50 descendants and 50 ancestors of ccbcf4b34ecf81e9
|
2026-09-27
| ||
| 04:47 | Rows lost to a failed or cancelled write count as dropped, and the ops page shows the count check-in: 9f7e279623 user: dmestas tags: ops-history-flush-off-request-path | |
| 04:45 | A yielded Tickets tab is reachable from the narrow-screen drawer closed check-in: 8366613005 user: zeitgeist tags: tickets-in-the-drawer-when-it-yields | |
| 04:44 | Ingest rounds shed at the lane back off and are journalled once past a deadline, and a round queued for a rebuild slot leaves the lane while it waits check-in: 4892dad217 user: dmestas tags: ingest-lane-shed-is-backpressure | |
| 04:44 | The stop tests wait on events, not clocks: a stub zit that blocks on a pipe, and in-flight witnesses for the streams check-in: 3855608e6a user: dmestas tags: stop-tests-wait-on-events | |
| 04:44 | Private-repository list answers private for every repository until a load has succeeded check-in: ffb4a5507c user: dmestas tags: visibility-list-refreshes-on-a-ttl | |
| 04:37 | Attempt limits come from the definition each run is pinned to, and an unfinished attempt is never called failed (#283) check-in: a54cc27949 user: dmestas tags: ci-attempt-of-and-settled-verdict | |
| 04:24 | A losing promotion reads the ticket before the redirect, so it is told where the ticket went check-in: 8a00a028b3 user: dmestas tags: losing-promotion-names-where-it-went | |
| 04:23 | land ci-steps-run-once check-in: 0d40c7a669 user: dmestas tags: trunk | |
| 04:19 | Test a stop during a checkout wait; keep the run-once rationale in one place closed check-in: e14d37468e user: dmestas tags: ci-steps-run-once | |
| 04:16 | Private-repository list reloads after five seconds so other nodes' visibility changes are seen check-in: 06d30f8721 user: dmestas tags: visibility-list-refreshes-on-a-ttl | |
| 04:14 | Run the race pipeline only on trunk and landing tips check-in: 8f1ae6a6aa user: dmestas tags: ci-race-on-landings | |
| 04:13 | land tag-once-per-repo-and-writer check-in: b2001ef3c5 user: dmestas tags: trunk | |
| 04:12 | Retried CI steps show attempt N of M and stop spinning once the verdict cannot change (#283) check-in: d338dd539a user: dmestas tags: ci-attempt-of-and-settled-verdict | |
| 04:01 | land ingest-queued-rebuild-honours-stop check-in: 1bb5bfa54d user: dmestas tags: trunk | |
| 04:00 | land ci-keep-every-attempt-log check-in: 650d6c85fc user: dmestas tags: trunk | |
| 04:00 | Trunk merged into tag-once-per-repo-and-writer, with the web test compile fix closed check-in: b0e524d0ab user: dmestas tags: tag-once-per-repo-and-writer | |
| 03:56 | CI steps run once; the runner retries only the checkout check-in: d386463602 user: dmestas tags: ci-steps-run-once | |
| 03:56 | Bring trunk's web test compile fix into the rebuild stop branch closed check-in: 1796c418bd user: dmestas tags: ingest-queued-rebuild-honours-stop | |
| 03:56 | Trunk merged into the ops history flush branch, taking trunk's copy of the shared op-timing test check-in: 84b6db5af3 user: dmestas tags: ops-history-flush-off-request-path | |
| 03:54 | Drained CI logs read past a failed lane and fsync the record before the rename closed check-in: ac78a08c6d user: dmestas tags: ci-keep-every-attempt-log | |
| 03:47 | The surface filter test lives in its own file, clear of the shared op-timing test check-in: d61ba606ba user: dmestas tags: ops-history-flush-off-request-path | |
| 03:46 | land handler-owned-page-templates check-in: 66d194b343 user: dmestas tags: trunk | |
| 03:43 | Flush ops history off the request path and keep web renders in their own recent window check-in: 3bdf2f891d user: dmestas tags: ops-history-flush-off-request-path | |
| 03:41 | A page that loses its hooks is journalled, and trunk is merged into the handler-owned page branch closed check-in: 9e18c78197 user: dmestas tags: handler-owned-page-templates | |
| 03:39 | Trunk merged into tag-once-per-repo-and-writer check-in: 18b79187d4 user: dmestas tags: tag-once-per-repo-and-writer | |
| 03:38 | Closes take turns per repository and carried verdicts are remembered per repository, with the store asked after a restart check-in: 7fd22cacb7 user: dmestas tags: tag-once-per-repo-and-writer | |
| 03:38 | merge trunk into handler-owned-page-templates check-in: 7ae575297f user: dmestas tags: handler-owned-page-templates | |
| 03:38 | land ci-race-step check-in: c68cc68d39 user: dmestas tags: trunk | |
| 03:34 | A stopped pipeline's queued rebuild is not logged as a failed first ingest, and Stop tolerates a pipeline built without NewPipeline check-in: 213dc0c5ab user: dmestas tags: ingest-queued-rebuild-honours-stop | |
| 03:32 | Drained CI logs survive a lane error and are written atomically; record-only view and drain rules are pinned by tests check-in: 30f439ed42 user: dmestas tags: ci-keep-every-attempt-log | |
| 03:26 | ingest: say why deleting the dead consumer on resubscribe is intended closed check-in: 969e393f5c user: dmestas tags: ingest-resubscribe-delete-comment | |
| 03:24 | The op-timing test passes the hooks argument withTimerFor now takes closed check-in: 7b6a3bde3e user: dmestas tags: web-tests-compile-again | |
| 03:24 | Select the six stop tests by exact name in the race target closed check-in: d54c5703c9 user: dmestas tags: ci-race-step | |
| 03:23 | web: pages are opaque handles drawn only through a handler's hooks; register ciRunPage and personEditPage check-in: b67646c5ba user: dmestas tags: handler-owned-page-templates | |
| 03:18 | ingest: a rebuild queued for a slot gives up when its pipeline stops check-in: b2aedcf60d user: dmestas tags: ingest-queued-rebuild-honours-stop | |
| 03:14 | land ci-race-step check-in: 14e778e6f1 user: dmestas tags: trunk | |
| 03:09 | ci: keep every attempt's log of a retried step and show each on the run page check-in: c782961ba5 user: dmestas tags: ci-keep-every-attempt-log | |
| 03:09 | The race target takes a COUNT for the slow packages, once by default check-in: 1f8de345fa user: dmestas tags: ci-race-step | |
| 03:08 | land tag-not-indexed-is-written check-in: 3a24fc9887 user: dmestas tags: trunk | |
| 03:08 | land ci-race-step check-in: 9b256af788 user: dmestas tags: trunk | |
| 03:08 | land ingest-failed-rebuild-rebuilds-search check-in: cc7bdc5ed3 user: dmestas tags: trunk | |
| 03:08 | land handler-owned-page-templates check-in: 4684d2bb0e user: dmestas tags: trunk | |
| 03:06 | checkin.close and the premerge carry-forward treat a tag written but not indexed as written and never add it twice closed check-in: 7b410efc19 user: dmestas tags: tag-not-indexed-is-written | |
| 03:02 | Keep the race step near three minutes: repeat only the cheap packages check-in: b3f17d04ee user: dmestas tags: ci-race-step | |
| 02:57 | Run the race detector as its own CI step and in make check check-in: fd593c91da user: dmestas tags: ci-race-step | |
| 02:53 | ingest: a failed rebuild also rebuilds search, so rows it folded can be updated closed check-in: 5fc9db224b user: dmestas tags: ingest-failed-rebuild-rebuilds-search | |
| 02:46 | land premerge-accept-rechecks-the-target check-in: 10ca664866 user: dmestas tags: trunk | |
| 02:46 | land web-rail-stream-test-race check-in: 5bb97692f3 user: dmestas tags: trunk | |
| 02:46 | land ingest-bound-concurrent-rebuilds check-in: 2bc5f8d576 user: dmestas tags: trunk | |
| 02:46 | land api-ops-are-timed-at-the-registry check-in: 26de248275 user: dmestas tags: trunk | |
| 02:39 | Time every op at the registry and file API calls beside web renders, tagged by surface closed check-in: ccbcf4b34e user: dmestas tags: api-ops-are-timed-at-the-registry | |
| 02:34 | land ingest-drop-resets-a-rebuild-covered check-in: 88e4f0e4b7 user: dmestas tags: trunk | |
| 02:34 | land ingest-resubscribe-after-stream-recreate check-in: a2cac610e2 user: dmestas tags: trunk | |
| 02:33 | land markdown-quote-fossil-extent check-in: da5b859333 user: dmestas tags: trunk | |
| 02:33 | land ingest-rebuild-resume-on-failure check-in: 6cbe042af2 user: dmestas tags: trunk | |
| 02:28 | ingest: a reset published before a rebuild no longer clears the rebuilt index closed check-in: 64eae210f6 user: dmestas tags: ingest-drop-resets-a-rebuild-covered | |
| 02:20 | markup: index fence runs so a quoted fence's closing lookup no longer scans to the end per quote closed check-in: 5916df2231 user: dmestas tags: markdown-quote-fossil-extent | |
| 02:11 | ingest: a fold whose consumer went with a recreated stream subscribes again closed check-in: ad165455fa user: dmestas tags: ingest-resubscribe-after-stream-recreate | |
| 02:07 | markup: give markdown quotes fossil's extent, so a quoted fence keeps its unmarked body check-in: dbd5881bcd user: dmestas tags: markdown-quote-fossil-extent | |
| 02:03 | ingest: a failed rebuild restores the search triggers and indexes it suspended closed check-in: e2e0d7965b user: dmestas tags: ingest-rebuild-resume-on-failure | |
|
2026-09-26
| ||
| 23:38 | land accept-expects-the-checked-tip check-in: a6e337766f user: dmestas tags: trunk | |
| 23:37 | land user-plan-test-reads-the-real-query check-in: 3b39b94022 user: dmestas tags: trunk | |
| 23:37 | land kinds-refresh-catches-up check-in: 966260d87e user: dmestas tags: trunk | |
| 22:23 | An accept whose zit version is unknown is refused before anything is written, and each accept probes zit again until it reads a good version closed check-in: afce4210f9 user: zeitgeist tags: accept-expects-the-checked-tip | |
| 22:17 | Accept tests pin the checked tip against an index that moved, and the auto-land sweep on an old zit logs once instead of every pass check-in: 1f591e20b9 user: zeitgeist tags: accept-expects-the-checked-tip | |
| 22:07 | An accept on a zit older than 0.3.31 is a 503 zit_too_old refusal naming both versions, and the version report and operations page show the gap check-in: 46b0914208 user: zeitgeist tags: accept-expects-the-checked-tip | |
| 21:57 | A landing accept tells zit the tip its checks judged, so a push since then is refused as branch_moved instead of merged unchecked check-in: c65b6e49b1 user: zeitgeist tags: accept-expects-the-checked-tip | |
| 21:56 | A closing index skips the kinds catch-up scan, so Close waits for one scan rather than two closed check-in: 9b913032de user: zeitgeist tags: kinds-refresh-catches-up | |
| 21:47 | The user timeline plan test explains the statement Timeline runs and names the index each variant may walk closed check-in: 152c0a5ddb user: zeitgeist tags: user-plan-test-reads-the-real-query | |
| 21:39 | A kinds refresh looks once more when it finishes, so the fold that landed during it does not leave the census a generation behind check-in: 40e40f3f00 user: zeitgeist tags: kinds-refresh-catches-up | |
| 21:36 | land user-timeline-walks-an-index check-in: 5574907f4e user: dmestas tags: trunk | |
| 21:36 | land zit-0.3.31 check-in: 5219570ca7 user: dmestas tags: trunk | |
| 21:36 | land composite-kind-timeline-merges check-in: a37e8361cd user: dmestas tags: trunk | |
| 21:32 | A person's timeline walks an index on (user, date, hash) and stops at the page instead of sorting everything they ever made closed check-in: 2d92dbeaa5 user: zeitgeist tags: user-timeline-walks-an-index | |
| 21:32 | Build with zit 0.3.31, which merges exactly the landing tip it read closed check-in: fbd7ac2047 user: zeitgeist tags: zit-0.3.31 | |
| 21:17 | A timeline chip that stands for several kinds walks each kind's index in date order and merges them, instead of sorting every matching event closed check-in: dba1f028a0 user: zeitgeist tags: composite-kind-timeline-merges | |
| 20:33 | land dashboard-kinds-stay-warm check-in: 153d9380dd user: dmestas tags: trunk | |
| 20:31 | Kinds census tests pin the generation rule: a fold during a scan stays stale and an older scan never overwrites a newer one; the boot warm runs after the zit reads closed check-in: 591092d027 user: zeitgeist tags: dashboard-kinds-stay-warm | |
| 20:14 | land a-missing-attachment-is-asked-for-once check-in: 8625678b5a user: dmestas tags: trunk | |
| 20:12 | The absent-class test is gofmt-clean, and artifact read errors name the hash rather than a flag closed check-in: 29bffbf5c3 user: dmestas tags: a-missing-attachment-is-asked-for-once | |
| 20:04 | The dashboard's kind counts are warmed at boot, survive a fold stale while one background refresh runs, and no longer hold the counts lock across the scan check-in: 39eead8fb9 user: zeitgeist tags: dashboard-kinds-stay-warm | |
| 20:01 | A blob is remembered missing only on zit's no-match answer, and ingest forgets the set whenever a content blob arrives without a fold check-in: 28fa9f63fa user: dmestas tags: a-missing-attachment-is-asked-for-once | |
| 19:57 | land a-missing-attachment-is-asked-for-once check-in: cabe2e3cd7 user: dmestas tags: trunk | |
| 19:46 | A mirror's missing attachment blob was asked for on every render; the store's answer is now kept until the next fold check-in: 7cebdf8681 user: dmestas tags: a-missing-attachment-is-asked-for-once | |
| 19:46 | land a-big-ticket-page-does-not-cost-a-second check-in: 20e4392eef user: dmestas tags: trunk | |
| 19:45 | land a-page-view-never-waits-for-the-writer check-in: 57caae7f6d user: dmestas tags: trunk | |
| 19:42 | Create new branch named "a-missing-attachment-is-asked-for-once" check-in: 00c8319a4c user: dmestas tags: a-missing-attachment-is-asked-for-once | |
| 19:34 | FinishBulk and Reset are tested to drop the tracker census memo closed check-in: 155e1c3ddd user: dmestas tags: a-big-ticket-page-does-not-cost-a-second | |
| 19:30 | A head memo batch is retried row by row only when a row failed, and the retries stop when the index closes closed check-in: daac6176ce user: dmestas tags: a-page-view-never-waits-for-the-writer | |
| 19:28 | land a-page-view-never-waits-for-the-writer check-in: 860e0d0bb1 user: dmestas tags: trunk | |
| 19:25 | The head memo flush stops when the index closes, a failed batch retries row by row, and the full-set and Reset-race cases are tested check-in: e7902bdd25 user: dmestas tags: a-page-view-never-waits-for-the-writer | |
| 19:24 | land a-big-ticket-page-does-not-cost-a-second check-in: c5ea47c7e1 user: dmestas tags: trunk | |
| 19:19 | Every ticket page scanned all of ticket_field twice, for the field names and the recorded numbers; both are now kept until the next fold check-in: 4c6ccdf7ce user: dmestas tags: a-big-ticket-page-does-not-cost-a-second | |
| 19:15 | land a-page-view-never-waits-for-the-writer check-in: 2c79fd6fc7 user: dmestas tags: trunk | |
| 19:15 | land the-forge-stops-in-order check-in: b93e6be77c user: dmestas tags: trunk | |
| 19:15 | land every-caps-write-is-guarded check-in: 6fa7bfd115 user: dmestas tags: trunk | |
| 19:11 | Create new branch named "a-big-ticket-page-does-not-cost-a-second" check-in: 6602bc2e46 user: dmestas tags: a-big-ticket-page-does-not-cost-a-second | |
| 19:09 | The attachment head memo was written through the index writer on the ticket page, so the page waited out every mirror fold; it is now remembered at once and flushed in the background check-in: 1af99a4829 user: dmestas tags: a-page-view-never-waits-for-the-writer | |
| 19:00 | A child's context is released as soon as the child has ended closed check-in: e2a4f12886 user: dmestas tags: the-forge-stops-in-order | |
| 19:00 | An empty claim token tombstones nothing, and the fossil-login link shows the guard's warning closed check-in: d827dea4f6 user: dmestas tags: every-caps-write-is-guarded | |
| 18:56 | land every-caps-write-is-guarded check-in: fcdce10acf user: dmestas tags: trunk | |