| Comment: | fix(tag): derive the check-in event time from direct date tags only (#372) (#380)
* fix(tag): derive the check-in event time from direct date tags only (#372) Section 4.2, step 6 of repo-state-01 makes a date tag applied directly to a check-in set its event time, whatever order the tag and the check-in arrive in. Crosslinking read the date tagxref row whatever its source, so re-crosslinking after a propagating date tag retimed the descendants that inherited it, and a cancel, an unparsable value or 0 left no time or a time of 0, as fossil 2.28 does. tag.RefreshEventTime now holds the rule, and both tag application and crosslinking call it: a direct, non-cancel date row whose value parses to 0 < r < 5373484.5 sets mtime; anything else restores the D-card time kept in omtime. The date-tag cancel, unparsable and zero cases are no longer deviations, and new before-checkin cases cover the other arrival order. Tracked in #372. * fix(tag): word the date-tag parse rule as step 6 does (#372) * docs(changelog): tidy the entry spacing (#372) |
|---|---|
| Timelines: | family | ancestors | descendants | both | main | trunk |
| Files: | files | file ages | folders |
| SHA3-256: |
30b9d650211f95a04a7dce67474ee839 |
| User & Date: | [email protected] on 2026-10-13 03:36:37.000 |
| Other Links: | branch diff | manifest | tags |
|
2026-10-13
| ||
| 03:51 |
fix(xfer): treat a payload cut short as its card's handler failure (#373) (#381)
sync-protocol-01 Section 5.1 (erratum s5.1-truncated-payload-receiver) makes a file, cfile or uvfile payload that the body ends inside a validation failure of the card's handler, never a stored artifact. The decoder failed the whole message instead, so a server answered HTTP 400 and a client dropped every card of the reply. The decoder now yields such a card as a BadFileCard carrying the handler's malformed-line text and stops, as it already does for a line it cannot frame. The server answers 200 with that error and keeps the cards before it; the client keeps the prefix it processed. fossil 2.28 stores the bytes present; -01 is followed here. check-in: 648f5d4479 user: [email protected] tags: main, trunk | |
| 03:36 |
fix(tag): derive the check-in event time from direct date tags only (#372) (#380)
* fix(tag): derive the check-in event time from direct date tags only (#372) Section 4.2, step 6 of repo-state-01 makes a date tag applied directly to a check-in set its event time, whatever order the tag and the check-in arrive in. Crosslinking read the date tagxref row whatever its source, so re-crosslinking after a propagating date tag retimed the descendants that inherited it, and a cancel, an unparsable value or 0 left no time or a time of 0, as fossil 2.28 does. tag.RefreshEventTime now holds the rule, and both tag application and crosslinking call it: a direct, non-cancel date row whose value parses to 0 < r < 5373484.5 sets mtime; anything else restores the D-card time kept in omtime. The date-tag cancel, unparsable and zero cases are no longer deviations, and new before-checkin cases cover the other arrival order. Tracked in #372. * fix(tag): word the date-tag parse rule as step 6 does (#372) * docs(changelog): tidy the entry spacing (#372) check-in: 30b9d65021 user: [email protected] tags: main, trunk | |
| 03:23 |
fix(manifest): resolve a delta manifest's baseline one level (artifact-format-01 4.7.3) (#379)
* fix(manifest): resolve a delta manifest's baseline one level (#371) A B card may name another delta manifest (artifact-format-01 4.7.3). Drop the #275 refusal in loadBaseline, which ListFiles and crosslink share, and use the named manifest's own F cards. Tracked in #371. * fix(manifest): drop baseline delete markers and never write delta-of-delta (#371) * docs(changelog): note delete markers and the writer rule for delta baselines (#371) check-in: 62396a79e5 user: [email protected] tags: main, trunk | |