fix(copies): a changed photo no longer wedges a merge's resumed verify

A merge's Verify stage journals a second version before filing it. When
the copy then failed or read other bytes (the file rewritten by an
editor or a sync tool, a drive dropping out before an edit), every retry
read new bytes for that journalled frame and stopped with changed,
keeping the journal, so the merge failed the same way until cancelled.

- When the kept sequence does not hold the journalled bytes, nothing was
  filed: the frame leaves SecondVersions and VersionShas and is verified
  as on a first visit, which files the photo as it is now.
- When it holds them (filed, then the exclusion failed), they are left
  out of Main and the merge stops and drops its journal, so the next run
  plans again instead of a cancel leaving them in Main for good.

Refs #73

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
This commit is contained in:
Alby96andClaude Opus 5.5 committed 2026-11-18 14:30:27 +01:00
1 parent 6db4c2e80c
commit 65b7cdf91b
5 files changed
+115 -3

No files matched your search

+3
View File
@@ -245,6 +245,9 @@ match. A 1.1 server works as before, without Integrate; a 1.2 server adds the ph
details instead of overwriting them.
- A second version a merge's verify files is left out of Main also when the run stopped right after filing it (a
drive dropping out, project.json held, the app closing): the resumed merge leaves it out then.
- A merge whose verify stopped before filing a second version, and whose photo changed before the retry, files the
photo as it is now instead of failing on every retry; one that had filed it leaves it out of Main and asks for a
new comparison.
- *Upload all local* leaves out again a sequence answered *Keep here* after its server copy was deleted, and the
answer stays; only a *Keep apart* given before the deletion lets it send the sequence.
- An upload that joined a server sequence this PC already knew still compares its details when it is resumed or
+1
View File
@@ -1031,3 +1031,4 @@ One line per decision: date · decision · alternative · reason.
| 2026-11-18 | A merge's Verify stage journals a second version (`SecondVersions`, and in `VersionShas` the SHA-256 it is filed as) before it files it; a resumed verify that finds such a frame in the kept sequence leaves it out of Main again (a no-op when it already is), and one that reads other bytes for it stops with "changed" (#73) | Save the journal right after filing | The ingest commits the frame before any save can follow: a failed exclusion or a crash in between left the version in Main for good, since a resumed verify read it as the kept sequence's own |
| 2026-11-18 | Supersedes the 2026-11-18 unlinked-404 rows on Upload all local: a Kept apart pair whose server sequence reads deleted on the server holds nothing back only when it was answered before the first 404 (`AnsweredAt < RemoteGoneAt`); a Keep here (answered after the deletion was shown) still skips the local sequence (#73) | A separate Keep here state or marker | Both answers are stored as Kept apart with the gone mark; the answer time already tells them apart, and Keep here's tooltip and design §6.3 promise Upload all local leaves it out |
| 2026-11-18 | An upload's checkpoint names the server sequence its first original reached with `:joined` appended when that was one this PC already knew; a resumed or retried upload sends all its details only when the checkpoint names the target without that mark (a plain id from an older run still means made by this job) (#73) | Look the sequence up again in the known remotes on resume | After the first attempt the link to it exists, so a resume cannot tell a sequence it made from one it joined; only the attempt that sent the first original knows |
| 2026-11-18 | Supersedes the 2026-11-18 `VersionShas` row where a resumed verify reads other bytes for a journalled second version: when the kept sequence does not hold the journalled SHA-256 (nothing was filed) the frame leaves `SecondVersions` and `VersionShas` and is verified as on a first visit, which files the bytes it reads now; when it holds them, they are left out of Main and the merge stops with "changed" and drops its journal, so the next run plans again (#73) | Treat the new bytes as another second version in the same run | A stop that kept the journal failed the same way on every retry once the file changed (an editor saving, a sync tool, an edit before a retry after a failed copy), and a cancel then left a filed version in Main; filing a third version unplanned goes further than the plan the user confirmed, while a new plan compares the new bytes and shows them |