at, finished_at, target_id);
+
+CREATE TABLE delivery_results (
+    id uuid,
+    created_at datetime,
+    updated_at datetime,
+    deleted_at datetime,
+    delivery_id uuid NOT NULL,
+    attempt_num integer NOT NULL,
+    success numeric,
+    status_code integer,
+    response_body text,
+    error text,
+    duration integer,
+    PRIMARY KEY (id),
+    CONSTRAINT fk_deliveries_delivery_results
+        FOREIGN KEY (delivery_id) REFERENCES deliveries (id)
+);
+
+CREATE INDEX idx_delivery_results_deleted_at
+    ON delivery_results (deleted_at);
+CREATE INDEX idx_delivery_results_delivery_id
+    ON delivery_results (delivery_id, deleted_at);
+
+CREATE TABLE event_totals (
+    id integer PRIMARY KEY AUTOINCREMENT,
+    events integer NOT NULL,
+    events_removed integer NOT NULL,
+    last_event_at datetime
+);
+
+-- The one row of event totals, all zero.
+INSERT INTO event_totals (events, events_removed) VALUES (0, 0);
+
+CREATE TABLE target_totals (
+    target_id uuid,
+    deliveries integer NOT NULL,
+    delivered integer NOT NULL,
+    failed integer NOT NULL,
+    deliveries_removed integer NOT NULL,
+    failed_removed integer NOT NULL,
+    PRIMARY KEY (target_id)
+);
+
+CREATE TABLE entrypoint_totals (
+    entrypoint_id uuid,
+    last_event_at datetime NOT NULL,
+    PRIMARY KEY (entrypoint_id)
+);

internal/db/migrations/main/000_migration.sql

@@ -0,0 +1,9 @@
+-- Migration 000: the table recording which migrations this file has
+-- had. Applied before the others; it records itself.
+
+CREATE TABLE IF NOT EXISTS schema_migrations (
+    version INTEGER PRIMARY KEY,
+    applied_at DATETIME DEFAULT CURRENT_TIMESTAMP
+);
+
+INSERT OR IGNORE INTO schema_migrations (version) VALUES (0);

internal/db/migrations/main/001_schema.sql

@@ -0,0 +1,89 @@
+-- Migration 001: the main database's schema. The models in
+-- internal/database must agree with it.
+
+CREATE TABLE settings (
+    "key" text,
+    value text NOT NULL,
+    PRIMARY KEY ("key")
+);
+
+CREATE TABLE users (
+    id uuid,
+    created_at datetime,
+    updated_at datetime,
+    deleted_at datetime,
+    username text NOT NULL,
+    password text NOT NULL,
+    PRIMARY KEY (id),
+    CONSTRAINT chk_users_username
+        CHECK (length(CAST(username AS BLOB)) <= 1024)
+);
+
+CREATE INDEX idx_users_deleted_at ON users (deleted_at);
+CREATE UNIQUE INDEX idx_users_username ON users (username);
+
+CREATE TABLE api_keys (
+    id uuid,
+    created_at datetime,
+    updated_at datetime,
+    deleted_at datetime,
+    user_id uuid NOT NULL,
+    "key" text NOT NULL,
+    description text,
+    last_used_at datetime,
+    PRIMARY KEY (id),
+    CONSTRAINT fk_users_api_keys FOREIGN KEY (user_id) REFERENCES users (id)
+);
+
+CREATE INDEX idx_api_keys_deleted_at ON api_keys (deleted_at);
+CREATE UNIQUE INDEX idx_api_keys_key ON api_keys ("key");
+
+CREATE TABLE webhooks (
+    id uuid,
+    created_at datetime,
+    updated_at datetime,
+    deleted_at datetime,
+    user_id uuid NOT NULL,
+    name text NOT NULL,
+    description text,
+    retention_days integer DEFAULT 30,
+    PRIMARY KEY (id),
+    CONSTRAINT fk_users_webhooks FOREIGN KEY (user_id) REFERENCES users (id)
+);
+
+CREATE INDEX idx_webhooks_deleted_at ON webhooks (deleted_at);
+
+CREATE TABLE entrypoints (
+    id uuid,
+    created_at datetime,
+    updated_at datetime,
+    deleted_at datetime,
+    webhook_id uuid NOT NULL,
+    path text NOT NULL,
+    description text,
+    active numeric DEFAULT true,
+    PRIMARY KEY (id),
+    CONSTRAINT fk_webhooks_entrypoints
+        FOREIGN KEY (webhook_id) REFERENCES webhooks (id)
+);
+
+CREATE INDEX idx_entrypoints_deleted_at ON entrypoints (deleted_at);
+CREATE UNIQUE INDEX idx_entrypoints_path ON entrypoints (path);
+
+CREATE TABLE targets (
+    id uuid,
+    created_at datetime,
+    updated_at datetime,
+    deleted_at datetime,
+    webhook_id uuid NOT NULL,
+    name text NOT NULL,
+    type text NOT NULL,
+    active numeric DEFAULT true,
+    config text,
+    max_retries integer,
+    PRIMARY KEY (id),
+    CONSTRAINT fk_webhooks_targets
+        FOREIGN KEY (webhook_id) REFERENCES webhooks (id)
+);
+
+CREATE INDEX idx_targets_deleted_at ON targets (deleted_at);

internal/db/migrations/migrations.go

@@ -0,0 +1,135 @@
+// Package migrations holds the schema of each kind of SQLite file
+// webhooker writes, as numbered SQL files embedded in the binary, and
+// applies it when a file is opened.
+//
+// Each kind of file has its own directory. In each, 000_migration.sql
+// creates only the schema_migrations table, and 001_schema.sql holds
+// the whole schema. Before 1.0 a schema change is an edit to
+// 001_schema.sql, never a new file.
+package migrations
+
+import (
+	"embed"
+	"fmt"
+	"io/fs"
+	"path"
+	"strconv"
+	"strings"
+
+	"gorm.io/gorm"
+)
+
+// The directories, one for each kind of database file.
+const (
+	// Main is the main database, webhooker.db.
+	Main = "main"
+
+	// Events is a webhook's event database, events-{uuid}.db.
+	Events = "events"
+
+	// Archive is a database target's archive file.
+	Archive = "archive"
+)
+
+//go:embed main/*.sql events/*.sql archive/*.sql
+var files embed.FS
+
+// Apply runs each SQL file in dir, which is Main, Events or Archive,
+// that the database has not had yet, in the order of the number its
+// name starts with, and records that number in schema_migrations.
+// 000_migration.sql creates that table and records itself, so it runs
+// first, and only when the table is missing.
+//
+// It all runs in one transaction, so a failure leaves the file as it
+// was.
+func Apply(db *gorm.DB, dir string) error {
+	return db.Transaction(func(tx *gorm.DB) error {
+		var tables int
+
+		// Row().Scan, not (*gorm.DB).Scan: see internal/gormlog.
+		err := tx.Raw(
+			"SELECT count(*) FROM sqlite_master "+
+				"WHERE type = 'table' AND name = ?",
+			"schema_migrations",
+		).Row().Scan(&tables)
+		if err != nil {
+			return fmt.Errorf("looking for schema_migrations: %w", err)
+		}
+
+		if tables == 0 {
+			err = run(tx, path.Join(dir, "000_migration.sql"))
+			if err != nil {
+				return err
+			}
+		}
+
+		// fs.Glob lists the files in name order, which is number order.
+		names, err := fs.Glob(files, path.Join(dir, "*.sql"))
+		if err != nil {
+			return fmt.Errorf("listing migrations: %w", err)
+		}
+
+		for _, name := range names {
+			err = applyOnce(tx, name)
+			if err != nil {
+				return err
+			}
+		}
+
+		return nil
+	})
+}
+
+// applyOnce runs the migration file name and records its number,
+// unless schema_migrations already has that number.
+func applyOnce(tx *gorm.DB, name string) error {
+	number, _, _ := strings.Cut(path.Base(name), "_")
+
+	version, err := strconv.Atoi(number)
+	if err != nil {
+		return fmt.Errorf("numbering migration %s: %w", name, err)
+	}
+
+	var applied int
+
+	err = tx.Raw(
+		"SELECT count(*) FROM schema_migrations WHERE version = ?",
+		version,
+	).Row().Scan(&applied)
+	if err != nil {
+		return fmt.Errorf("checking migration %s: %w", name, err)
+	}
+
+	if applied > 0 {
+		return nil
+	}
+
+	err = run(tx, name)
+	if err != nil {
+		return err
+	}
+
+	err = tx.Exec(
+		"INSERT INTO schema_migrations (version) VALUES (?)", version,
+	).Error
+	if err != nil {
+		return fmt.Errorf("recording migration %s: %w", name, err)
+	}
+
+	return nil
+}
+
+// run executes every statement in the migration file name.
+func run(tx *gorm.DB, name string) error {
+	script, err := files.ReadFile(name)
+	if err != nil {
+		return fmt.Errorf("reading migration %s: %w", name, err)
+	}
+
+	err = tx.Exec(string(script)).Error
+	if err != nil {
+		return fmt.Errorf("applying migration %s: %w", name, err)
+	}
+
+	return nil
+}

internal/db/migrations/migrations_test.go

@@ -0,0 +1,57 @@
+package migrations_test
+
+import (
+	"database/sql"
+	"log/slog"
+	"path/filepath"
+	"testing"
+
+	"github.com/stretchr/testify/assert"
+	"github.com/stretchr/testify/require"
+	"gorm.io/driver/sqlite"
+	"gorm.io/gorm"
+	_ "modernc.org/sqlite"
+	"sneak.berlin/go/webhooker/internal/db/migrations"
+	"sneak.berlin/go/webhooker/internal/gormlog"
+)
+
+// TestApplyRunsEachFileOnce applies each kind of database file's
+// migrations to a new file twice. The first time records every file.
+// The second finds them all recorded and runs none, which is the only
+// way it can succeed: 001_schema.sql creates tables that then exist.
+func TestApplyRunsEachFileOnce(t *testing.T) {
+	t.Parallel()
+
+	for _, dir := range []string{
+		migrations.Main, migrations.Events, migrations.Archive,
+	} {
+		t.Run(dir, func(t *testing.T) {
+			t.Parallel()
+
+			sqlDB, err := sql.Open(
+				"sqlite", filepath.Join(t.TempDir(), "test.db"),
+			)
+			require.NoError(t, err)
+
+			t.Cleanup(func() { _ = sqlDB.Close() })
+
+			db, err := gorm.Open(
+				sqlite.Dialector{Conn: sqlDB},
+				&gorm.Config{
+					Logger: gormlog.New(slog.New(slog.DiscardHandler)),
+				},
+			)
+			require.NoError(t, err)
+
+			require.NoError(t, migrations.Apply(db, dir))
+			require.NoError(t, migrations.Apply(db, dir))
+
+			var versions string
+
+			require.NoError(t, sqlDB.QueryRowContext(t.Context(),
+				"SELECT group_concat(version) FROM schema_migrations",
+			).Scan(&versions))
+			assert.Equal(t, "0,1", versions)
+		})
+	}
+}

internal/delivery/archive_sweeper.go

@@ -201,8 +201,7 @@ func (s *ArchiveSweeper) sweep(ctx context.Context) {
 
 // sweepTarget prunes the archive of a single database target.
 // A missing, empty, or "never" expiry parses as a zero duration
-// and is skipped entirely, so those archives keep exactly the
+// and is skipped entirely, so those archives are never pruned.
-// behaviour they had before the sweep existed.
 func (s *ArchiveSweeper) sweepTarget(target *database.Target) {
 	expiry, err := parseArchiveExpiry(target.Config)
 	if err != nil {

internal/delivery/archive_sweeper_test.go

@@ -22,6 +22,7 @@ import (
 	_ "modernc.org/sqlite" // Pure Go SQLite driver.
 	"sneak.berlin/go/webhooker/internal/database"
 	"sneak.berlin/go/webhooker/internal/database/databasetest"
+	"sneak.berlin/go/webhooker/internal/db/migrations"
 	"sneak.berlin/go/webhooker/internal/delivery"
 	"sneak.berlin/go/webhooker/internal/gormlog"
 )
@@ -172,9 +173,7 @@ func seedArchiveFile(
 	)
 	require.NoError(t, err)
 
-	require.NoError(
+	require.NoError(t, migrations.Apply(gdb, migrations.Archive))
-		t, gdb.AutoMigrate(&delivery.ExportArchivedEvent{}),
-	)
 
 	for i, at := range archivedAt {
 		row := delivery.ExportArchivedEvent{
@@ -781,9 +780,9 @@ func TestArchiveSweep_NeverExpiryUntouched(t *testing.T) {
 //
 // The spec is stronger than that — a "never" archive is skipped
 // before any file is touched — so the archive here exists but has
-// never been migrated. Opening it at all would run AutoMigrate
+// never been migrated. Opening it at all would apply the archive
-// and create the archive table, which is exactly what must not
+// schema and create the archive table, which is exactly what must
-// happen.
+// not happen.
 func TestArchiveSweep_NeverExpirySkipsBeforeOpening(
 	t *testing.T,
 ) {
@@ -807,7 +806,7 @@ func TestArchiveSweep_NeverExpirySkipsBeforeOpening(
 
 // seedUnmigratedArchive creates an archive file that exists but
 // carries no archive schema, so any open of it is observable: the
-// archive table appears only if something ran AutoMigrate.
+// archive table appears only if something applied the schema.
 func seedUnmigratedArchive(t *testing.T, path string) {
 	t.Helper()
 

internal/delivery/delivery_totals_test.go

@@ -32,8 +32,9 @@ func targetTotals(
 
 // TestUpdateDeliveryStatus_FinishTimeAndTargetTotals pins what a status
 // write records for the webhook page's statistics: the time a delivery
-// finished, set only when it becomes delivered or failed, and one more
+// finished, set only when it becomes delivered or failed and stored in
-// on its target's delivered or failed total.
+// the same form as created_at, and one more on its target's delivered
+// or failed total.
 func TestUpdateDeliveryStatus_FinishTimeAndTargetTotals(t *testing.T) {
 	t.Parallel()
 
@@ -75,6 +76,16 @@ func TestUpdateDeliveryStatus_FinishTimeAndTargetTotals(t *testing.T) {
 			if tt.finished {
 				require.NotNil(t, stored.FinishedAt)
 				assert.False(t, stored.FinishedAt.Before(before))
+
+				var createdAt, finishedAt string
+
+				require.NoError(t, db.Raw(
+					`SELECT CAST(created_at AS TEXT),
+						CAST(finished_at AS TEXT)
+					FROM deliveries WHERE id = ?`, d.ID,
+				).Row().Scan(&createdAt, &finishedAt))
+				assertStoredTime(t, createdAt)
+				assertStoredTime(t, finishedAt)
 			} else {
 				assert.Nil(t, stored.FinishedAt)
 			}

internal/delivery/engine.go

@@ -1701,8 +1701,11 @@ func writeDeliveryStatus(
 			database.DeliveryStatusFailed,
 		}).
 		Updates(map[string]any{
-			"status":      status,
+			"status": status,
-			"finished_at": time.Now(),
+			// GORM's clock, the one that stamps created_at. A bare
+			// time.Now() carries a monotonic clock reading, which the
+			// driver would store as an " m=+…" suffix.
+			"finished_at": tx.NowFunc(),
 		})
 	if res.Error != nil || res.RowsAffected == 0 {
 		return res.Error
@@ -1906,12 +1909,13 @@ func (e *Engine) takeForRedispatch(
 		return false
 	}
 
+	// GORM's clock, as writeDeliveryStatus explains.
 	res := webhookDB.
 		Model(&database.Delivery{}).
 		Where(
 			"id = ? AND status = ?", deliveryID, eligible,
 		).
-		UpdateColumn("updated_at", time.Now())
+		UpdateColumn("updated_at", webhookDB.NowFunc())
 
 	if res.Error != nil {
 		e.log.Error(

internal/delivery/engine_integration_test.go

@@ -963,29 +963,12 @@ func TestFailUnretryableRetry_WritesNoTargetRow(
 		database.DeliveryStatusFailed,
 	)
 
-	// The table exists in the per-webhook database because GORM
+	// The per-webhook database has no targets table, so writing a
-	// migrates the Delivery relation's model alongside it. It must
+	// target row there would have failed the status write above.
-	// stay empty.
+	assert.False(t, s.WebhookDB.Migrator().HasTable("targets"),
-	var targetRows int64
-
-	require.NoError(t, s.WebhookDB.
-		Table("targets").
-		Count(&targetRows).Error)
-
-	assert.Zero(t, targetRows,
 		"orphaned-retry terminal failure wrote a target row "+
 			"into the per-webhook event database",
 	)
-
-	var configs []string
-
-	require.NoError(t, s.WebhookDB.
-		Table("targets").
-		Pluck("config", &configs).Error)
-
-	assert.NotContains(
-		t, strings.Join(configs, " "), hookURL,
-	)
 }
 
 func TestRecoverSingleRetry_UnknownTargetType(

internal/delivery/engine_lifecycle_test.go

@@ -2,11 +2,16 @@ package delivery_test
 
 import (
 	"context"
+	"io"
+	"net/http"
+	"net/http/httptest"
 	"path/filepath"
+	"sync"
 	"testing"
 	"time"
 
 	"github.com/google/uuid"
+	"github.com/stretchr/testify/assert"
 	"github.com/stretchr/testify/require"
 	"go.uber.org/fx"
 	"gorm.io/gorm/clause"
@@ -272,6 +277,154 @@ func TestEngine_StopHookHonoursStopTimeout(t *testing.T) {
 	requireStopHookExpires(t, lc.hooks[0], "delivery engine")
 }
 
+// seedHeldDelivery gives the setup's webhook an http target with
+// maxRetries, pointed at a receiver that keeps every request open until
+// the client gives up on it, and seeds one delivery to that target at
+// status. It returns the task that sends attempt attemptNum of the
+// delivery, and a channel that is closed when a request reaches the
+// receiver.
+func seedHeldDelivery(
+	t *testing.T,
+	s iSetup,
+	maxRetries int,
+	status database.DeliveryStatus,
+	attemptNum int,
+) (delivery.Task, <-chan struct{}) {
+	t.Helper()
+
+	arrived := make(chan struct{})
+
+	var once sync.Once
+
+	ts := httptest.NewServer(http.HandlerFunc(
+		func(_ http.ResponseWriter, r *http.Request) {
+			// The server notices the client hanging up only once the
+			// body has been read; until then r.Context() never ends.
+			_, _ = io.Copy(io.Discard, r.Body)
+
+			once.Do(func() { close(arrived) })
+
+			<-r.Context().Done()
+		},
+	))
+	t.Cleanup(ts.Close)
+
+	cfg := iHTTPConfig(ts.URL)
+	targetID := uuid.New().String()
+
+	iCreateWebhook(t, s.MainDB, s.WebhookID, "hook")
+	iCreateTarget(
+		t, s.MainDB, targetID, s.WebhookID, "held",
+		database.TargetTypeHTTP, cfg, maxRetries,
+	)
+
+	event := iSeedEvent(t, s.WebhookDB, s.WebhookID, `{"held":true}`)
+	d := iSeedDelivery(t, s.WebhookDB, event.ID, targetID, status)
+
+	bodyStr := event.Body
+
+	return iTask(
+		d, event, s.WebhookID, targetID, "held", cfg,
+		maxRetries, attemptNum, &bodyStr,
+	), arrived
+}
+
+// stopDuringAttempt waits until the receiver holds an attempt open, then
+// stops the engine through its real stop hook.
+func stopDuringAttempt(
+	t *testing.T, lc *recordingLifecycle, arrived <-chan struct{},
+) {
+	t.Helper()
+
+	select {
+	case <-arrived:
+	case <-time.After(5 * time.Second):
+		t.Fatal("the delivery attempt never reached the receiver")
+	}
+
+	require.NoError(t, lc.hooks[0].OnStop(context.Background()))
+}
+
+// requireRequeued runs restart recovery on the stopped engine and
+// requires it to queue task's attempt again on ch.
+func requireRequeued(
+	t *testing.T, s iSetup, ch <-chan delivery.Task, task delivery.Task,
+) {
+	t.Helper()
+
+	s.Engine.ExportRecoverWebhookDeliveries(
+		context.Background(), s.WebhookID,
+	)
+
+	select {
+	case got := <-ch:
+		assert.Equal(t, task.DeliveryID, got.DeliveryID)
+		assert.Equal(t, task.AttemptNum, got.AttemptNum)
+	case <-time.After(5 * time.Second):
+		t.Fatal("restart recovery did not queue the delivery again")
+	}
+}
+
+// TestEngine_StopLeavesFireAndForgetAttemptPending is the regression
+// test for a stop that turned an attempt in flight into a terminal
+// failure. Stopping the engine cancels the request; for a target with
+// no retries that was the delivery's only attempt, recorded as failed
+// with "context canceled", and nothing ever sent it again. The
+// delivery must instead stay pending, with no result and no count in
+// the attempt metrics, for restart recovery to send.
+func TestEngine_StopLeavesFireAndForgetAttemptPending(t *testing.T) {
+	t.Parallel()
+
+	s := newISetup(t)
+	reg := mIsolate(t, s)
+	lc := startEngineViaHook(t, s.Engine)
+
+	task, arrived := seedHeldDelivery(
+		t, s, 0, database.DeliveryStatusPending, 1,
+	)
+
+	s.Engine.Notify([]delivery.Task{task})
+	stopDuringAttempt(t, lc, arrived)
+
+	iAssertStatus(
+		t, s.WebhookDB, task.DeliveryID, database.DeliveryStatusPending,
+	)
+	assert.Empty(t, iResults(t, s.WebhookDB, task.DeliveryID))
+	assert.Zero(t, mCounter(t, reg, mAttempts, mTypeHTTP))
+	assert.Zero(t, mHTTPDurations(t, reg))
+
+	requireRequeued(t, s, s.Engine.ExportDeliveryCh(), task)
+}
+
+// TestEngine_StopLeavesLastRetryAttemptRetrying is the same regression
+// for a target with retries, on its last attempt: the cancelled attempt
+// used to exhaust the retries and end the delivery failed. It must stay
+// retrying with only its earlier attempt recorded, and restart recovery
+// must send that last attempt again.
+func TestEngine_StopLeavesLastRetryAttemptRetrying(t *testing.T) {
+	t.Parallel()
+
+	const maxRetries = 2
+
+	s := newISetup(t)
+	lc := startEngineViaHook(t, s.Engine)
+
+	task, arrived := seedHeldDelivery(
+		t, s, maxRetries, database.DeliveryStatusRetrying, maxRetries,
+	)
+	iSeedFailedResult(t, s.WebhookDB, task.DeliveryID)
+
+	s.Engine.ExportScheduleRetry(task, 0)
+	stopDuringAttempt(t, lc, arrived)
+
+	iAssertStatus(
+		t, s.WebhookDB, task.DeliveryID, database.DeliveryStatusRetrying,
+	)
+	assert.Len(t, iResults(t, s.WebhookDB, task.DeliveryID), 1)
+
+	requireRequeued(t, s, s.Engine.ExportRetryCh(), task)
+}
+
 // deliverToArchive gives the setup's webhook a database target,
 // runs one delivery to it through the running engine, and returns
 // the target's ID and archive file path. The archive writer holds

internal/delivery/engine_test.go

@@ -10,7 +10,6 @@ import (
 	"net/http"
 	"net/http/httptest"
 	"os"
-	"path/filepath"
 	"strconv"
 	"strings"
 	"sync"
@@ -22,12 +21,11 @@ import (
 	"github.com/prometheus/client_golang/prometheus"
 	"github.com/stretchr/testify/assert"
 	"github.com/stretchr/testify/require"
-	"gorm.io/driver/sqlite"
 	"gorm.io/gorm"
 	_ "modernc.org/sqlite"
 	"sneak.berlin/go/webhooker/internal/database"
+	"sneak.berlin/go/webhooker/internal/database/databasetest"
 	"sneak.berlin/go/webhooker/internal/delivery"
-	"sneak.berlin/go/webhooker/internal/gormlog"
 	"sneak.berlin/go/webhooker/internal/metrics"
 )
 
@@ -37,35 +35,18 @@ const testContentType = "application/json"
 func testWebhookDB(t *testing.T) *gorm.DB {
 	t.Helper()
 
-	dbPath := filepath.Join(
+	// Opened by the manager the service uses, so these tests cannot
-		t.TempDir(), "events-test.db",
+	// pass against a schema, journal and locking settings or write
+	// guard production does not use.
+	mgr := databasetest.NewWebhookDBManagerWithLogger(
+		t, t.TempDir(), slog.New(slog.DiscardHandler),
 	)
 
-	// Opened the way the service opens a per-webhook database, so
+	t.Cleanup(func() { _ = mgr.CloseAll() })
-	// these tests cannot pass against journal and locking settings
+
-	// production does not use.
+	db, err := mgr.GetDB(uuid.New().String())
-	sqlDB, err := database.OpenSQLite(
-		dbPath, database.SQLiteModeCreate,
-	)
 	require.NoError(t, err)
 
-	t.Cleanup(func() { _ = sqlDB.Close() })
-
-	db, err := gorm.Open(
-		sqlite.Dialector{Conn: sqlDB},
-		&gorm.Config{Logger: gormlog.New(slog.New(slog.DiscardHandler))},
-	)
-	require.NoError(t, err)
-
-	require.NoError(t, db.AutoMigrate(
-		&database.Event{},
-		&database.Delivery{},
-		&database.DeliveryResult{},
-		&database.EventTotals{},
-		&database.TargetTotals{},
-	))
-	require.NoError(t, db.Create(&database.EventTotals{}).Error)
-
 	return db
 }
 
@@ -136,6 +117,18 @@ func seedDelivery(
 	return d
 }
 
+// assertStoredTime asserts that text, a timestamp as SQLite holds it,
+// is in the form GORM stores created_at in: the layout of
+// time.Time.String, with no " m=+…" monotonic clock reading after it.
+func assertStoredTime(t *testing.T, text string) {
+	t.Helper()
+
+	_, err := time.Parse(
+		"2006-01-02 15:04:05.999999999 -0700 MST", text,
+	)
+	assert.NoError(t, err, "stored as %q", text)
+}
+
 // httpDeliveryFixture holds the task and delivery needed
 // for HTTP delivery tests.
 type httpDeliveryFixture struct {

internal/delivery/inflight_test.go

@@ -462,7 +462,8 @@ func TestScheduledRetryIsNotSweptDuringBackoff(t *testing.T) {
 
 // TestRedispatchStampsTheRow pins the cadence control: a stranded
 // delivery that has just been handed out is not selected again by the
-// next tick a minute later.
+// next tick a minute later. The stamp is stored in the same form as
+// created_at.
 func TestRedispatchStampsTheRow(t *testing.T) {
 	t.Parallel()
 
@@ -493,4 +494,13 @@ func TestRedispatchStampsTheRow(t *testing.T) {
 		"a re-dispatched delivery must be stamped so the "+
 			"next tick does not select it again",
 	)
+
+	var createdAt, updatedAt string
+
+	require.NoError(t, s.WebhookDB.Raw(
+		`SELECT CAST(created_at AS TEXT), CAST(updated_at AS TEXT)
+		FROM deliveries WHERE id = ?`, d.ID,
+	).Row().Scan(&createdAt, &updatedAt))
+	assertStoredTime(t, createdAt)
+	assertStoredTime(t, updatedAt)
 }

internal/delivery/target.go

@@ -32,7 +32,8 @@ type Scheduler interface {
 // Engine.observeAttempt, alongside the DeliveryResult it records for
 // it. Deliver is also entered for attempts that never happen — an
 // open circuit breaker refuses one — so the count cannot be taken
-// from around this call.
+// from around this call. An attempt cut off by a stop is neither
+// counted nor recorded.
 type Target interface {
 	Deliver(
 		ctx context.Context,

internal/delivery/target_database_archive.go

@@ -14,6 +14,7 @@ import (
 	"gorm.io/driver/sqlite"
 	"gorm.io/gorm"
 	"sneak.berlin/go/webhooker/internal/database"
+	"sneak.berlin/go/webhooker/internal/db/migrations"
 	"sneak.berlin/go/webhooker/internal/gormlog"
 )
 
@@ -276,7 +277,11 @@ func (w *archiveWriter) write(
 		}
 	}
 
-	row.ArchivedAt = time.Now()
+	// GORM's clock, the one that stamps created_at in the other
+	// databases, so this is stored in the same form. A bare time.Now()
+	// carries a monotonic clock reading, which the driver would store
+	// as an " m=+…" suffix.
+	row.ArchivedAt = w.db.NowFunc()
 
 	err := w.db.Create(&row).Error
 	if err != nil {
@@ -335,7 +340,7 @@ func (w *archiveWriter) openMode(
 		)
 	}
 
-	err = gdb.AutoMigrate(&archivedEvent{})
+	err = migrations.Apply(gdb, migrations.Archive)
 	if err != nil {
 		_ = sqlDB.Close()
 

internal/delivery/target_database_test.go

@@ -150,6 +150,13 @@ func TestArchiveWriter_WritesRow(t *testing.T) {
 	assert.Equal(t, row.ContentType, got[0].ContentType)
 	assert.JSONEq(t, `{"hello":"world"}`, got[0].Body)
 	assert.False(t, got[0].ArchivedAt.IsZero())
+
+	var archivedAt string
+
+	require.NoError(t, w.DB().Raw(
+		"SELECT CAST(archived_at AS TEXT) FROM archived_events",
+	).Row().Scan(&archivedAt))
+	assertStoredTime(t, archivedAt)
 }
 
 func TestArchiveWriter_RecreatesAfterRemoval(

internal/delivery/target_headers.go

@@ -82,8 +82,8 @@ func isReservedTargetHeader(name string) bool {
 // one pair silently overwriting the other.
 //
 // An input with no pairs yields an empty map, which omitempty drops
-// from the stored config: a target configured with no headers keeps
+// from the stored config, so a target configured with no headers
-// the same config JSON it had before this field existed.
+// stores no headers key at all.
 func ParseTargetHeaders(raw string) (map[string]string, error) {
 	headers := make(map[string]string)
 

internal/delivery/target_http.go

@@ -57,7 +57,11 @@ type httpCore struct {
 // recorded and no circuit breaker is consulted. A positive
 // maxRetries gates the attempt on the circuit breaker and
 // schedules a backed-off retry on failure.
+//
+// ctx is the context attempt sends its request with. Stopping the
+// engine cancels it; see interruptedByStop.
 func (c *httpCore) deliver(
+	ctx context.Context,
 	webhookDB *gorm.DB,
 	d *database.Delivery,
 	task *Task,
@@ -66,21 +70,55 @@ func (c *httpCore) deliver(
 	attempt func() attemptResult,
 ) {
 	if maxRetries == 0 {
-		c.fireAndForget(webhookDB, d, attempt())
+		c.fireAndForget(ctx, webhookDB, d, attempt())
 
 		return
 	}
 
 	c.withRetry(
-		webhookDB, d, task, sched, maxRetries, attempt,
+		ctx, webhookDB, d, task, sched, maxRetries, attempt,
 	)
 }
 
+// interruptedByStop reports whether an attempt failed because the
+// engine is stopping, which cancels the context its request is made
+// with. Such an attempt is not recorded: no DeliveryResult, no circuit
+// breaker outcome, no status change, no count in the attempt metrics.
+// The delivery stays pending or retrying, and startup recovery sends
+// it again on the next start.
+//
+// A target's own timeout leaves that context alone, so its attempt is
+// still recorded as the failure it is.
+func (c *httpCore) interruptedByStop(
+	ctx context.Context,
+	d *database.Delivery,
+	res attemptResult,
+) bool {
+	if res.success || ctx.Err() == nil {
+		return false
+	}
+
+	c.eng.log.Warn(
+		"delivery attempt interrupted by shutdown; not recorded, "+
+			"it is sent again on the next start",
+		"delivery_id", d.ID,
+		"event_id", d.EventID,
+		"target_id", d.TargetID,
+	)
+
+	return true
+}
+
 func (c *httpCore) fireAndForget(
+	ctx context.Context,
 	webhookDB *gorm.DB,
 	d *database.Delivery,
 	res attemptResult,
 ) {
+	if c.interruptedByStop(ctx, d, res) {
+		return
+	}
+
 	c.eng.observeAttempt(d.Target.Type, res.elapsed())
 
 	err := c.eng.recordResult(
@@ -110,6 +148,7 @@ func (c *httpCore) fireAndForget(
 }
 
 func (c *httpCore) withRetry(
+	ctx context.Context,
 	webhookDB *gorm.DB,
 	d *database.Delivery,
 	task *Task,
@@ -131,6 +170,10 @@ func (c *httpCore) withRetry(
 
 	res := attempt()
 
+	if c.interruptedByStop(ctx, d, res) {
+		return
+	}
+
 	c.eng.observeAttempt(d.Target.Type, res.elapsed())
 
 	err := c.eng.recordResult(
@@ -395,7 +438,7 @@ func (t *httpTarget) Deliver(
 	}
 
 	t.deliver(
-		webhookDB, d, task, sched,
+		ctx, webhookDB, d, task, sched,
 		d.Target.MaxRetries, attempt,
 	)
 }

internal/delivery/target_redact.go

@@ -13,9 +13,9 @@ import (
 const RedactionMarker = "(redacted)"
 
 // Redactor removes one target's own credential material from
-// text that target's remote peer chose: a delivery response
+// a delivery response body, whose text the target's remote
-// body, or a delivery error stored before the delivery path
+// peer chose, and from a stored delivery error as a second
-// learned to mask the URLs it embeds.
+// line behind the delivery path's own URL masking.
 //
 // It removes byte-identical echoes of strings taken from the
 // target's stored configuration, and nothing else. Anything

internal/delivery/target_slack.go

@@ -82,7 +82,7 @@ func (t *slackTarget) Deliver(
 	}
 
 	t.deliver(
-		webhookDB, d, task, sched,
+		ctx, webhookDB, d, task, sched,
 		d.Target.MaxRetries, attempt,
 	)
 }

internal/delivery/terminal_state_test.go

@@ -312,13 +312,9 @@ func TestFailMissingTarget_WritesNoTargetRow(
 		database.DeliveryStatusFailed,
 	)
 
-	var configs []string
+	// The per-webhook database has no targets table, so writing a
-
+	// target row there would have failed the status write above.
-	require.NoError(t, s.WebhookDB.
+	assert.False(t, s.WebhookDB.Migrator().HasTable("targets"),
-		Table("targets").
-		Pluck("config", &configs).Error)
-
-	assert.Empty(t, configs,
 		"the deleted-target terminal path wrote a target row "+
 			"into the per-webhook event database",
 	)

internal/gormlog/gormlog_test.go

@@ -135,9 +135,11 @@ func openDB(
 	)
 	require.NoError(t, err)
 
-	require.NoError(t, gdb.AutoMigrate(&thing{}))
+	require.NoError(t, gdb.Exec(
+		"CREATE TABLE things (id text PRIMARY KEY, name text)",
+	).Error)
 
-	// Migration chatter is not what any of these cases is about.
+	// Creating the table is not what any of these cases is about.
 	buf.Reset()
 
 	return gdb

internal/gormlog/scan_guard_test.go

@@ -115,8 +115,8 @@ func skipDir(name string) bool {
 }
 
 // walkNonTestGo parses every non-test .go file under root. It returns
-// the directories, relative to root, it parsed a file in, along with
+// the directories, relative to root, it parsed a file in or below,
-// every unguarded Scan it found.
+// along with every unguarded Scan it found.
 func walkNonTestGo(t *testing.T, root string) (map[string]bool, []string) {
 	t.Helper()
 
@@ -155,7 +155,12 @@ func walkNonTestGo(t *testing.T, root string) (map[string]bool, []string) {
 				return err
 			}
 
-			walked[dir] = true
+			// The directories above count too, so that one holding
+			// only a nested package, as internal/db does, is not
+			// taken for one the walk skipped.
+			for ; dir != "."; dir = filepath.Dir(dir) {
+				walked[dir] = true
+			}
 
 			for _, pos := range unguardedScans(fset, file) {
 				hits = append(hits, relPosition(root, pos))
@@ -205,9 +210,9 @@ func TestGormScanIsNeverCalledOutsideTests(t *testing.T) {
 	walked, offenders := walkNonTestGo(t, root)
 
 	// The module's packages are static, templates, and every directory
-	// directly under cmd and internal. Each holds non-test code, so one
+	// directly under cmd and internal. Each holds non-test code, itself
-	// the walk parsed nothing in was skipped, and a Scan there would
+	// or in a package below it, so one the walk parsed nothing in or
-	// pass unseen.
+	// below was skipped, and a Scan there would pass unseen.
 	packages := []string{"static", "templates"}
 
 	for _, parent := range []string{"cmd", "internal"} {

internal/handlers/delivery_attempts_test.go

@@ -43,7 +43,7 @@ func TestDeliveryAttempts_ReadInTheTargetTypesOwnTerms(t *testing.T) {
 	}
 
 	for _, tc := range cases {
-		t.Run(string(tc.targetType), func(t *testing.T) {
+		t.Run(string(tc.targetType)+" "+tc.outcome, func(t *testing.T) {
 			t.Parallel()
 
 			f := newRecentEventsFixture(t)

internal/handlers/delivery_replay_test.go

@@ -314,21 +314,15 @@ func assertReplayTask(
 	)
 }
 
-// assertNoLeakedTarget proves the per-webhook database holds no target
+// assertNoLeakedTarget proves the per-webhook database holds no
-// rows. AutoMigrate creates the table there because Delivery declares
+// target: an association write would have upserted the whole target,
-// the relation, so it is a ROW that signals a leak: an association
+// plaintext config and all, into the event database. See
-// write would have upserted the whole target, plaintext config and
-// all, into the event database. See
 // https://git.eeqj.de/sneak/webhooker/issues/206.
 func assertNoLeakedTarget(t *testing.T, webhookDB *gorm.DB) {
 	t.Helper()
 
-	var leaked int64
+	assert.False(
-
+		t, webhookDB.Migrator().HasTable(&database.Target{}),
-	require.NoError(t, webhookDB.Unscoped().
-		Model(&database.Target{}).Count(&leaked).Error)
-	assert.Zero(
-		t, leaked,
 		"replay must not write the target into the event database",
 	)
 }

internal/handlers/delivery_result_view.go

@@ -12,13 +12,12 @@ import (
 //
 // The bound is enforced here and in SQL, because this page's
 // memory profile must not depend on a constant in another
-// package staying where it is, and because rows predating the
+// package staying where it is, or on every stored row having
-// delivery engine's own cap or restored from an archive are
+// been written by the delivery engine under its own cap.
-// not covered by it at all.
 //
 // It happens to equal that engine cap, so a row written by the
-// current engine reaches this one exactly and is never cut
+// engine reaches this one exactly and is never cut twice.
-// twice. Nothing here may assume the two differ: see view.
+// Nothing here may assume the two differ: see view.
 const maxRenderedResponseBytes = 4096
 
 // deliveryResultColumns is the delivery attempt projection.
@@ -40,9 +39,8 @@ const deliveryResultColumns = "delivery_id, attempt_num, created_at, " +
 //
 // Both Error and ResponseBody have been through the target's
 // Redactor. The engine already masks the URL out of the
-// errors it stores, so for errors this is a second line
+// errors it stores, so for errors this is a second line; for
-// covering rows written before it did; for response bodies it
+// response bodies it is the only line, and its reach is what
-// is the only line, and its reach is what
 // delivery.Redactor documents.
 type DeliveryResultView struct {
 	AttemptNum int

internal/handlers/delivery_result_view_test.go

@@ -158,8 +158,7 @@ func TestHandleSourceLogs_RendersFailedAttempt(t *testing.T) {
 // TestHandleSourceLogs_EscapesResponseBody proves the
 // response body is treated as the untrusted remote content it
 // is. The remote chooses these bytes and the page is rendered
-// inside the operator's authenticated origin, where the
+// inside the operator's authenticated origin.
-// application's own CSP allows inline script from 'self'.
 func TestHandleSourceLogs_EscapesResponseBody(t *testing.T) {
 	t.Parallel()
 

internal/handlers/event_body.go

@@ -35,11 +35,10 @@ const eventBodyQuery = "SELECT cast(body as blob) " +
 // response is deliberately not renderable. Content-Disposition
 // makes the browser download rather than display it, and the
 // octet-stream type plus nosniff stop it being interpreted as
-// HTML or script. Without those a stored payload would execute
+// HTML or script. Without those a stored payload would be shown
-// as the logged-in operator. The application's CSP does not
+// as a page of the operator's own origin. The application's CSP
-// help here: script-src allows 'unsafe-inline' from 'self', so
+// is a second defence behind them: it allows no inline script,
-// a document served from this origin could run its own inline
+// so such a page could not run script of its own.
-// script.
 func (h *Handlers) HandleEventBodyDownload() http.HandlerFunc {
 	return func(w http.ResponseWriter, r *http.Request) {
 		webhook, ok := h.ownedWebhook(w, r)

internal/handlers/event_detail_test.go

@@ -84,7 +84,7 @@ func TestHandleEventDetail_ShowsEventWholeWithDeliveries(t *testing.T) {
 	assert.Contains(t, page, strconv.Itoa(len(body))+" bytes")
 	assert.Contains(t, page, "{\n  "pad": "xxx")
 	assert.Contains(t, page, ""tail": ""+sentinel+""\n}")
-	assert.Contains(t, page, `style="max-height: 32rem; overflow-y: auto"`)
+	assert.Contains(t, page, `break-all max-h-[32rem] overflow-y-auto"`)
 	assert.NotContains(t, page, "Showing the first")
 	assert.Contains(t, page, target.Name)
 	assert.Contains(t, page, ">failed</span>")

internal/handlers/logbound_test.go

@@ -1,19 +1,15 @@
 package handlers_test
 
-// The handler-side half of the log-field audit. Two slog calls in
+// The handler-side half of the log-field audit. The slog calls in
-// this package reach a value an UNAUTHENTICATED client picks outright
+// this package that reach a value an UNAUTHENTICATED client picks
-// and of a length it picks outright:
+// outright, and of a length it picks outright, are the failed-login
+// DEBUG lines, whose username is a form field.
 //
-//   - the unknown-entrypoint DEBUG line on /h/{uuid}, whose
+// They are at DEBUG, which is off in production by default. That is
-//     path segment matched no stored entrypoint and so is bounded by
-//     nothing;
-//   - the failed-login DEBUG lines, whose username is a form field.
-//
-// Both are at DEBUG, which is off in production by default. That is
 // not a bound: an operator turning DEBUG on to diagnose a flood must
-// not thereby hand the flood an unbounded write. Both spend the same
+// not thereby hand the flood an unbounded write. They spend the same
-// internal/logfield budget as the access log, and both are held here
+// internal/logfield budget as the access log, and are held here to
-// to middleware.MaxAccessLogLineBytes.
+// middleware.MaxAccessLogLineBytes.
 //
 // The two login lines past the username lookup — "invalid password"
 // and "user logged in" — carry the same cap without needing it, since
@@ -260,43 +256,6 @@ func postLoginWithPassword(
 	return w.Code
 }
 
-// TestUnknownEntrypoint_LogLineDoesNotTrackPathSize drives 8 KB of
-// client-chosen path at the unauthenticated receiver's
-// unknown-entrypoint DEBUG line and holds it to the same ceiling the
-// access log states.
-func TestUnknownEntrypoint_LogLineDoesNotTrackPathSize(t *testing.T) {
-	t.Parallel()
-
-	for handlerName, newHandler := range logHandlers() {
-		for fillName, fill := range escapeFills() {
-			t.Run(handlerName+"/"+fillName, func(t *testing.T) {
-				t.Parallel()
-
-				h, buf := capturingHandlers(t, newHandler)
-				router := receiverRouter(h)
-
-				for i := range floodRequests {
-					assert.Equal(
-						t,
-						http.StatusNotFound,
-						postReceiver(
-							t, router,
-							oversizedFill(fill)+
-								strings.Repeat("y", i),
-						),
-					)
-				}
-
-				lines := logLines(t, buf)
-				require.Len(t, lines, floodRequests)
-				assertNoClientText(t, buf)
-
-				assertBoundedFlood(t, buf.Len())
-			})
-		}
-	}
-}
-
 // TestFailedLogin_LogLineDoesNotTrackUsernameSize drives 8 KB of
 // client-chosen username at the unauthenticated login endpoint's
 // DEBUG line and holds it to the same ceiling.

internal/handlers/source_detail_test.go

@@ -273,7 +273,7 @@ func TestHandleSourceDetail_FitsWideAndNarrowWindows(t *testing.T) {
 
 	assert.Contains(
 		t, body,
-		`<div class="mx-auto px-6 py-8" style="max-width: 108rem"`,
+		`<div class="max-w-[108rem] mx-auto px-6 py-8">`,
 	)
 	assert.Contains(
 		t, body,
@@ -286,8 +286,8 @@ func TestHandleSourceDetail_FitsWideAndNarrowWindows(t *testing.T) {
 // target and says what deleting it loses, that the webhook's gives its
 // number of stored events (5 received, 2 removed by retention, so 3,
 // the statistics pane's "Within retention" figure), and that an
-// entrypoint with no description is named by its URL. The template
+// entrypoint with no description is named by its URL. Each prompt is
-// writes the slashes after http: as \/, which the browser reads as /.
+// its form's data-confirm attribute, where &#10; is a line break.
 func TestHandleSourceDetail_DeletePromptsNameWhatIsLost(t *testing.T) {
 	t.Parallel()
 
@@ -325,29 +325,28 @@ func TestHandleSourceDetail_DeletePromptsNameWhatIsLost(t *testing.T) {
 	body := renderSourceDetailPage(t, h, sess, wh.ID)
 
 	assert.Contains(t, body,
-		`Delete webhook &quot;delete-me&quot;?\n\n`+
+		`data-confirm="Delete webhook &quot;delete-me&quot;?&#10;&#10;`+
 			`This deletes its stored events (3) and their deliveries. `+
-			`Any archive files it wrote are kept.`)
+			`Any archive files it wrote are kept."`)
 	assert.Contains(t, body,
-		`Delete entrypoint &quot;Stripe&quot;?\n\n`+
+		`data-confirm="Delete entrypoint &quot;Stripe&quot;?&#10;&#10;`+
 			`Senders using its URL get an error from now on, `+
-			`and the URL cannot be restored.`)
+			`and the URL cannot be restored."`)
 	assert.Contains(t, body,
-		`Delete entrypoint &quot;http:\/\/example.com/h/`+
+		`data-confirm="Delete entrypoint &quot;http://example.com/h/`+
-			unnamed.Path+`&quot;?`)
+			unnamed.Path+`&quot;?&#10;&#10;`)
 	assert.Contains(t, body,
-		`Delete target &quot;t-log&quot;?\n\n`+
+		`data-confirm="Delete target &quot;t-log&quot;?&#10;&#10;`+
 			`Nothing more is delivered to it. `+
-			`Its past deliveries stay in the event log.`)
+			`Its past deliveries stay in the event log."`)
 }
 
 // TestHandleSourceDetail_DeletePromptKeepsQuotesInName checks that a
 // webhook name with quotes, a backslash, a closing script tag and a
-// newline reaches its delete prompt escaped for the script, which the
+// newline reaches its delete prompt escaped for the attribute, which
-// browser reads back as the name typed: each quote and angle bracket
+// the browser reads back as the name typed: each quote and angle
-// as a \u escape, the slash as \/, the newline as \n and the backslash
+// bracket as a character reference, the rest as typed. An unescaped
-// doubled. An unescaped newline would break the prompt's script, and
+// double quote would end the attribute and cut the prompt short.
-// the form would then submit without asking.
 func TestHandleSourceDetail_DeletePromptKeepsQuotesInName(t *testing.T) {
 	t.Parallel()
 
@@ -373,6 +372,6 @@ func TestHandleSourceDetail_DeletePromptKeepsQuotesInName(t *testing.T) {
 	body := renderSourceDetailPage(t, h, sess, wh.ID)
 
 	assert.Contains(t, body,
-		"Delete webhook &quot;Bob\\u0027s \\u0022best\\u0022 \\\\ hook"+
+		"Delete webhook &quot;Bob&#39;s &#34;best&#34; \\ hook"+
-			"\\u003c\\/script\\u003e\\nline two&quot;?")
+			"&lt;/script&gt;\nline two&quot;?")
 }

internal/handlers/target_retries.go

@@ -49,13 +49,6 @@ var (
 // working retry configuration over a typo. A default answers a
 // question that was not asked; it never answers one that was asked
 // badly.
-//
-// A target stored with a count above the ceiling before this
-// validation existed keeps rendering and keeps delivering — nothing
-// clamps the row. Re-saving it from the edit form does have to bring
-// it into range, because the form submits the pre-filled value back
-// and accepting it would be the ceiling not applying to the edit
-// path. The 400 names the ceiling, so the fix is one field.
 func parseMaxRetries(raw string, fallback int) (int, error) {
 	raw = strings.TrimSpace(raw)
 	if raw == "" {

internal/handlers/webhook.go

@@ -10,7 +10,6 @@ import (
 	"gorm.io/gorm"
 	"sneak.berlin/go/webhooker/internal/database"
 	"sneak.berlin/go/webhooker/internal/delivery"
-	"sneak.berlin/go/webhooker/internal/logfield"
 	"sneak.berlin/go/webhooker/internal/middleware"
 )
 
@@ -49,14 +48,13 @@ func (h *Handlers) HandleWebhook() http.HandlerFunc {
 			return
 		}
 
-		// Logged only once the UUID is known to name a real
+		// Logged only once the lookup has found the entrypoint, so
-		// entrypoint. The UUID comes straight out of the path on
+		// an invented path writes no INFO line; a miss is logged at
-		// the one unauthenticated endpoint, so logging it before
+		// DEBUG. The entrypoint is named by its ID: the UUID is its
-		// the lookup let a client write an INFO line per invented
+		// credential and is never logged.
-		// path; the request itself is already in the access log
-		// and a miss is already logged at DEBUG.
 		h.log.Info("webhook request received",
-			"entrypoint_uuid", entrypointUUID,
+			"entrypoint_id", entrypoint.ID,
+			"webhook_id", entrypoint.WebhookID,
 			"method", r.Method,
 			"remoteIP", middleware.RemoteIP(r),
 			"clientIP", middleware.ClientIP(r),
@@ -133,17 +131,10 @@ func (h *Handlers) lookupEntrypoint(
 		"path = ?", entrypointUUID,
 	).First(&entrypoint)
 	if result.Error != nil {
-		// The receiver is unauthenticated and /h/{uuid}
+		// Not even an unknown UUID is logged. The receiver's log
-		// matches any single segment, so this value is entirely
+		// lines never carry the path segment, which is a credential
-		// client-chosen on exactly the branch where the lookup
+		// whenever it names an entrypoint.
-		// failed. DEBUG is off by default; the cap is what keeps
+		h.log.Debug("entrypoint not found")
-		// turning it on from restoring an unbounded write.
-		h.log.Debug(
-			"entrypoint not found",
-			"path", logfield.Truncate(
-				entrypointUUID, logfield.MaxBytes,
-			),
-		)
 		http.NotFound(w, r)
 
 		return entrypoint, false

internal/handlers/webhook_test.go

@@ -12,12 +12,15 @@ import (
 	"testing"
 
 	"github.com/go-chi/chi"
+	"github.com/google/uuid"
 	"github.com/stretchr/testify/assert"
 	"github.com/stretchr/testify/require"
+	"gorm.io/gorm/clause"
 	"sneak.berlin/go/webhooker/internal/config"
 	"sneak.berlin/go/webhooker/internal/database"
 	"sneak.berlin/go/webhooker/internal/handlers"
 	"sneak.berlin/go/webhooker/internal/middleware"
+	"sneak.berlin/go/webhooker/internal/middleware/middlewaretest"
 )
 
 // TestHandleWebhook_LogsClientNextToThePeer checks that the
@@ -99,6 +102,87 @@ func TestHandleWebhook_LogsClientNextToThePeer(t *testing.T) {
 	}
 }
 
+// TestHandleWebhook_LogsNoEntrypointUUID sends an accepted event, a
+// request over the receiver rate limit and a request for an unknown
+// UUID through the receiver as routes.go mounts it, and checks that
+// nothing logged, at any level, carries either UUID.
+func TestHandleWebhook_LogsNoEntrypointUUID(t *testing.T) {
+	t.Parallel()
+
+	var (
+		h  *handlers.Handlers
+		db *database.Database
+	)
+
+	cfg := &config.Config{DataDir: t.TempDir(), ReceiverRateLimit: 1}
+
+	app := newTestAppWithConfig(t, cfg, &h, &db)
+	app.RequireStart()
+
+	t.Cleanup(app.RequireStop)
+
+	buf := new(bytes.Buffer)
+	log := slog.New(slog.NewJSONHandler(
+		buf, &slog.HandlerOptions{Level: slog.LevelDebug},
+	))
+	h.SetLogForTest(log)
+	mw := middlewaretest.New(t, log, cfg, nil)
+
+	webhook := seedWebhook(t, db)
+	entrypoint := &database.Entrypoint{
+		WebhookID: webhook.ID,
+		Path:      uuid.New().String(),
+		Active:    true,
+	}
+	require.NoError(t, db.DB().Omit(clause.Associations).
+		Create(entrypoint).Error)
+
+	unknownUUID := uuid.New().String()
+
+	router := chi.NewRouter()
+	router.Use(mw.Logging())
+	router.With(mw.ReceiverRateLimit()).HandleFunc(
+		middleware.ReceiverRoute, h.HandleWebhook(),
+	)
+
+	// In order: the limit is one request per minute per path, so the
+	// second request for the known UUID is the rejected one.
+	for _, step := range []struct {
+		entrypointUUID string
+		wantStatus     int
+	}{
+		{entrypoint.Path, http.StatusOK},
+		{entrypoint.Path, http.StatusTooManyRequests},
+		{unknownUUID, http.StatusNotFound},
+	} {
+		req := httptest.NewRequestWithContext(
+			context.Background(), http.MethodPost,
+			"/h/"+step.entrypointUUID, strings.NewReader("{}"),
+		)
+		w := httptest.NewRecorder()
+		router.ServeHTTP(w, req)
+
+		require.Equal(t, step.wantStatus, w.Code)
+	}
+
+	// The lines this test is about were all written, so the checks
+	// below are not passing on an empty log.
+	for _, msg := range []string{
+		"webhook request received",
+		"webhook receiver rate limit exceeded",
+		"entrypoint not found",
+	} {
+		assert.Contains(t, buf.String(), `"msg":"`+msg+`"`)
+	}
+
+	assert.NotContains(t, buf.String(), entrypoint.Path)
+	assert.NotContains(t, buf.String(), unknownUUID)
+
+	line := receivedLine(t, buf)
+	assert.Equal(t, entrypoint.ID, line["entrypoint_id"])
+	assert.Equal(t, webhook.ID, line["webhook_id"])
+}
+
 // receivedLine returns the one "webhook request received" line in the
 // captured JSON log.
 func receivedLine(t *testing.T, buf *bytes.Buffer) map[string]any {

internal/middleware/accesslog_test.go

@@ -7,6 +7,7 @@ import (
 	"log/slog"
 	"net/http"
 	"net/http/httptest"
+	"strconv"
 	"strings"
 	"testing"
 
@@ -27,6 +28,13 @@ const floodRequests = 64
 // line for a redirected or rejected request may contain it.
 const attackerMarker = "QQATTACKERTEXTQQ"
 
+// knownEntrypointUUID and failingEntrypointUUID are the UUIDs the
+// stand-in receiver in accessLogRouter answers 200 and 500 to.
+const (
+	knownEntrypointUUID   = "6f1e2a9c-4b7d-4e3a-9c2f-1d8b5a7e3c60"
+	failingEntrypointUUID = "0b7c3e5a-2d9f-4a61-8e4b-7c1d6f2a9e35"
+)
+
 // maxLineBytes bounds a single access log line whose client-supplied
 // fields are of ordinary size. Well above what the fixed fields need,
 // well below the length of the oversized input the amplification tests
@@ -132,17 +140,21 @@ func accessLogRouter(m *middleware.Middleware) *chi.Mux {
 	)
 
 	router.HandleFunc(
-		"/h/{uuid}",
+		middleware.ReceiverRoute,
 		func(w http.ResponseWriter, r *http.Request) {
-			// Stands in for the real handler: an unknown entrypoint
+			// Stands in for the real handler: a known entrypoint
-			// UUID 404s, a known one succeeds.
+			// UUID succeeds, a failing one is a server error, and
-			if chi.URLParam(r, "uuid") != "known" {
+			// any other 404s.
+			switch chi.URLParam(r, "uuid") {
+			case knownEntrypointUUID:
+				w.WriteHeader(http.StatusOK)
+			case failingEntrypointUUID:
+				http.Error(
+					w, "boom", http.StatusInternalServerError,
+				)
+			default:
 				http.Error(w, "not found", http.StatusNotFound)
-
-				return
 			}
-
-			w.WriteHeader(http.StatusOK)
 		},
 	)
 
@@ -335,11 +347,13 @@ func oversizedValue(ch string) string {
 }
 
 // oversizedHeaders fills every client-supplied header the access log
-// reads with the same value.
+// reads with the same value. The Referer carries it as the host of an
+// absolute URL, because the access log keeps only a Referer's scheme
+// and host.
 func oversizedHeaders(value string) map[string]string {
 	return map[string]string{
 		"User-Agent":   value,
-		"Referer":      value,
+		"Referer":      "http://" + value,
 		"X-Request-Id": value,
 	}
 }
@@ -649,17 +663,111 @@ func TestAccessLog_SuccessKeepsConcretePathAndRedactsQuery(
 	router := accessLogRouter(m)
 
 	assert.Equal(
-		t, http.StatusOK, get(t, router, "/h/known?src=ci"),
+		t, http.StatusOK,
+		get(t, router, "/.well-known/healthcheck?src=ci"),
 	)
 
-	// The path resolved against a stored entrypoint, so it stays. The
-	// query never does: see TestAccessLog_UnauthenticatedSuccess...
 	entries := accessLogEntries(t, buf)
 	require.Len(t, entries, 1)
-	assert.Equal(t, "/h/known?(redacted)", entries[0]["url"])
+	assert.Equal(
+		t, "/.well-known/healthcheck?(redacted)", entries[0]["url"],
+	)
 	assert.NotContains(t, buf.String(), "src=ci")
 }
 
+// TestAccessLog_ReceiverNeverLogsTheUUID covers the two statuses that
+// keep the concrete path on every other route. On the receiver that
+// path carries the entrypoint UUID, so the line names the route
+// pattern instead.
+func TestAccessLog_ReceiverNeverLogsTheUUID(t *testing.T) {
+	t.Parallel()
+
+	for entrypointUUID, wantStatus := range map[string]int{
+		knownEntrypointUUID:   http.StatusOK,
+		failingEntrypointUUID: http.StatusInternalServerError,
+	} {
+		t.Run(strconv.Itoa(wantStatus), func(t *testing.T) {
+			t.Parallel()
+
+			m, buf := capturingMiddleware(t)
+			router := accessLogRouter(m)
+
+			assert.Equal(
+				t, wantStatus,
+				get(t, router, "/h/"+entrypointUUID+"?src=ci"),
+			)
+
+			entries := accessLogEntries(t, buf)
+			require.Len(t, entries, 1)
+			assert.Equal(t, middleware.ReceiverRoute, entries[0]["url"])
+			assert.NotContains(t, buf.String(), entrypointUUID)
+		})
+	}
+}
+
+// TestAccessLog_ReceiverPreflightNeverLogsTheUUID sends a CORS
+// preflight to the receiver in dev, in the production middleware
+// order. CORS answers it with 200 before chi routes it, so the line
+// names no route and carries no UUID.
+func TestAccessLog_ReceiverPreflightNeverLogsTheUUID(t *testing.T) {
+	t.Parallel()
+
+	m, buf := capturingMiddleware(t)
+
+	router := chi.NewRouter()
+	router.Use(m.Logging())
+	router.Use(m.CORS())
+	router.HandleFunc(
+		middleware.ReceiverRoute,
+		func(w http.ResponseWriter, _ *http.Request) {
+			w.WriteHeader(http.StatusOK)
+		},
+	)
+
+	req := httptest.NewRequestWithContext(
+		context.Background(), http.MethodOptions,
+		"/h/"+knownEntrypointUUID, nil,
+	)
+	req.Header.Set("Origin", "https://sender.example")
+	req.Header.Set("Access-Control-Request-Method", http.MethodPost)
+
+	w := httptest.NewRecorder()
+	router.ServeHTTP(w, req)
+	assert.Equal(t, http.StatusOK, w.Code)
+
+	entries := accessLogEntries(t, buf)
+	require.Len(t, entries, 1)
+	assert.Equal(t, unmatchedRouteLiteral, entries[0]["url"])
+	assert.NotContains(t, buf.String(), knownEntrypointUUID)
+}
+
+// TestAccessLog_RefererNeverLogsTheUUID sends receiver requests whose
+// Referer is the entrypoint URL, as a sender's is after it follows a
+// redirect from http:// to https://. The referer field keeps only the
+// scheme and host, and a Referer without them is logged as a marker.
+func TestAccessLog_RefererNeverLogsTheUUID(t *testing.T) {
+	t.Parallel()
+
+	m, buf := capturingMiddleware(t)
+	router := accessLogRouter(m)
+
+	for _, referer := range []string{
+		"http://webhooker.example/h/" + knownEntrypointUUID + "?src=ci",
+		"/h/" + knownEntrypointUUID,
+	} {
+		assert.Equal(t, http.StatusOK, getWithHeaders(
+			t, router, "/h/"+knownEntrypointUUID,
+			map[string]string{"Referer": referer},
+		))
+	}
+
+	entries := accessLogEntries(t, buf)
+	require.Len(t, entries, 2)
+	assert.Equal(t, "http://webhooker.example", entries[0]["referer"])
+	assert.Equal(t, "(redacted)", entries[1]["referer"])
+	assert.NotContains(t, buf.String(), knownEntrypointUUID)
+}
+
 func TestAccessLog_ServerErrorKeepsConcreteURL(t *testing.T) {
 	t.Parallel()
 

internal/middleware/logbound_test.go

@@ -10,8 +10,10 @@ package middleware_test
 //     the client's choosing.
 //   - CSRF's 403 rejection, at WARN, also registered ahead of
 //     RequireAuth.
-//   - The rate limiters' 429 rejection, at WARN, on the
+//   - The rate limiters' 429 rejection, at WARN. On the receiver
-//     unauthenticated receiver among others.
+//     route it names the route pattern rather than the path, and
+//     the limiters that log a path sit behind RequireAuth, so its cap
+//     is defensive.
 //   - RequireAuth's own unauthenticated-request line, at DEBUG.
 //   - RecordLoginFailure's throttle rejection, at WARN. Its cap is
 //     defensive rather than load-bearing today: chi pins the one
@@ -268,10 +270,14 @@ func logSites() map[string]logSite {
 			send:       postNoToken,
 			wantStatus: http.StatusForbidden,
 		},
-		// The per-entrypoint receiver limiter, unauthenticated. Its
+		// The 429 line every limiter but the aggregate one shares.
-		// bucket is keyed on the path, so the first request through a
+		// Mounted outside any route, as here, it logs the path, as
-		// fresh path is served and only the ones after it are
+		// it does for the limiters behind RequireAuth; on
-		// rejected; sendUntilLimited absorbs that.
+		// ReceiverRoute it names the pattern instead. The
+		// per-entrypoint receiver limiter's bucket is keyed on the
+		// path, so the first request through a fresh path is served
+		// and only the ones after it are rejected; sendUntilLimited
+		// absorbs that.
 		"receiver rate limit 429": {
 			build: func(
 				t *testing.T, m *middleware.Middleware,

internal/middleware/metrics.go

@@ -41,7 +41,8 @@ const unmatchedMethod = unmatchedRoute
 // after /h/ is client-chosen, so labelling by path lets any
 // unauthenticated client mint permanent series at will, and publishes
 // the entrypoint UUID (the receiver's only credential) in the scrape
-// while doing it.
+// while doing it. The access log and the 429 line read it for the
+// same reason: see ReceiverRoute.
 //
 // chi populates the route context during routeHTTP, so this is only
 // valid once routing has run. Every caller below is on the recording

internal/middleware/middleware.go

@@ -13,7 +13,6 @@ import (
 	"time"
 
 	basicauth "github.com/99designs/basicauth-go"
-	"github.com/go-chi/chi"
 	"github.com/go-chi/chi/middleware"
 	"github.com/go-chi/cors"
 	"github.com/prometheus/client_golang/prometheus"
@@ -49,6 +48,10 @@ const (
 	// pick the size of the line it writes.
 	redactedQuery = "?(redacted)"
 
+	// redactedReferer stands in for a Referer the access log cannot
+	// reduce to its scheme and host; see accessLogReferer.
+	redactedReferer = "(redacted)"
+
 	// maxLogRequestIDBytes bounds the request id, which is also
 	// client-supplied: chi's RequestID middleware passes an inbound
 	// X-Request-Id header through verbatim. Its generated form is an
@@ -96,14 +99,14 @@ const (
 	// It is also the ceiling on every OTHER line this service writes
 	// THROUGH SLOG that carries text an UNAUTHENTICATED client
 	// supplies. Those lines — the MaxBodySize rejection, the CSRF
-	// rejection, the rate-limit rejection, the unauthenticated-request
+	// rejection, the unauthenticated-request DEBUG line, the
-	// and unknown-entrypoint DEBUG lines, the failed-login DEBUG
+	// failed-login DEBUG lines, and the two login-throttle WARN
-	// lines, and the two login-throttle WARN lines ("login failure
+	// lines ("login failure limit exceeded" in loginguard.go and
-	// limit exceeded" in loginguard.go and "password verification
+	// "password verification capacity exhausted" in
-	// capacity exhausted" in internal/handlers/auth.go) — spend the
+	// internal/handlers/auth.go) — spend the same per-field budgets,
-	// same per-field budgets, and each carries strictly fewer
+	// and each carries strictly fewer client-supplied fields than the
-	// client-supplied fields than the access log does,
+	// access log does, so none of them can reach a width the access
-	// so none of them can reach a width the access log cannot. That is
+	// log cannot. That is
 	// asserted directly, per line and under both handlers, rather than
 	// left to the reasoning: see logbound_test.go in this package and
 	// in internal/handlers.
@@ -151,6 +154,12 @@ const (
 	MaxAccessLogLineBytes = 2560
 )
 
+// ReceiverRoute is the route pattern the webhook receiver is served
+// at. Its path carries the entrypoint UUID, which is the receiver's
+// only credential, so the log lines this package writes name a
+// request on this route by the pattern, never by its path.
+const ReceiverRoute = "/h/{uuid}"
+
 //nolint:revive // MiddlewareParams is a standard fx naming convention.
 type MiddlewareParams struct {
 	fx.In
@@ -268,14 +277,14 @@ func (lrw *loggingResponseWriter) Unwrap() http.ResponseWriter {
 // marker.
 //
 // The path on those branches is bounded by the service's routes or by
-// the operator's data — a 2xx on the receiver means the UUID named a
+// the operator's data — a 2xx under /s means the file is in the
-// stored entrypoint, a 2xx under /s means the file is in the embedded
+// embedded tree. The query is not bounded by anything:
-// tree. The query is not bounded by anything: /.well-known/healthcheck
+// /.well-known/healthcheck and /s/* take no authentication and sit
-// and /s/* take no authentication and sit behind no rate limiter, and
+// behind no rate limiter, and /pages/login behind only the login
-// /pages/login behind only the login limiter, so any of them will
+// limiter, so any of them will answer 200 to a URL carrying an
-// answer 200 to a URL carrying an arbitrary number of arbitrary bytes
+// arbitrary number of arbitrary bytes after the '?'. Keeping the path
-// after the '?'. Keeping the path and dropping the query is what makes
+// and dropping the query is what makes this branch as bounded as the
-// this branch as bounded as the pattern branches below.
+// pattern branches below.
 //
 // Nothing debuggable is lost. The only query parameters the service
 // reads are the login page's `next`, the page to return to,
@@ -297,36 +306,61 @@ func concreteLogURL(r *http.Request) string {
 
 // accessLogURL returns the value for the access log's url field.
 //
-// 2xx and 5xx responses get the concrete path (see concreteLogURL). A
+// On ReceiverRoute it is the route pattern whatever the status, so
-// success resolved against a static route or against the operator's
+// the entrypoint UUID in the path is never logged. A request chi never
-// own data — on the receiver, a 2xx means the UUID named a stored
+// routed gets unmatchedRoute whatever the status, for the same reason:
-// entrypoint — and a server error is our own bug, where the exact URL
+// in dev, CORS answers a preflight to the receiver before routing.
-// is the primary evidence and which no client can provoke at will.
+//
+// On every other route, 2xx and 5xx responses get the concrete path
+// (see concreteLogURL). A success resolved against a static route or
+// against the operator's own data, and a server error is our own bug,
+// where the exact URL is the primary evidence and which no client can
+// provoke at will.
 //
 // 3xx and 4xx responses get the chi route pattern instead. Those are
-// the outcomes an unauthenticated client drives for free: 404 or 429
+// the outcomes an unauthenticated client drives for free, such as a
-// on any invented /h/ path, 303 to the login page on any
+// 303 to the login page on any invented /user/ path. Logging the
-// invented /user/ path. Logging the concrete URL there lets a flood
+// concrete URL there lets a flood write attacker-chosen text, of
-// write attacker-chosen text, of attacker-chosen length, into the
+// attacker-chosen length, into the operator's log at one line per
-// operator's log at one line per request. The pattern comes from the
+// request. The pattern comes from the router's own table, so it is
-// router's own table, so it is bounded by the service's routes while
+// bounded by the service's routes while still naming which class of
-// still naming which class of request was rejected.
+// request was rejected.
 //
 // The pattern is only populated once routing has run, so this must be
 // called after the handler returns, not before.
 func accessLogURL(r *http.Request, status int) string {
+	pattern := routePatternID(r.Context())
+	if pattern == ReceiverRoute || pattern == unmatchedRoute {
+		return pattern
+	}
+
 	if status < http.StatusMultipleChoices ||
 		status >= http.StatusInternalServerError {
 		return concreteLogURL(r)
 	}
 
-	if rc := chi.RouteContext(r.Context()); rc != nil {
+	return pattern
-		if pattern := rc.RoutePattern(); pattern != "" {
+}
-			return pattern
+
-		}
+// accessLogReferer returns the value for the access log's referer
+// field: the Referer's scheme and host, never its path or query. The
+// path can be an entrypoint URL, which carries the UUID: a sender that
+// follows a redirect from http:// to https:// sends the URL it was
+// redirected from, and a browser sends the 404 page's URL under /h/
+// when it fetches that page's stylesheet. A Referer without both a
+// scheme and a host is logged as redactedReferer.
+func accessLogReferer(r *http.Request) string {
+	referer := r.Referer()
+	if referer == "" {
+		return ""
 	}
 
-	return unmatchedRoute
+	parsed, err := url.Parse(referer)
+	if err != nil || parsed.Scheme == "" || parsed.Host == "" {
+		return redactedReferer
+	}
+
+	return parsed.Scheme + "://" + parsed.Host
 }
 
 // Logging returns middleware that logs each HTTP request with
@@ -379,7 +413,7 @@ func (s *Middleware) Logging() func(http.Handler) http.Handler {
 						requestID, maxLogRequestIDBytes,
 					),
 					"referer", logfield.Truncate(
-						r.Referer(), logfield.MaxBytes,
+						accessLogReferer(r), logfield.MaxBytes,
 					),
 					"proto", r.Proto,
 					"remoteIP", RemoteIP(r),
@@ -538,7 +572,9 @@ func (s *Middleware) MetricsAuth() func(http.Handler) http.Handler {
 // SecurityHeaders returns middleware that sets production security
 // headers on every response: HSTS (unless HSTS_MAX_AGE is zero),
 // X-Content-Type-Options, X-Frame-Options, CSP, Referrer-Policy, and
-// Permissions-Policy.
+// Permissions-Policy. The CSP takes scripts and styles only from
+// files on this origin, so a template may hold no inline script,
+// event handler or style attribute.
 func (s *Middleware) SecurityHeaders() func(http.Handler) http.Handler {
 	maxAge := s.params.Config.HSTSMaxAge
 	hsts := "max-age=" + strconv.FormatInt(int64(maxAge/time.Second), 10) +
@@ -560,8 +596,8 @@ func (s *Middleware) SecurityHeaders() func(http.Handler) http.Handler {
 			w.Header().Set(
 				"Content-Security-Policy",
 				"default-src 'self'; "+
-					"script-src 'self' 'unsafe-inline'; "+
+					"script-src 'self'; "+
-					"style-src 'self' 'unsafe-inline'",
+					"style-src 'self'",
 			)
 			w.Header().Set(
 				"Referrer-Policy",

internal/middleware/middleware_test.go

@@ -747,10 +747,10 @@ func TestNoCache_SetsHeaders(t *testing.T) {
 
 // --- SecurityHeaders Middleware Tests ---
 
-// hstsHeaders serves one request through SecurityHeaders built over
+// securityHeaders serves one request through SecurityHeaders built
-// a Config with the given HSTSMaxAge, and returns every
+// over a Config with the given HSTSMaxAge, and returns the response's
-// Strict-Transport-Security value on the response.
+// headers.
-func hstsHeaders(t *testing.T, maxAge time.Duration) []string {
+func securityHeaders(t *testing.T, maxAge time.Duration) http.Header {
 	t.Helper()
 
 	cfg := &config.Config{HSTSMaxAge: maxAge}
@@ -772,7 +772,7 @@ func hstsHeaders(t *testing.T, maxAge time.Duration) []string {
 
 	handler.ServeHTTP(w, req)
 
-	return w.Header().Values("Strict-Transport-Security")
+	return w.Header()
 }
 
 func TestSecurityHeaders_HSTS(t *testing.T) {
@@ -804,11 +804,28 @@ func TestSecurityHeaders_HSTS(t *testing.T) {
 		t.Run(tt.name, func(t *testing.T) {
 			t.Parallel()
 
-			assert.Equal(t, tt.expected, hstsHeaders(t, tt.maxAge))
+			assert.Equal(
+				t, tt.expected,
+				securityHeaders(t, tt.maxAge).Values(
+					"Strict-Transport-Security",
+				),
+			)
 		})
 	}
 }
 
+// TestSecurityHeaders_CSP pins the Content-Security-Policy: a page
+// takes its scripts and styles only from files on its own origin,
+// never from an inline script, event handler or style.
+func TestSecurityHeaders_CSP(t *testing.T) {
+	t.Parallel()
+
+	assert.Equal(
+		t, "default-src 'self'; script-src 'self'; style-src 'self'",
+		securityHeaders(t, 0).Get("Content-Security-Policy"),
+	)
+}
+
 // --- MaxBodySize Middleware Tests ---
 
 const testBodyLimit int64 = 64

internal/middleware/ratelimit.go

@@ -259,26 +259,28 @@ func (m *Middleware) clientAddr(r *http.Request) (netip.Addr, bool) {
 	return peer, true
 }
 
-// tooManyRequests returns the 429 handler used by the
+// tooManyRequests returns the 429 handler used by the postRateLimit
-// password-change and per-entrypoint receiver limiters: it logs the
+// limiters and the per-entrypoint receiver limiter: it logs the
 // rejection with logMessage and answers with responseMessage.
 // httprate adds the Retry-After header (RFC 6585). The aggregate
 // receiver limiter uses floodTooManyRequests instead.
 //
-// The path is capped against the same budget as the access log's url
+// On ReceiverRoute the line names the route pattern, so the
-// field. The per-entrypoint receiver limiter is unauthenticated and
+// entrypoint UUID in the path is never logged. Elsewhere it names the
-// its path is a client-chosen segment of client-chosen length, so at
+// path, capped against the same budget as the access log's url field;
-// WARN an uncapped path would let a sender pick the size of the line
+// the limiters there sit behind RequireAuth, so the cap is defensive.
-// it writes — the same defect the access log capping closed.
 func (m *Middleware) tooManyRequests(
 	logMessage, responseMessage string,
 ) http.HandlerFunc {
 	return func(w http.ResponseWriter, r *http.Request) {
+		path := r.URL.Path
+		if routePatternID(r.Context()) == ReceiverRoute {
+			path = ReceiverRoute
+		}
+
 		m.log.Warn(
 			logMessage,
-			"path", logfield.Truncate(
+			"path", logfield.Truncate(path, logfield.MaxBytes),
-				r.URL.Path, logfield.MaxBytes,
-			),
 			"remoteIP", RemoteIP(r),
 			"clientIP", ClientIP(r),
 		)

internal/server/alpine_browser_test.go

@@ -60,8 +60,8 @@ const (
 // TestAlpineRunsUnderTheSecurityPolicy loads the webhook page and the
 // event log in a headless browser, served by the real router and so
 // under the real Content-Security-Policy, and checks that the pages'
-// Alpine.js directives and the copy control work, and that a target's
+// Alpine.js directives, the copy control and the delete prompts work,
-// row shows its delivery counts.
+// and that a target's row shows its delivery counts.
 func TestAlpineRunsUnderTheSecurityPolicy(t *testing.T) {
 	t.Parallel()
 
@@ -74,6 +74,10 @@ func TestAlpineRunsUnderTheSecurityPolicy(t *testing.T) {
 	userID, _ := env.seedUser(t, "browser", "browser-password")
 	webhook, older, event, target := seedBrowserWebhook(t, env, userID)
 
+	// The sign-in page is loaded before the browser signs in, since a
+	// signed-in visit to it is sent on to another page.
+	checkOtherPages(ctx, t, srv.URL+"/pages/login")
+
 	require.NoError(t, chromedp.Run(
 		ctx, setCookies(srv.URL, env.authCookies(t, userID, "browser")),
 	))
@@ -99,6 +103,11 @@ func TestAlpineRunsUnderTheSecurityPolicy(t *testing.T) {
 	checkEventLog(ctx, t, page+"/events", event.ID, older.ID, target.Name)
 	checkMobileMenu(ctx, t, page)
 	checkPhoneWidth(ctx, t, page, page+"/events", target.Name)
+	checkOtherPages(ctx, t,
+		srv.URL+"/hooks", srv.URL+"/settings", srv.URL+"/user/browser",
+		// A path that does not exist, which gets the 404 error page.
+		srv.URL+"/no-such-page")
+	checkDeletePrompt(ctx, t, page, target)
 
 	assert.Empty(t, problems(), "the browser reported problems")
 }
@@ -1388,3 +1397,75 @@ func checkPhoneWidth(
 	assert.False(t, sideways, "the event log scrolls sideways on a phone")
 	assert.Empty(t, cutOff, "the event log cuts these off on a phone")
 }
+
+// checkOtherPages loads the pages at urls, which no other check opens,
+// so that anything the browser reports going wrong on them fails the
+// test too.
+func checkOtherPages(ctx context.Context, t *testing.T, urls ...string) {
+	t.Helper()
+
+	for _, url := range urls {
+		require.NoError(t, chromedp.Run(ctx, loadPage(url)))
+	}
+}
+
+// answerPrompt returns a script that stands in for the user at the
+// browser's prompt: it makes window.confirm record the question in
+// window.asked and give answer. It also records in window.sent whether
+// a form is sent once its submit event is handled, from a listener on
+// window, which the event reaches after the listener static/js/app.js
+// adds to document.
+func answerPrompt(answer bool) string {
+	return fmt.Sprintf(`
+		window.asked = "";
+		window.sent = false;
+		window.confirm = (question) => {
+			window.asked = question;
+			return %t;
+		};
+		window.addEventListener("submit", (event) => {
+			window.sent = !event.defaultPrevented;
+		});`, answer)
+}
+
+// checkDeletePrompt loads a webhook page and clicks Delete beside
+// target. It checks that the browser asks first, naming the target,
+// that answering no leaves the form unsent, and that answering yes
+// sends it and the target is deleted.
+func checkDeletePrompt(
+	ctx context.Context, t *testing.T, url string, target *database.Target,
+) {
+	t.Helper()
+
+	form := `form[action$="/targets/` + target.ID + `/delete"]`
+	question := `Delete target "` + target.Name + `"?` + "\n\n" +
+		"Nothing more is delivered to it. " +
+		"Its past deliveries stay in the event log."
+
+	var (
+		asked  string
+		sent   bool
+		listed bool
+	)
+
+	require.NoError(t, chromedp.Run(
+		ctx, loadPage(url), chromedp.Evaluate(answerPrompt(false), nil),
+	))
+	click(ctx, t, form+" button")
+	require.NoError(t, chromedp.Run(
+		ctx,
+		chromedp.Evaluate(`window.asked`, &asked),
+		chromedp.Evaluate(`window.sent`, &sent),
+	))
+	assert.Equal(t, question, asked, "Delete does not ask about the target")
+	assert.False(t, sent, "answering no still sends the delete form")
+
+	require.NoError(t, chromedp.Run(
+		ctx, chromedp.Evaluate(answerPrompt(true), nil),
+	))
+	clickAndLoad(ctx, t, form+" button")
+	require.NoError(t, chromedp.Run(ctx, chromedp.Evaluate(
+		`document.querySelector('`+form+`') !== null`, &listed,
+	)))
+	assert.False(t, listed, "answering yes does not delete the target")
+}

internal/server/bind_address_test.go

@@ -5,6 +5,7 @@ import (
 	"net"
 	"net/http"
 	"strconv"
+	"syscall"
 	"testing"
 	"time"
 
@@ -25,6 +26,11 @@ const (
 	// its network namespace even with a published port.
 	wildcardV4 = "0.0.0.0"
 
+	// loopbackV6 and wildcardV6 are the IPv6 counterparts of the two
+	// above.
+	loopbackV6 = "::1"
+	wildcardV6 = "::"
+
 	// unavailableAddr is a TEST-NET-1 address (RFC 5737). It is a
 	// well-formed literal that no host is assigned, so binding it
 	// fails with EADDRNOTAVAIL rather than succeeding somewhere
@@ -116,6 +122,24 @@ func otherLocalAddr(t *testing.T) string {
 	return ""
 }
 
+// skipWithoutIPv6 skips the test when this host cannot bind
+// loopbackV6. Without IPv6 there is no second family for a bind to
+// claim or leave alone.
+func skipWithoutIPv6(t *testing.T) {
+	t.Helper()
+
+	var listenCfg net.ListenConfig
+
+	l, err := listenCfg.Listen(
+		t.Context(), "tcp6", net.JoinHostPort(loopbackV6, "0"),
+	)
+	if err != nil {
+		t.Skip("host has no IPv6 loopback to bind")
+	}
+
+	require.NoError(t, l.Close())
+}
+
 // startBoundServer starts the wired app with the given bind address
 // on a free port and returns that port. The app is stopped on
 // cleanup.
@@ -292,6 +316,59 @@ func TestBindAddress_WildcardReachesOtherAddresses(t *testing.T) {
 	requireDialable(t, net.JoinHostPort(other, strconv.Itoa(port)))
 }
 
+// TestBindAddress_WildcardV4IsIPv4Only pins that 0.0.0.0 answers on
+// IPv4 alone, as the README says, whether it is written as 0.0.0.0 or
+// as the IPv4-mapped ::ffff:0.0.0.0. Go listens on 0.0.0.0 with a
+// socket that answers IPv6 too unless told otherwise, which put the
+// admin UI and the receiver on every IPv6 address of a host whose
+// operator had chosen an IPv4 bind.
+func TestBindAddress_WildcardV4IsIPv4Only(t *testing.T) {
+	t.Parallel()
+
+	skipWithoutIPv6(t)
+
+	for _, bindAddress := range []string{wildcardV4, "::ffff:0.0.0.0"} {
+		t.Run(bindAddress, func(t *testing.T) {
+			t.Parallel()
+
+			port := startBoundServer(t, bindAddress)
+
+			// Positive control: the service really is up, on IPv4.
+			requireDialable(
+				t, net.JoinHostPort(loopbackV4, strconv.Itoa(port)),
+			)
+
+			dialer := net.Dialer{Timeout: dialTimeout}
+
+			conn, err := dialer.DialContext(
+				t.Context(), "tcp",
+				net.JoinHostPort(loopbackV6, strconv.Itoa(port)),
+			)
+			if err == nil {
+				_ = conn.Close()
+			}
+
+			require.ErrorIs(
+				t, err, syscall.ECONNREFUSED,
+				"a %s bind answered on %s", bindAddress, loopbackV6,
+			)
+		})
+	}
+}
+
+// TestBindAddress_WildcardV6ReachesBothFamilies is the counterpart:
+// :: is the value that binds IPv6, and it answers IPv4 as well.
+func TestBindAddress_WildcardV6ReachesBothFamilies(t *testing.T) {
+	t.Parallel()
+
+	skipWithoutIPv6(t)
+
+	port := startBoundServer(t, wildcardV6)
+
+	requireDialable(t, net.JoinHostPort(loopbackV6, strconv.Itoa(port)))
+	requireDialable(t, net.JoinHostPort(loopbackV4, strconv.Itoa(port)))
+}
+
 // TestBindAddress_ServesRequestsOnConfiguredAddress proves the bound
 // listener serves the application rather than merely accepting TCP,
 // so a bind address that is honoured cannot be mistaken for one that

internal/server/export_test.go

@@ -91,6 +91,24 @@ func NewRouterForTest(
 	return s.router
 }
 
+// NewRouterWithSentryForTest is NewRouterForTest with the sentryhttp
+// handler registered, as a set SENTRY_DSN does in production.
+func NewRouterWithSentryForTest(
+	t *testing.T,
+	log *logger.Logger,
+	cfg *config.Config,
+	mw *middleware.Middleware,
+	h *handlers.Handlers,
+) http.Handler {
+	t.Helper()
+
+	s := newServerForTest(t, log, cfg, mw, h)
+	s.sentryEnabled.Store(true)
+	s.SetupRoutes()
+
+	return s.router
+}
+
 // ListenAddrForTest exposes the address the HTTP listener binds for
 // a given Config, so the rendering of host and port — IPv6
 // bracketing above all — can be pinned without standing up a

internal/server/http.go

@@ -1,9 +1,11 @@
 package server
 
 import (
+	"context"
 	"errors"
 	"net"
 	"net/http"
+	"net/netip"
 	"strconv"
 	"time"
 )
@@ -39,6 +41,23 @@ func (s *Server) listenAddr() string {
 	)
 }
 
+// listenNetwork returns the network the HTTP listener binds on, "tcp4"
+// for an IPv4 BindAddress and "tcp" for any other.
+//
+// On "tcp", Go binds 0.0.0.0 with a socket that answers IPv6 as well,
+// exactly as it binds ::. "tcp4" keeps 0.0.0.0 to IPv4, which is what
+// the README says it means. :: stays on "tcp" and answers both. An
+// IPv4-mapped address such as ::ffff:0.0.0.0 is the IPv4 address it
+// carries to Go's listener, so it gets "tcp4" too.
+func (s *Server) listenNetwork() string {
+	addr, err := netip.ParseAddr(s.params.Config.BindAddress)
+	if err == nil && addr.Unmap().Is4() {
+		return "tcp4"
+	}
+
+	return "tcp"
+}
+
 // newHTTPServer builds the HTTP server for this Server's
 // configuration.
 //
@@ -67,7 +86,17 @@ func (s *Server) serveUntilShutdown() {
 
 	s.log.Info("http begin listen", "listenaddr", s.httpServer.Addr)
 
-	err := s.httpServer.ListenAndServe()
+	// ListenAndServe always listens on "tcp", which would undo
+	// listenNetwork.
+	var listenCfg net.ListenConfig
+
+	listener, err := listenCfg.Listen(
+		context.Background(), s.listenNetwork(), s.httpServer.Addr,
+	)
+	if err == nil {
+		err = s.httpServer.Serve(listener)
+	}
+
 	if err != nil && !errors.Is(err, http.ErrServerClosed) {
 		s.log.Error("listen error", "error", err)
 		s.shutdownWithFailure()

internal/server/routes.go

@@ -6,7 +6,8 @@ import (
 
 	sentryhttp "github.com/getsentry/sentry-go/http"
 	"github.com/go-chi/chi"
-	"github.com/go-chi/chi/middleware"
+	chimw "github.com/go-chi/chi/middleware"
+	"sneak.berlin/go/webhooker/internal/middleware"
 	"sneak.berlin/go/webhooker/static"
 )
 
@@ -60,7 +61,7 @@ func (s *Server) SetupRoutes() {
 }
 
 func (s *Server) setupGlobalMiddleware() {
-	s.router.Use(middleware.RequestID)
+	s.router.Use(chimw.RequestID)
 	s.router.Use(s.mw.SecurityHeaders())
 	s.router.Use(s.mw.Logging())
 
@@ -90,8 +91,8 @@ func (s *Server) setupGlobalMiddleware() {
 
 // recoverPanics installs on r the recoverer, answering a panic with
 // page (a plain 500 when page is nil), and inside it the Sentry error
-// reporting (if SENTRY_DSN is set). Repanic is true so panics still
+// reporting and transaction naming (if SENTRY_DSN is set). Repanic is
-// bubble up to the recoverer.
+// true so panics still bubble up to the recoverer.
 //
 // Each admin page route group installs its own, with the error page,
 // as its first middleware. A panic there is logged, reported and
@@ -105,6 +106,11 @@ func (s *Server) recoverPanics(r chi.Router, page http.Handler) {
 			Repanic: true,
 		})
 		r.Use(sentryHandler.Handle)
+		// Installed with every sentryhttp handler rather than once
+		// globally: in an admin page route group, the group's own
+		// handler finishes the transaction as the group returns, so
+		// a rename further out would come too late.
+		r.Use(nameSentryTransaction)
 	}
 }
 
@@ -340,7 +346,7 @@ func (s *Server) setupWebhookRoutes() {
 	// unauthenticated endpoint's body; TestReceiver_OversizeBodyRefused
 	// pins it.
 	s.router.With(s.mw.ReceiverRateLimit()).HandleFunc(
-		"/h/{uuid}",
+		middleware.ReceiverRoute,
 		s.h.HandleWebhook(),
 	)
 }

internal/server/sentry.go

@@ -1,6 +1,7 @@
 package server
 
 import (
+	"context"
 	"net/http"
 	"net/url"
 	"strings"
@@ -71,10 +72,15 @@ func scrubSentryRequest(
 
 	pattern := sentryRoutePattern(hint)
 
-	// Only transaction events carry a Transaction name, and the SDK
+	// Only transaction events carry a Transaction name. sentryhttp
 	// builds it from the concrete path too (sentryhttp.go:105 via
-	// tracing.go:553). Rewritten on the same terms.
+	// tracing.go:553) and marks it SourceURL, so it is rewritten on
-	if event.Transaction != "" {
+	// the same terms. Only a name nameSentryTransaction set from the
+	// route pattern, which it marks SourceRoute, is kept.
+	namedByRoute := event.TransactionInfo != nil &&
+		event.TransactionInfo.Source == sentry.SourceRoute
+
+	if event.Transaction != "" && !namedByRoute {
 		event.Transaction = sentryTransactionName(
 			event.Transaction, pattern,
 		)
@@ -120,7 +126,8 @@ func scrubSentryRequest(
 // The transaction dispatch has no such request: Span.doFinish calls
 // hub.CaptureEvent (tracing.go:356), which passes a nil hint that the
 // client replaces with an empty one (client.go:620-622). The pattern
-// is therefore always "" there, and the callers fall back.
+// is therefore always "" there, and the callers fall back. A routed
+// transaction is named by nameSentryTransaction instead.
 func sentryRoutePattern(hint *sentry.EventHint) string {
 	if hint == nil || hint.Context == nil {
 		return ""
@@ -142,6 +149,52 @@ func sentryRoutePattern(hint *sentry.EventHint) string {
 	return rctx.RoutePattern()
 }
 
+// nameSentryTransaction renames the request's Sentry transaction to
+// "METHOD <route pattern>", such as "POST /h/{uuid}". sentryhttp names
+// it after the concrete path, which scrubSentryRequest withholds;
+// since Sentry groups transactions by name, every transaction in the
+// service would otherwise land in one group per method, such as
+// "POST /(redacted)".
+//
+// It has to run inside the sentryhttp handler, where the transaction
+// is on the request context. It renames on the way out, once chi has
+// routed the request, and does so in a defer so that a request whose
+// handler panics is named too. The context is taken when the defer is
+// set up, before routing, but the chi routing context on it is a
+// pointer that chi fills in as it routes.
+func nameSentryTransaction(next http.Handler) http.Handler {
+	return http.HandlerFunc(
+		func(w http.ResponseWriter, r *http.Request) {
+			defer setSentryTransactionName(r.Context(), r.Method)
+
+			next.ServeHTTP(w, r)
+		},
+	)
+}
+
+// setSentryTransactionName does the rename for nameSentryTransaction.
+// A path that matches nothing at the top level has no pattern; its
+// name is left as the SDK set it, for scrubSentryRequest to withhold.
+// Inside a route group, a path that matches no route in the group, or
+// a request the group's own middleware turns away before routing, has
+// the group's pattern, such as "/pages/*", and is named after it.
+func setSentryTransactionName(ctx context.Context, method string) {
+	transaction := sentry.TransactionFromContext(ctx)
+	rctx := chi.RouteContext(ctx)
+
+	if transaction == nil || rctx == nil {
+		return
+	}
+
+	pattern := rctx.RoutePattern()
+	if pattern == "" {
+		return
+	}
+
+	transaction.Name = method + " " + pattern
+	transaction.Source = sentry.SourceRoute
+}
+
 // sentryRouteURL rebuilds an event's request URL with the route
 // pattern in place of the concrete path.
 //

internal/server/sentry_test.go

@@ -16,6 +16,7 @@ import (
 	"github.com/go-chi/chi"
 	"github.com/stretchr/testify/assert"
 	"github.com/stretchr/testify/require"
+	"sneak.berlin/go/webhooker/internal/config"
 	"sneak.berlin/go/webhooker/internal/server"
 )
 
@@ -130,11 +131,15 @@ func (c sentryCase) capture(t *testing.T) []*sentry.Event {
 // router mirrors the one ordering these tests depend on, over the two
 // route patterns they need: a recovering middleware outside, then the
 // sentryhttp handler registered with Use and Repanic set, exactly as
-// routes.go orders the two. The bare recover stands in for
+// routes.go orders the two. It leaves out nameSentryTransaction,
-// Middleware.Recoverer, which holds that outer slot in production; it
+// which routes.go installs inside the sentryhttp handler, so a
-// is here only to keep panic stacks out of the test output. That the
+// transaction captured here keeps the SDK's name and shows what the
-// production one really does catch what sentryhttp re-raises is
+// hook does with a transaction that missed the rename. The bare
-// pinned separately, by TestSentryStillSeesAPanic.
+// recover stands in for Middleware.Recoverer, which holds that outer
+// slot in production; it is here only to keep panic stacks out of the
+// test output. That the production one really does catch what
+// sentryhttp re-raises is pinned separately, by
+// TestSentryStillSeesAPanic.
 func (c sentryCase) router() http.Handler {
 	handler := func(_ http.ResponseWriter, r *http.Request) {
 		// This call is what drains the body tee and fills the
@@ -353,8 +358,9 @@ func TestSentryScrub_KeepsTheRoutingContext(t *testing.T) {
 // Span.doFinish captures with a nil hint, so BeforeSendTransaction
 // gets one with no context and no request: the route pattern is out of
 // reach and both the URL and the SDK-built transaction name have to
-// fall back. Tracing is off in this service, so no transaction event
+// fall back. This router has no nameSentryTransaction, so this is the
-// is produced today; the hook is a floor against that changing.
+// floor under a transaction that missed the rename;
+// TestSentryTransaction_NamedByRoute covers the routes as they ship.
 func TestSentryScrub_RedactsTheTransactionDispatch(t *testing.T) {
 	t.Parallel()
 
@@ -376,6 +382,83 @@ func TestSentryScrub_RedactsTheTransactionDispatch(t *testing.T) {
 	assert.Equal(t, "POST /(redacted)", event.Transaction)
 }
 
+// TestSentryTransaction_NamedByRoute serves requests through the real
+// route tree with tracing on, which the service leaves off. A routed
+// request's transaction is named after its route pattern, so Sentry
+// can group transactions by route, and the entrypoint UUID in the
+// receiver path appears nowhere in the event. A path that matches
+// nothing at the top level has no pattern, and its name is still
+// withheld; one that matches no route inside a route group is named
+// after the group's pattern.
+func TestSentryTransaction_NamedByRoute(t *testing.T) {
+	t.Parallel()
+
+	// newTestEnv leaves the receiver rate limit at zero, which
+	// refuses every request before it reaches the receiver.
+	env := newTestEnvWithConfig(t, &config.Config{
+		DataDir:           t.TempDir(),
+		Environment:       config.EnvironmentDev,
+		ReceiverRateLimit: 10,
+	})
+
+	router := server.NewRouterWithSentryForTest(
+		t, env.log, env.cfg, env.mw, env.hnd,
+	)
+
+	for _, tc := range []struct {
+		method, path, want string
+	}{
+		{
+			http.MethodPost,
+			"/h/" + sentryReceiverUUID,
+			"POST /h/{uuid}",
+		},
+		// An admin page route group, which installs its own
+		// sentryhttp handler.
+		{http.MethodGet, "/pages/login", "GET /pages/login"},
+		{http.MethodGet, "/pages/unrouted", "GET /pages/*"},
+		{
+			http.MethodPost,
+			"/h/" + sentryReceiverUUID + "/unrouted",
+			"POST /(redacted)",
+		},
+	} {
+		t.Run(tc.want, func(t *testing.T) {
+			t.Parallel()
+
+			transport := &captureTransport{}
+
+			opts := server.SentryClientOptionsForTest(
+				"https://public@sentry.invalid/1",
+				"webhooker-test",
+			)
+			opts.Transport = transport
+			opts.EnableTracing = true
+			opts.TracesSampleRate = 1.0
+
+			client, err := sentry.NewClient(opts)
+			require.NoError(t, err)
+
+			req := httptest.NewRequestWithContext(
+				sentry.SetHubOnContext(
+					context.Background(),
+					sentry.NewHub(client, sentry.NewScope()),
+				),
+				tc.method, tc.path, nil,
+			)
+			router.ServeHTTP(httptest.NewRecorder(), req)
+
+			event := onlyEvent(t, transport.events)
+			require.Equal(t, "transaction", event.Type)
+
+			assert.Equal(t, tc.want, event.Transaction)
+			assert.NotContains(
+				t, marshalEvent(t, event), sentryReceiverUUID,
+			)
+		})
+	}
+}
+
 // TestSentryScrub_TransactionDispatchIsUnscrubbedWithoutTheHook pins
 // that dispatch's premise the same way, since it is the one the
 // service does not exercise today.

internal/server/server.go

@@ -135,6 +135,8 @@ func New(lc fx.Lifecycle, params ServerParams) (*Server, error) {
 	s.httpServer = s.newHTTPServer()
 
 	lc.Append(fx.Hook{
+		//nolint:contextcheck // The listener outlives this hook, whose
+		// context fx cancels once the app has started.
 		OnStart: func(_ context.Context) error {
 			s.startupTime = time.Now()
 			go s.Run()

internal/session/session.go

@@ -441,10 +441,9 @@ func (s *Session) Regenerate(
 //
 // Whichever comes first ends the session.
 //
-// A session that claims to be authenticated but carries no
+// A session that claims to be authenticated but lacks the timestamp
-// timestamps predates this check; it is treated as expired so the
+// a deadline is measured from is treated as expired, so the user
-// user re-authenticates rather than being granted an unbounded
+// re-authenticates rather than being granted an unbounded session.
-// session.
 func (s *Session) expired(sess *sessions.Session) bool {
 	now := s.now()
 

internal/versionscript/version_script_test.go

@@ -153,17 +153,15 @@ func TestVersion_IsStableAcrossInvocations(t *testing.T) {
 	require.Equal(t, first, second)
 }
 
-// The build target is the only place composing linker flags. If a
+// The build target is the only place that stamps the version. If a
-// future edit drops either half, the binary silently reports "dev"
+// future edit drops the -X flag, the binary silently reports "dev"
-// again (the defect this package exists for) or fails to link on
+// again, the defect this package exists for.
-// Alpine.
+func TestMakefile_BuildStampsTheVersion(t *testing.T) {
-func TestMakefile_BuildComposesVersionAndExtraFlags(t *testing.T) {
 	t.Parallel()
 
 	makefile := read(t, makefilePath)
 
 	require.Contains(t, makefile, "-X main.version=$(VERSION)")
-	require.Contains(t, makefile, "$(GO_LDFLAGS)")
 	require.Contains(t, makefile, "VERSION ?= $(shell script/version)")
 }
 
@@ -222,8 +220,8 @@ func requireMake(t *testing.T) {
 	}
 }
 
-// Every compile in the image goes through the build target, so the
+// The image compiles through the build target, so it carries the
-// static relink cannot replace the flags that carry the stamp.
+// target's version stamp.
 func TestDockerfile_BuildsThroughTheMakeTarget(t *testing.T) {
 	t.Parallel()
 
@@ -232,8 +230,7 @@ func TestDockerfile_BuildsThroughTheMakeTarget(t *testing.T) {
 	require.NotContains(t, dockerfile, "go build",
 		"a raw go build bypasses the Makefile's -X flag")
 	require.Contains(t, dockerfile, "ARG VERSION")
-	require.Contains(t, dockerfile,
+	require.Contains(t, dockerfile, `RUN make build VERSION="$VERSION"`)
-		`make build VERSION="$VERSION" GO_LDFLAGS='-extldflags "-static"'`)
 }
 
 func read(t *testing.T, path string) string {

static/css/style.css

@@ -1,7 +1,7 @@
 /*
- * The two shared styles for the controls a user clicks. Every page loads
+ * The two shared styles for the controls a user clicks, and the rule that
- * this file after tailwind.css. It is plain CSS: make css does not build
+ * hides an element marked x-cloak. Every page loads this file after
- * it.
+ * tailwind.css. It is plain CSS: make css does not build it.
  *
  * A button is btn-primary, btn-secondary or btn-danger, from input.css.
  * A secondary or inline action, such as Copy beside an entrypoint URL or
@@ -46,3 +46,11 @@
         outline-offset: 2px;
     }
 }
+
+/*
+ * Alpine.js removes x-cloak from an element once it has started, so
+ * something Alpine hides is not seen on the page while it loads.
+ */
+[x-cloak] {
+    display: none !important;
+}

static/js/app.js

@@ -58,6 +58,18 @@
     }
 })();
 
+// Asks before a form with a data-confirm attribute is sent, such as a
+// delete form on the webhook page. The browser's own prompt shows the
+// attribute's text, and the form is sent only if the user accepts.
+document.addEventListener("submit", function (event) {
+    "use strict";
+
+    const message = event.target.getAttribute("data-confirm");
+    if (message !== null && !window.confirm(message)) {
+        event.preventDefault();
+    }
+});
+
 // Alpine.js components.
 //
 // The pages' Content-Security-Policy forbids eval, so the UI loads

templates/event_body.html

@@ -5,7 +5,7 @@
 {{if .Binary}}
 <p class="text-xs text-gray-500">This body is binary ({{.Size}} bytes) and is not shown. <a href="{{.EventURL}}/body" class="btn-small">Download the body</a></p>
 {{else if .Text}}
-<pre class="rounded-md border border-gray-200 bg-white p-2 text-xs text-gray-700 overflow-x-auto whitespace-pre-wrap break-all"{{if .Scroll}} style="max-height: 32rem; overflow-y: auto"{{end}}>{{.Text}}</pre>
+<pre class="rounded-md border border-gray-200 bg-white p-2 text-xs text-gray-700 overflow-x-auto whitespace-pre-wrap break-all{{if .Scroll}} max-h-[32rem] overflow-y-auto{{end}}">{{.Text}}</pre>
 {{if .Cut}}
 <p class="mt-2 text-xs text-gray-500">Showing the first {{.ShownBytes}} of {{.Size}} bytes, unformatted. <a href="{{.EventURL}}" class="btn-small">Show the whole body</a> <a href="{{.EventURL}}/body" class="btn-small">Download the body</a></p>
 {{end}}

templates/htmlheader.html

@@ -4,6 +4,5 @@
 <title>{{block "title" .}}Webhooker{{end}}</title>
 <link rel="stylesheet" href="/s/css/tailwind.css">
 <link rel="stylesheet" href="/s/css/style.css">
-<style>[x-cloak] { display: none !important; }</style>
 {{block "head" .}}{{end}}
 {{end}}

templates/settings.html

@@ -13,13 +13,11 @@
             <div class="p-4">
                 <!-- A value too wide to sit beside its name moves to the
                      next line, where a list breaks only at the spaces
-                     between its entries. overflow-wrap: anywhere breaks
+                     between its entries. wrap-anywhere breaks inside a
-                     inside a value only when it alone is wider than the
+                     value only when it alone is wider than the line. -->
-                     line; an inline style, because the committed
-                     tailwind.css has no class for it. -->
                 <div class="flex flex-wrap justify-between items-start gap-4">
                     <code class="text-sm font-medium text-gray-900">{{.Name}}</code>
-                    <code class="text-sm text-gray-900" style="overflow-wrap: anywhere">{{.Value}}</code>
+                    <code class="text-sm text-gray-900 wrap-anywhere">{{.Value}}</code>
                 </div>
                 <p class="text-sm text-gray-500 mt-1">{{.Description}}</p>
             </div>

templates/source_detail.html

@@ -5,12 +5,10 @@
 {{define "content"}}
 <!-- 108rem, half again the 72rem (max-w-6xl) of the webhook list, the
      event log, the navbar and the footer, so an entrypoint URL fits on
-     one line. An inline style, because the committed tailwind.css has
+     one line. wrap-anywhere goes only on names and descriptions: a
-     no class this wide. wrap-anywhere goes only on names and
+     row too wide for a phone must still run past the edge, where the
-     descriptions: a row too wide for a phone must still run past the
+     browser test sees it, rather than break its controls mid-word. -->
-     edge, where the browser test sees it, rather than break its
+<div class="max-w-[108rem] mx-auto px-6 py-8">
-     controls mid-word. -->
-<div class="mx-auto px-6 py-8" style="max-width: 108rem">
     <div class="mb-6">
         <a href="/hooks" class="btn-small">&larr; Back to webhooks</a>
         <div class="flex flex-wrap justify-between items-center gap-2 mt-2">
@@ -23,11 +21,12 @@
             <div class="flex gap-2">
                 <a href="/hook/{{.Webhook.ID}}/events" class="btn-secondary">Full Event Log</a>
                 <a href="/hook/{{.Webhook.ID}}/edit" class="btn-secondary">Edit</a>
-                <!-- The delete prompts are the browser's own, so they
+                <!-- Before a delete form is sent, static/js/app.js
-                     work without the page's scripts. The template
+                     asks in the browser's own prompt, with the form's
-                     escapes each name for the script, so a quote or a
+                     data-confirm text; &#10; is a line break. The
-                     backslash in it shows as typed. -->
+                     template escapes each name for the attribute, so a
-                <form method="POST" action="/hook/{{.Webhook.ID}}/delete" onsubmit="return confirm('Delete webhook &quot;{{.Webhook.Name}}&quot;?\n\nThis deletes its stored events{{with .Stats}} ({{.WithinRetention.Events}}){{end}} and their deliveries. Any archive files it wrote are kept.')">
+                     quote in it shows as typed. -->
+                <form method="POST" action="/hook/{{.Webhook.ID}}/delete" data-confirm="Delete webhook &quot;{{.Webhook.Name}}&quot;?&#10;&#10;This deletes its stored events{{with .Stats}} ({{.WithinRetention.Events}}){{end}} and their deliveries. Any archive files it wrote are kept.">
                     <input type="hidden" name="csrf_token" value="{{.CSRFToken}}">
                     <button type="submit" class="btn-danger">Delete</button>
                 </form>
@@ -90,7 +89,7 @@
                                     {{if .Active}}Deactivate{{else}}Activate{{end}}
                                 </button>
                             </form>
-                            <form method="POST" action="/hook/{{$.Webhook.ID}}/entrypoints/{{.ID}}/delete" onsubmit="return confirm('Delete entrypoint &quot;{{if .Description}}{{.Description}}{{else}}{{$.BaseURL}}/h/{{.Path}}{{end}}&quot;?\n\nSenders using its URL get an error from now on, and the URL cannot be restored.')" class="inline">
+                            <form method="POST" action="/hook/{{$.Webhook.ID}}/entrypoints/{{.ID}}/delete" data-confirm="Delete entrypoint &quot;{{if .Description}}{{.Description}}{{else}}{{$.BaseURL}}/h/{{.Path}}{{end}}&quot;?&#10;&#10;Senders using its URL get an error from now on, and the URL cannot be restored." class="inline">
                                 <input type="hidden" name="csrf_token" value="{{$.CSRFToken}}">
                                 <button type="submit" class="btn-small text-red-600" title="Delete">Delete</button>
                             </form>
@@ -278,7 +277,7 @@
                                     {{if .Active}}Deactivate{{else}}Activate{{end}}
                                 </button>
                             </form>
-                            <form method="POST" action="/hook/{{$.Webhook.ID}}/targets/{{.ID}}/delete" onsubmit="return confirm('Delete target &quot;{{.Name}}&quot;?\n\nNothing more is delivered to it. Its past deliveries stay in the event log.')" class="inline">
+                            <form method="POST" action="/hook/{{$.Webhook.ID}}/targets/{{.ID}}/delete" data-confirm="Delete target &quot;{{.Name}}&quot;?&#10;&#10;Nothing more is delivered to it. Its past deliveries stay in the event log." class="inline">
                                 <input type="hidden" name="csrf_token" value="{{$.CSRFToken}}">
                                 <button type="submit" class="btn-small text-red-600" title="Delete">Delete</button>
                             </form>