crates/engine/src/cmd/library.rs: kept main's split; the PR's parse_gps changes and tests are ported to crates/engine/src/cmd/library/gps.rs.
docs/gaps.md, docs/parity-checklist.md: both sides kept (header items, rows, LR-META-LOCATION note).
crates/ui-egui/tests/it/main.rs: both mod lines kept. docs/whats-new.md, docs/parity-checklist.md: both sides' entries, header and LR-BEHAV-ACCESS row combined.
docs/parity-checklist.md, docs/ui-parity.md: headers, the KEYC-METADATA and LRC-LIB-KEYWORDS rows carry both sides' changes; revision rows kept from both.
photo.clearLabel takes no colour, but the toast code read params[label] for
it as for photo.label, so photo.clearLabel {label: red} cleared the photo and
showed 'Red Label'. The colour is read for photo.label only.
Setting a colour label is photo.label with the colour as a parameter, and the
keyboard shortcuts editor binds a key to a command, not to a parameter, so
only 6-9 had keys and clearing a label could not be given one. photo.clearLabel
is the same as photo.label with none, as a command of its own. The Photo >
Set Color Label > None row runs it and shows its key.
The note now goes when another folder is watched, a listing of a folder no longer watched is ignored, and "can be read again" is logged only for the folder that failed.
A data line such as `nan 0 0` starts with a letter and was skipped as an
unknown keyword, so a file with enough other entries still parsed.
(CodeRabbit review on #2540.)
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
A watched folder that was removed, unmounted or not allowed looked like one with nothing new: the listing error was logged at debug only. It is now logged as a warning once when it starts (and once when the folder can be read again), and Settings > Import shows the reason under the watched folder.
The previous commit's sections.rs went through a PowerShell read and
write in the Windows ANSI code page, which turned …, – and ▸ into
mojibake in five album-menu labels (Create Album…, Sort Albums A–Z, …)
and two doc comments, so the labels missed their translation keys
(CodeRabbit on #3392). The file's text outside missing_count is now
byte for byte what it was before.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
A catalog change seen while a timed refresh of the Missing Photos count
is running cannot start its recount until the refresh ends, and the
refresh is not pending work, so settle() could return before either
finished and the row the recount adds moved the folder rows afterwards
(CodeRabbit on #3392).
The change is now counted as pending from the frame it is seen
(`owed`); the recount takes that entry over when it starts and drops it
once its count is stored. settling_waits_for_a_recount_owed_during_a_timed_refresh
adds a missing photo while a slow refresh runs and fails without this.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
While a headless run settles, its simulated clock runs several times
faster than wall time, so with every Missing Photos count pending, a
count that takes about a second started the next 5 s refresh before the
last one finished and settle() never ended (120 s timeout with a 1 s
count, 48 s with 600 ms). That would also stall lightcraft-cli snapshot
on a library on a slow share.
Only the first count and a recount after a catalog change, which add or
take away the Missing Photos row, are pending work now; the timed
refresh is not. Settling takes as long as the count (1.0 s for 1 s,
3.0 s for 3 s). The regression test also checks that settling ends
while a slow count refreshes.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
aiongg