docs/parity-checklist.md: header keeps main's date plus the PR's item; the LR-IMP-ADD-DIALOG row keeps main's text and gets the PR's sentence; revision rows union.
docs/whats-new.md: both entries kept.
docs/gaps.md, docs/parity-checklist.md: header keeps the newer date plus both sides' items; revision row added newest first.
locales/ru.json: union of keys.
docs/gaps.md, docs/parity-checklist.md: kept main's header and rows, added the PR's video text and revision row.
crates/engine/src/lightroom_catalog.rs: the PR's two tests moved to lightroom_catalog/tests.rs (main split the module).
Dropped the PR's unrelated hunks (slide_overlay.rs, tests_export.rs, raw corpus tests) per the review.
Require a covering embedded image only for renderable import candidates. Keep the available camera JPEG for preview-only raws and failed or unavailable render fallbacks; ordinary quick views and grid thumbnails retain their immediate stand-in behavior. Also release import preview state when canceled through the control channel.
Co-Authored-By: Codex <noreply@openai.com>
Where the folder listing can't vouch for a name (non-ASCII names, a listing that is gone), each photo's sidecar names are asked of the disk, nine per photo, one photo after another. The batch's lookups now run 16 at a time; the loop that plans the photos takes each one's answer in order.
They use the same `at_once` as the probe checks, so the latency stand-in measured with the previous commit applies to them as well (16 at a time against one at a time); the lookups themselves were not timed on a share for this version.
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
Reusing a scan's probes costs one stat per file to see that it is as it was. They ran one after another, so on a network share every file waited for a round trip. They now run 16 at a time, in the import's batches and in the scan's check of an earlier scan's probes.
Measured on this version with each stat made to take 2 ms (a stand-in for a share's round trip): 5,990 stats took 15.1 s one at a time and 0.95 s through `stat_all`, with the answers in the same order. An earlier version of this change, tried on a real network share, spent most of the import's time in these stats; that figure is not repeated here because it was not taken on this code.
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
The restore actions use the recorded folder without checking that it still
exists, so an upgrade creates it again rather than using the default folder.
`rating:>=3`, `iso:<=400`, `f:>=2.8` read the `=` as part of the number, the
parse failed, and the token silently matched no photo. Parse the two-character
operators first and compare inclusively; `>`, `<`, `=` and bare numbers are
unchanged.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01SqKgSkHi5MfTQFrZvDKBjm
* LR-PREF-SHORTCUTS: the shortcuts panel's Fixed keys ignore keys saved for unknown commands
* KEYC-METADATA: an Add Keywords request the Keywords panel can't take in time expires
* LR-META-INFO: Target Photo saves an Info edit to the photo it shows
Preserve current main, adapt frame/input and IME handling, and keep library
check results stable across layout passes. Include the matching Nix font lock
and reviewed attribution changes.
aiongg