fix: a popup save no longer undoes a network switch the background saved during it (closes #641)
check / check (push) Waiting to run
e2e / e2e-chrome (push) Waiting to run
e2e / e2e-firefox (push) Waiting to run

The popup's saveState() and the background's updateState() each read the
stored record, change it and write it back, each ordered only within its own
context, so a write by one could land between the other's read and write and
be lost. Both now hold the browser's lock (navigator.locks) from the read to
the write, through withLock() in src/shared/browserApi.js, and the recovery
screen's reset takes it too. A page in a Firefox private window gets locks of
its own and is not covered. The lock also orders a page's own saves, so both
per-context queues are gone. The unit tests get a small lock from a Jest setup
file.

Model: opus-5-5
This commit was merged in pull request #644.
This commit is contained in:
2026-10-20 20:13:19 +02:00
parent 4472746b89
commit 93c34e1a2a
19 changed files with 240 additions and 67 deletions
+19 -26
View File
@@ -16,17 +16,20 @@
// across any number of awaits and no concurrent work can
// move it.
// updateState(fn) read fresh, apply fn to that fresh record, write it
// back — all inside a queue, so two background writes
// never interleave, and the read is one storage round trip
// ahead of the write rather than a page lifetime ahead of
// it (which is what made the popup's saveState() need a
// back — all holding the lock every write of the stored
// record takes, so no other write lands in between, by
// the background or by any page except a page in a
// Firefox private window (README, "Stored state and its
// version"). The read is one storage round trip ahead of
// the write rather than a page lifetime ahead of it
// (which is what made the popup's saveState() need a
// per-field merge against a baseline at all).
//
// A handler that must both read and write therefore does its network work
// against a snapshot it owns, and applies the RESULT inside updateState().
// It never publishes an object other in-flight work is holding.
const { storageGet, storageSet } = require("../shared/browserApi");
const { storageGet, storageSet, withLock } = require("../shared/browserApi");
const { normalizePersisted } = require("../shared/persistedState");
const {
STATE_SCHEMA_VERSION,
@@ -51,11 +54,6 @@ async function getState() {
return normalizePersisted(result.autistmask);
}
// Serializes the read-modify-write turns below. Two of them interleaved would
// each read before the other wrote, and the second write would carry the first
// one's fields back to their pre-turn values.
let updateQueue = Promise.resolve();
async function updateStateOnce(mutate) {
const s = await getState();
await mutate(s);
@@ -70,27 +68,22 @@ async function updateStateOnce(mutate) {
// Apply `mutate` to a record read fresh from storage and write the result
// back. `mutate` receives a detached, normalized profile and mutates it in
// place; it may be async, but it must not do anything slow — the window
// between the read and the write is the window in which another context's
// write is lost, and keeping it to one storage round trip is what makes a
// whole-record write safe here. Concretely: the write is the WHOLE record, so
// a popup write that lands inside that window is reverted, in every field, by
// the record this turn read before it. That is accepted because the window is
// one round trip long and the popup is not writing while the worker is;
// widening it is what would make it a real hazard.
// place; it may be async, but it must not do anything slow. The write is the
// WHOLE record, which is safe only because the lock (withLock() in
// src/shared/browserApi.js) is held from the read to the write: two
// background turns, or a turn and a page's save, never interleave
// (https://git.eeqj.de/sneak/AutistMask/issues/641). Every page's save waits
// while it is held, except a page in a Firefox private window, which Firefox
// gives locks of its own (README, "Stored state and its version").
//
// `mutate` must also not call updateState() itself, directly or through
// anything it awaits: the queue is strictly serial, so the inner turn waits on
// the outer one, which is waiting on the inner one. That deadlocks the whole
// background, not just the caller. Mutate the record you were handed.
// anything it awaits: the inner turn waits for the lock the outer one holds,
// and the outer one is waiting on the inner one. That deadlocks every write of
// the stored record, not just the caller. Mutate the record you were handed.
//
// Resolves with the record that was written.
function updateState(mutate) {
const turn = updateQueue.then(() => updateStateOnce(mutate));
// The queue must advance even when a turn rejects, or every update after
// it queues behind a promise that never settles.
updateQueue = turn.catch(() => {});
return turn;
return withLock("autistmask", () => updateStateOnce(mutate));
}
module.exports = { getState, updateState };