ComfyAssets-ComfyUI_PromptManager

Author SHA1 Message Date

Author	SHA1	Message	Date
Vito	f95f66d76a	ci(browser): run page workflows with parallel workers and CI shards (#274 ) * test(browser): one test server per Playwright worker The page workflows started one global webServer and ran with one worker. Each worker now owns its own tests/browser/server.py on an OS-chosen port, with its own temp root and database, through a shared `test` in tests/browser/fixtures.mjs. baseURL, and so page, request and browser.newContext, point at that worker's server. A server that dies fails the running test with its stderr tail and the next test gets a fresh one. The worker count comes from PM_TEST_WORKERS, else 1 under CI, else a quarter of the CPUs up to 8. Chromium workflows in the pinned image: 510 tests in 606 s with one worker on main, 511 (one new isolation spec) in 119 s with eight. * test(browser): test servers bind an OS-chosen port themselves tests/browser/server.py accepts --port 0: it binds 127.0.0.1:0 before building the app, holds that socket for its whole life and prints PM_TEST_SERVER_PORT=<n> to stdout. startBrowserServer defaults to port 0 and reads the announced port, so no port is reserved and released before the server binds and two launchers can no longer race for one. The "address already in use" retry and free-port.mjs are gone; --port N still works and is checked against the announcement. * ci(coverage): shard and merge modes for the frontend coverage run frontend-coverage.mjs now plans its steps from the command line. No arguments still runs everything, then reports and checks. --shard I/N runs one slice of the chromium+touch workflows into coverage/js/raw without wiping it or reporting; shard 1 also runs the node units and the viewer suite. --merge-only reports and checks over raw coverage the shards wrote, with --update-baseline as before, and refuses an empty or missing raw dir. npm scripts: test:coverage:shard and coverage:merge. * ci(viewerjs): run browser suites with workers and shards Frontend coverage runs as three shards (PM_TEST_WORKERS=3 each) that upload their raw V8 files as coverage-raw-N; a merge job downloads them, fails unless all three arrived with raw files and no clashing names, then runs coverage:merge. Firefox and WebKit page workflows run as two shards each with three workers. The required viewerjs context keeps its name and now needs the shards, the merge and the browser shards, failing unless every one succeeded. * ci(viewerjs): cache npm downloads and size timeouts from the first sharded run The first sharded run lost a webkit shard to its 12-minute limit after all 267 tests passed, and npm ci stalled for 5 minutes in six of eight jobs. Each npm ci now restores npm's download cache keyed on the vendor lockfile, and browser job timeouts are the measured times plus at least 50%, never under 20 minutes. The workflow comment records the measurements. * test(browser): wait for rendering frames, not timers or luck, under load Two Library workflows failed in CI's webkit shard 2/2 (run 39714840809) under parallel load: - "library infinite loading ..." counted cards right after the last page hid "Load more". That render queues the window update for the next frame, so a slow frame left the 120 cards of the all-rows layout in place (received 120, expected < 80). It now waits for the next frame, which runs after every callback the page queued. - "row measurement preserves anchor ..." settled with a 60 ms timer. A scroll event is delivered in a rendering frame, so on a starved renderer the timer ended before the scroll's frame and no card was in view (null anchor). settleLibrary now waits for two consecutive native frames with no Library frame queued, checked inside the frame callback. Before/after: a scratch copy that delays every frame by 120 ms (as a starved runner) failed the old helpers 19 of 40 runs, at the CI lines, and passed the new ones 40 of 40 on chromium and webkit. Throttled to 1 CPU with 4 workers, the original spec failed 1 of 42 for each test. The Tags viewer workflow closed the viewer as soon as its image was visible, before ViewerJS rendered the title, so CI's coverage never ran the page's title callback (tags-page.js functions 26/46 = 56.52%, one function above the floor of 56; locally 27/46). The test now asserts the title names the shown file, which always runs the callback. * test(browser): stable under parallel workers Document how to run the browser suites in parallel and sharded, with the measured times, and what keeps specs stable beside each other: wait for a rendering frame, a response or a visible state, never for time. With `b665aa6` the full workflows suite passed on chromium, touch, firefox and webkit with 8 workers, each test 3 times (6132 passed, 0 failed, 24 minutes), and once at CI size (docker --cpus=4, 3 workers: 2044 passed, 25.5 minutes). The npm download cache removed the npm ci stall: 0 to 2 s in every job of run 39722172529, against about 300 s in most jobs before. The workflow's timing comment now records that run; the whole workflow took 8m54s against about 33 minutes before sharding. Shard imbalance (shard 1 carries the concept-comparison specs) is documented, not tuned. * test(browser): server stderr on failures, no orphan servers, waits on state Final review fixes for the per-worker test servers (#274): - A failed test gets the stderr its server wrote during that test attached as `server-stderr`, so an aiohttp traceback behind a 500 reaches the CI report. The launcher keeps the newest 256 KB with a stream position per test (stderrMark/stderrSince), and the fixture attaches it only when the test did not end as expected. - The port announcement counts only as a complete line; a chunk ending mid-number was read as a shorter port. - server.py --exit-on-stdin-eof shuts down gracefully (database closed, temp root removed) when stdin closes. The launcher holds that pipe, so a SIGKILLed worker leaves no server behind, on every OS. - settled() gives up after 600 frames with the observation stats instead of hanging to the test timeout. - The Classic text-sync test waits for the page's 50 ms timer to install serializeValue instead of sleeping 60 ms. With that timer run 100 ms late (a scratch copy), the old wait failed 20/20 and the poll passed 20/20; the spec passed 240/240 at 8 workers x10 and 240/240 at 1 CPU with 4 workers. - The workflow test checks each npm cache key names its job's pinned image tag. Every new unit test failed before its change (missing export, the old regex timing out on the split announcement, an unknown server flag, a hung settle, a mismatched cache key). * docs(testing): server lifetime, stderr attachments and the final measured CI times TESTING.md describes the stdin-EOF shutdown and the server-stderr attachment, states that the npm cache key names the job's image tag (now tested), quotes run 39725176320 at `cbc8a61` (8m38s) and explains the 20-minute browser job limit. The workflow timing comment quotes the same run. * fix(browser): server shutdown and its tests on Windows, Python < 3.13 and CI Three CI failures on `ed809f7`: - Windows: server.py exited 1 on stdin EOF because create_app works with the temp root as its cwd, and Windows cannot remove a process's cwd. main() now restores the original cwd before the root is removed. - Python 3.10 to 3.12: the lifetime test's cleanup called communicate() after closing stdin, which flushes the closed pipe (ValueError); 3.13 tolerates it. It now waits and closes the pipes itself, and the exit assertion shows the server's stderr. Under 3.12: errors=2 before, OK after. - Tooling (CI container): the SIGKILLed-worker tests treated an exited server as alive, because the container's PID 1 does not reap orphans and kill(pid, 0) succeeds on a zombie. Reproduced locally with docker without --init (2 failed); a zombie now counts as exited (2 passed).	2026-10-21 23:10:57 -07:00
Vito	2cecfcf0ac	feat(activity): scan new and unresolved files by default, and download a job report (#273 ) * feat(activity): scan new and unresolved files by default, all files on request Scan for metadata now skips, before reading any metadata, the files the Library already links with a resolved link (one generated_images row with exactly one pair link), found with one streamed query that keeps only the scanned paths and compares paths with os.path.normcase on both sides. Skipped files count as Already known ("Skipped · already linked"). Unlinked, fallback-linked, ambiguous and unresolved files are read again. POST /prompt_manager/jobs/scan takes {scope: "new" \| "all"} (default "new", anything else is a 400). The review dialog shows the concept's "New and unresolved files" and adds an "All files" switch (D58). The legacy /prompt_manager/scan route keeps scanning all files with unchanged SSE payloads. * feat(activity): download a job report GET /prompt_manager/jobs/{id}/report returns the summary, counters, every outcome group's listed rows with truncated flags, the scope and the log as a JSON attachment (prompt-manager-<kind>-<YYYYMMDD-HHMM>.json), with no absolute paths. The job view's Download report saves it through the api-client URL, and the scan log opens at its newest line and follows new lines only while the reader is at the bottom. * fix(activity): re-read known files whose size changed, and review follow-ups The new scope now skips a known, resolved file only while its stored generated_images.file_size equals its size on disk; a NULL stored size is re-read. A file replaced at the same path is read again and its conflict shows under Needs review. The scan stores each file's size when it ingests it, and both sides of the known-set match go through normalize_image_path. A failed known-set read logs and falls back to reading every file, and an empty request body asks for the default scope. The All files switch is described by its hint, a lost job's Download report is disabled, the report toast is skipped once the view is gone, and the scan log opens at its newest line every time. Fixtures use the real scope shape. * test(activity): wait for the job view's settled detail read, not a fixed delay The 200-row cap test now starts counting detail reads once one read has returned the full listing and none is in flight. The job view re-reads after an in-flight read only when the summary still lists more than that read returned, so no redundant read is left due after opening.	2026-10-21 18:07:16 -07:00
Vito	a6f6697c50	feat(activity): shared job registry, Activity page and Scan for metadata (#269 ) * feat(activity): shared session job registry with the scan as its first kind The output scan now runs as the first kind of an in-memory JobRegistry (py/api/jobs.py, replacing py/api/scan_job.py). Each job keeps exact counters, at most 200 outcome rows per group and a log capped at 2,000 lines, with root-relative paths and public root labels only. New routes list, inspect, follow (SSE) and cancel jobs and start the scan (POST /prompt_manager/jobs/scan, 409 busy while another record-changing job runs). The legacy /prompt_manager/scan routes attach to the same job and stream unchanged payloads. * feat(activity): concept Activity page, top-bar button and progress pill The workspace gains the concept's Activity page (renderActivity), its top-bar button and sidebar row, fed by one session job store: a /jobs read, one SSE stream per running job, a 1-30 s capped backoff after a dropped stream and a 15 s list poll while visible. A job the server no longer lists shows as ended, so no pill stays stuck after a restart. D52: the counts show running jobs and hide at 0; the phone top-bar row may wrap. D53: the mockup's progress pill links the running job. D54: real job items with state chips, detail lines and Cancel job. Start and finish toasts follow the spec copy. Fake browser jobs come from tests/browser/workspace_activity_fixture.py. * feat(activity): Scan for metadata with its review dialog and job view Images -> Import keeps Rescan and adds the concept's "Review scan" row as "Scan for metadata". It opens the concept's "Review metadata scan" with the real source count, the fixed "All files" scope, a backup box that downloads the existing database backup, and Cancel / Start scan. Start scan starts the session scan job and opens its view; a start refused by another record-changing job toasts its title as a link. #/activity/<id> is now the approved job-view mockup: the heading with Download report (disabled until Activity-2) and Cancel scan, the Progress surface with the five counter tiles, the Results surface with the outcome list, path filter and capped rows (Added and Already known open the viewer through the listing identity), and the collapsed Scan log. Activity item titles link their views. The scan's progress line groups thousands. Deviations D55-D57 are recorded in TESTING.md. * fix(activity): stop the job store on pagehide and let proxied routes outlive teardown CI's touch coverage run failed "a resolver failure for the ComfyUI link never stops the shell from mounting" with "route.fulfill: Fetch response has been disposed": its trace shows two Library thumbnail requests started by the last route still in the spec's proxy-mount handler while the page fixture was torn down. - The proxy-mount handler ignores only that teardown error, and the workspace and Activity specs unroute with ignoreErrors before teardown. - The Activity store now ends its work on pagehide: the poll and backoff timers stop, the job streams close and a list read in flight is aborted (an older read is also aborted when a newer one starts). A pageshow from the back-forward cache reads the list again. * test(activity): compare the job view log's line height to the hundredth of a pixel WebKit reports the mockup's 11px/1.7 line height as 18.700001px, so the job view comparison failed on WebKit only. The measurement now rounds it, as other engine-dependent values are normalized. * fix(activity): final review round for the job store, job view and scan rows - Store: one failed list read no longer marks running jobs as ended; two in a row, or a good list without them, do (M1). A good read resets the backoff, except for a stream that keeps dropping while the list reads fine (M2). Finished jobs the server pruned are dropped (M3). Numbers group as en-US, like the server's lines (M13). - Job view: a job the store's list no longer has shows the not-found state (M3). Focus moves to the job heading when Cancel scan goes away, and to the item's title link in the list (M4). The detail is read only when its listed rows change or the open log grows, using the summary's new listed and log_length (M12). - Viewer rows open by exact identity: each scan row records its root's listing index, the source must still carry the row's label, and the workspace listing gains an exact `path` filter (M8, M9). - Scan rows name the prompt by its Library title, stored title first (M10). - A toast with a link stays 20 s, and any toast stays while hovered or focused (M14). - A real two-batch scan cancelled after batch 1 keeps partial counts (M7). - TESTING.md: D54, D56 and D57 updated (M5, M6). * test(activity): cover the held and linked toast and a job view pruned by the server Unit tests for the toast's 6.5 s and 20 s timeouts and its hover and focus hold, and a browser test for a finished job's open view after the server stops listing it. Coverage floors are only raised.	2026-10-21 14:22:30 -07:00

Vito

f95f66d76a

ci(browser): run page workflows with parallel workers and CI shards (#274 )

* test(browser): one test server per Playwright worker

The page workflows started one global webServer and ran with one worker.
Each worker now owns its own tests/browser/server.py on an OS-chosen port,
with its own temp root and database, through a shared `test` in
tests/browser/fixtures.mjs. baseURL, and so page, request and
browser.newContext, point at that worker's server. A server that dies
fails the running test with its stderr tail and the next test gets a
fresh one. The worker count comes from PM_TEST_WORKERS, else 1 under CI,
else a quarter of the CPUs up to 8.

Chromium workflows in the pinned image: 510 tests in 606 s with one
worker on main, 511 (one new isolation spec) in 119 s with eight.

* test(browser): test servers bind an OS-chosen port themselves

tests/browser/server.py accepts --port 0: it binds 127.0.0.1:0 before
building the app, holds that socket for its whole life and prints
PM_TEST_SERVER_PORT=<n> to stdout. startBrowserServer defaults to port 0
and reads the announced port, so no port is reserved and released before
the server binds and two launchers can no longer race for one. The
"address already in use" retry and free-port.mjs are gone; --port N still
works and is checked against the announcement.

* ci(coverage): shard and merge modes for the frontend coverage run

frontend-coverage.mjs now plans its steps from the command line.
No arguments still runs everything, then reports and checks.
--shard I/N runs one slice of the chromium+touch workflows into
coverage/js/raw without wiping it or reporting; shard 1 also runs the
node units and the viewer suite. --merge-only reports and checks over raw
coverage the shards wrote, with --update-baseline as before, and refuses
an empty or missing raw dir. npm scripts: test:coverage:shard and
coverage:merge.

* ci(viewerjs): run browser suites with workers and shards

Frontend coverage runs as three shards (PM_TEST_WORKERS=3 each) that upload
their raw V8 files as coverage-raw-N; a merge job downloads them, fails unless
all three arrived with raw files and no clashing names, then runs
coverage:merge. Firefox and WebKit page workflows run as two shards each with
three workers. The required viewerjs context keeps its name and now needs the
shards, the merge and the browser shards, failing unless every one succeeded.

* ci(viewerjs): cache npm downloads and size timeouts from the first sharded run

The first sharded run lost a webkit shard to its 12-minute limit after all
267 tests passed, and npm ci stalled for 5 minutes in six of eight jobs. Each
npm ci now restores npm's download cache keyed on the vendor lockfile, and
browser job timeouts are the measured times plus at least 50%, never under 20
minutes. The workflow comment records the measurements.

* test(browser): wait for rendering frames, not timers or luck, under load

Two Library workflows failed in CI's webkit shard 2/2 (run 39714840809)
under parallel load:

- "library infinite loading ..." counted cards right after the last
  page hid "Load more". That render queues the window update for the
  next frame, so a slow frame left the 120 cards of the all-rows
  layout in place (received 120, expected < 80). It now waits for the
  next frame, which runs after every callback the page queued.
- "row measurement preserves anchor ..." settled with a 60 ms timer. A
  scroll event is delivered in a rendering frame, so on a starved
  renderer the timer ended before the scroll's frame and no card was in
  view (null anchor). settleLibrary now waits for two consecutive
  native frames with no Library frame queued, checked inside the frame
  callback.

Before/after: a scratch copy that delays every frame by 120 ms (as a
starved runner) failed the old helpers 19 of 40 runs, at the CI lines,
and passed the new ones 40 of 40 on chromium and webkit. Throttled to
1 CPU with 4 workers, the original spec failed 1 of 42 for each test.

The Tags viewer workflow closed the viewer as soon as its image was
visible, before ViewerJS rendered the title, so CI's coverage never ran
the page's title callback (tags-page.js functions 26/46 = 56.52%, one
function above the floor of 56; locally 27/46). The test now asserts the
title names the shown file, which always runs the callback.

* test(browser): stable under parallel workers

Document how to run the browser suites in parallel and sharded, with the
measured times, and what keeps specs stable beside each other: wait for
a rendering frame, a response or a visible state, never for time.

With b665aa6 the full workflows suite passed on chromium, touch, firefox
and webkit with 8 workers, each test 3 times (6132 passed, 0 failed,
24 minutes), and once at CI size (docker --cpus=4, 3 workers: 2044
passed, 25.5 minutes).

The npm download cache removed the npm ci stall: 0 to 2 s in every job
of run 39722172529, against about 300 s in most jobs before. The
workflow's timing comment now records that run; the whole workflow took
8m54s against about 33 minutes before sharding. Shard imbalance (shard 1
carries the concept-comparison specs) is documented, not tuned.

* test(browser): server stderr on failures, no orphan servers, waits on state

Final review fixes for the per-worker test servers (#274):

- A failed test gets the stderr its server wrote during that test
  attached as `server-stderr`, so an aiohttp traceback behind a 500
  reaches the CI report. The launcher keeps the newest 256 KB with a
  stream position per test (stderrMark/stderrSince), and the fixture
  attaches it only when the test did not end as expected.
- The port announcement counts only as a complete line; a chunk ending
  mid-number was read as a shorter port.
- server.py --exit-on-stdin-eof shuts down gracefully (database closed,
  temp root removed) when stdin closes. The launcher holds that pipe,
  so a SIGKILLed worker leaves no server behind, on every OS.
- settled() gives up after 600 frames with the observation stats
  instead of hanging to the test timeout.
- The Classic text-sync test waits for the page's 50 ms timer to
  install serializeValue instead of sleeping 60 ms. With that timer run
  100 ms late (a scratch copy), the old wait failed 20/20 and the poll
  passed 20/20; the spec passed 240/240 at 8 workers x10 and 240/240
  at 1 CPU with 4 workers.
- The workflow test checks each npm cache key names its job's pinned
  image tag.

Every new unit test failed before its change (missing export, the old
regex timing out on the split announcement, an unknown server flag, a
hung settle, a mismatched cache key).

* docs(testing): server lifetime, stderr attachments and the final measured CI times

TESTING.md describes the stdin-EOF shutdown and the server-stderr
attachment, states that the npm cache key names the job's image tag
(now tested), quotes run 39725176320 at cbc8a61 (8m38s) and explains the
20-minute browser job limit. The workflow timing comment quotes the same
run.

* fix(browser): server shutdown and its tests on Windows, Python < 3.13 and CI

Three CI failures on ed809f7:

- Windows: server.py exited 1 on stdin EOF because create_app works with
  the temp root as its cwd, and Windows cannot remove a process's cwd.
  main() now restores the original cwd before the root is removed.
- Python 3.10 to 3.12: the lifetime test's cleanup called communicate()
  after closing stdin, which flushes the closed pipe (ValueError); 3.13
  tolerates it. It now waits and closes the pipes itself, and the exit
  assertion shows the server's stderr. Under 3.12: errors=2 before, OK
  after.
- Tooling (CI container): the SIGKILLed-worker tests treated an exited
  server as alive, because the container's PID 1 does not reap orphans
  and kill(pid, 0) succeeds on a zombie. Reproduced locally with docker
  without --init (2 failed); a zombie now counts as exited (2 passed).

2026-10-21 23:10:57 -07:00

Vito

2cecfcf0ac

feat(activity): scan new and unresolved files by default, and download a job report (#273 )

* feat(activity): scan new and unresolved files by default, all files on request

Scan for metadata now skips, before reading any metadata, the files the
Library already links with a resolved link (one generated_images row with
exactly one pair link), found with one streamed query that keeps only the
scanned paths and compares paths with os.path.normcase on both sides.
Skipped files count as Already known ("Skipped · already linked").
Unlinked, fallback-linked, ambiguous and unresolved files are read again.

POST /prompt_manager/jobs/scan takes {scope: "new" | "all"} (default
"new", anything else is a 400). The review dialog shows the concept's
"New and unresolved files" and adds an "All files" switch (D58). The
legacy /prompt_manager/scan route keeps scanning all files with unchanged
SSE payloads.

* feat(activity): download a job report

GET /prompt_manager/jobs/{id}/report returns the summary, counters, every
outcome group's listed rows with truncated flags, the scope and the log as a
JSON attachment (prompt-manager-<kind>-<YYYYMMDD-HHMM>.json), with no absolute
paths. The job view's Download report saves it through the api-client URL, and
the scan log opens at its newest line and follows new lines only while the
reader is at the bottom.

* fix(activity): re-read known files whose size changed, and review follow-ups

The new scope now skips a known, resolved file only while its stored
generated_images.file_size equals its size on disk; a NULL stored size is
re-read. A file replaced at the same path is read again and its conflict
shows under Needs review. The scan stores each file's size when it ingests
it, and both sides of the known-set match go through normalize_image_path.

A failed known-set read logs and falls back to reading every file, and an
empty request body asks for the default scope. The All files switch is
described by its hint, a lost job's Download report is disabled, the report
toast is skipped once the view is gone, and the scan log opens at its
newest line every time. Fixtures use the real scope shape.

* test(activity): wait for the job view's settled detail read, not a fixed delay

The 200-row cap test now starts counting detail reads once one read has
returned the full listing and none is in flight. The job view re-reads after
an in-flight read only when the summary still lists more than that read
returned, so no redundant read is left due after opening.

2026-10-21 18:07:16 -07:00

Vito

a6f6697c50

feat(activity): shared job registry, Activity page and Scan for metadata (#269 )

* feat(activity): shared session job registry with the scan as its first kind

The output scan now runs as the first kind of an in-memory JobRegistry
(py/api/jobs.py, replacing py/api/scan_job.py). Each job keeps exact
counters, at most 200 outcome rows per group and a log capped at 2,000
lines, with root-relative paths and public root labels only. New routes
list, inspect, follow (SSE) and cancel jobs and start the scan
(POST /prompt_manager/jobs/scan, 409 busy while another record-changing
job runs). The legacy /prompt_manager/scan routes attach to the same job
and stream unchanged payloads.

* feat(activity): concept Activity page, top-bar button and progress pill

The workspace gains the concept's Activity page (renderActivity), its
top-bar button and sidebar row, fed by one session job store: a /jobs
read, one SSE stream per running job, a 1-30 s capped backoff after a
dropped stream and a 15 s list poll while visible. A job the server no
longer lists shows as ended, so no pill stays stuck after a restart.

D52: the counts show running jobs and hide at 0; the phone top-bar row
may wrap. D53: the mockup's progress pill links the running job. D54:
real job items with state chips, detail lines and Cancel job. Start and
finish toasts follow the spec copy. Fake browser jobs come from
tests/browser/workspace_activity_fixture.py.

* feat(activity): Scan for metadata with its review dialog and job view

Images -> Import keeps Rescan and adds the concept's "Review scan" row as
"Scan for metadata". It opens the concept's "Review metadata scan" with the
real source count, the fixed "All files" scope, a backup box that downloads
the existing database backup, and Cancel / Start scan. Start scan starts the
session scan job and opens its view; a start refused by another
record-changing job toasts its title as a link.

#/activity/<id> is now the approved job-view mockup: the heading with
Download report (disabled until Activity-2) and Cancel scan, the Progress
surface with the five counter tiles, the Results surface with the outcome
list, path filter and capped rows (Added and Already known open the viewer
through the listing identity), and the collapsed Scan log. Activity item
titles link their views. The scan's progress line groups thousands.

Deviations D55-D57 are recorded in TESTING.md.

* fix(activity): stop the job store on pagehide and let proxied routes outlive teardown

CI's touch coverage run failed "a resolver failure for the ComfyUI link never
stops the shell from mounting" with "route.fulfill: Fetch response has been
disposed": its trace shows two Library thumbnail requests started by the last
route still in the spec's proxy-mount handler while the page fixture was torn
down.

- The proxy-mount handler ignores only that teardown error, and the workspace
and Activity specs unroute with ignoreErrors before teardown.
- The Activity store now ends its work on pagehide: the poll and backoff timers
stop, the job streams close and a list read in flight is aborted (an older
read is also aborted when a newer one starts). A pageshow from the
back-forward cache reads the list again.

* test(activity): compare the job view log's line height to the hundredth of a pixel

WebKit reports the mockup's 11px/1.7 line height as 18.700001px, so the job view comparison failed on WebKit only. The measurement now rounds it, as other engine-dependent values are normalized.

* fix(activity): final review round for the job store, job view and scan rows

- Store: one failed list read no longer marks running jobs as ended; two in
a row, or a good list without them, do (M1). A good read resets the
backoff, except for a stream that keeps dropping while the list reads fine
(M2). Finished jobs the server pruned are dropped (M3). Numbers group as
en-US, like the server's lines (M13).
- Job view: a job the store's list no longer has shows the not-found state
(M3). Focus moves to the job heading when Cancel scan goes away, and to the
item's title link in the list (M4). The detail is read only when its listed
rows change or the open log grows, using the summary's new listed and
log_length (M12).
- Viewer rows open by exact identity: each scan row records its root's
listing index, the source must still carry the row's label, and the
workspace listing gains an exact `path` filter (M8, M9).
- Scan rows name the prompt by its Library title, stored title first (M10).
- A toast with a link stays 20 s, and any toast stays while hovered or
focused (M14).
- A real two-batch scan cancelled after batch 1 keeps partial counts (M7).
- TESTING.md: D54, D56 and D57 updated (M5, M6).

* test(activity): cover the held and linked toast and a job view pruned by the server

Unit tests for the toast's 6.5 s and 20 s timeouts and its hover and focus hold, and a browser test for a finished job's open view after the server stops listing it. Coverage floors are only raised.

2026-10-21 14:22:30 -07:00