Repository navigation
fix(github): match pulled issues on owner/repo, not number alone - #7794
NahumLitvin wants to merge 1 commit into
Conversation
Pulling issue 42 from a second repository into the same beads database overwrote the bead already linked to issue 42 of the first repository. The pull engine falls back to matching by ExtractIdentifier, and the GitHub tracker returned the bare issue number for any URL, so refs from another repository collided with the configured one. Once initialized, ExtractIdentifier now resolves refs through RefScope, which relationship sync already uses, and returns "" for refs outside the configured repository. Assisted-by: Claude Code Agent-Signature: claude-code-claude-opus-5-5-medium on behalf of Nahum Litvin Signed-off-by: Nahum Litvin <14837912+NahumLitvin@users.noreply.github.com>
|
Triage: |
bee-ghosttrack
left a comment
There was a problem hiding this comment.
Reviewed at 83fd4db (pinned locally as refs/pr/7794-r990). Approving; the Bazel Farm is the merge gate, see the CI note.
What I verified
go test ./internal/github/ -race -count=1andgo test ./internal/tracker/ -short -count=1pass on the PR head.- Control: with
internal/github/tracker.goreset to the merge-base (9d6fff7) and the new test kept,TestExtractIdentifierScopedToRepositoryfails on exactly the two cross-repo rows (acme/repo-aandother/repo-bboth yield"42"). So the test pins the fix, not the surrounding behaviour. gofmt -lis clean;strconvwas already imported. The test lives in the existingtracker_test.go, so noBUILD.bazelsrcs change is needed.- main has moved only by the two dependency merges (#7646, #6355) since this branch forked; nothing under
internal/github/orinternal/tracker/changed.
Call-site read (internal/tracker/engine.go): every caller already treats "" as "skip this ref", so routing initialized trackers through RefScope.IssueNumberFromRef is safe at all seven sites. Besides the pull fallback map that #7781 is about (the localByExternalIdentifier build and the existing == nil lookup), the same change quietly fixes two siblings: the push-update path no longer FetchIssues the configured repo's issue of the same number for a bead linked elsewhere (it counts as skipped), and the pre-linked hydrate pass no longer does either. The uninitialized (client == nil) path keeps the old regex behaviour, which is what the pre-existing TestExtractIdentifier on a bare &Tracker{} relies on.
Two notes, neither blocking
- Behaviour change worth a line in the PR description or release notes: a bead whose ref points at the repo's previous owner/name (after a transfer or rename) will no longer match, since the scope compares owner/repo exactly (case-insensitively). That is the intended effect of the scope, but multi-repo users who rely on GitHub's redirect will notice.
- Selective pull by bead id (
bd github pull <bead-id>) for a bead linked to another repo now logs "No external ref found for local issue …" and skips. The ref exists, it just isn't this repo's; a follow-up could sharpen the wording, but skipping is the right outcome.
CI note: this is a first-time fork PR, so the Bazel Farm run shows skipping with the check suite in action_required until a maintainer approves the workflow run. The approve above rests on the local runs and the control; the Farm going green after run approval is the merge gate. The GitLab tracker's ExtractIdentifier has the same number-only shape, as the description says; an issue for it would keep that from getting lost.
|
thanks for the careful read. added the rename/transfer note to the description. I'll file the GitLab one as a separate issue. claude-code-claude-opus-5-5-unknown-reasoning on behalf of Nahum Litvin |
GitHub and GitLab fallback lookups now require the configured repository/project and origin, preserving distinct issue numbers across remotes and repeat-pull idempotency. Retain Nahum Litvin's gastownhall#7794 regressions and add numeric GitLab metadata/stale-path controls for gastownhall#7795. Pin the merged engine/analyzer forks and Go 1.27.3 in the root and independent examples, preserving driver 2.4.1 and prior customizations. Update the compiler-compatible lint pin and Nix dependency hash. Qualify engine/storage regressions, plan suites, real CLI pulls and isolated old/new native-backup round trips; keep strict conflicting-virtual-base refusal.
|
Adapted and verified this correction in our fork: BlueboxC@0429f95 . Thank you, Nahum Litvin; the five supplied scenarios are retained, with additional configured-origin, GitHub Enterprise API/base-path and malformed/foreign-reference controls. One integration detail worth preserving: provider-family recognition and identifier extraction serve different purposes. A foreign GitHub URL must still be recognized as a GitHub reference by push, while producing no fallback identifier for the current repository. Narrowing both checks to the current repository could instead make push create an unwanted issue. The GitHub package passes. A real CLI/embedded-Dolt fixture reproduces the lost second import before the fix and retains two separate issue-7 records afterward; repeating the second pull is idempotent. It also passes with our exact published and installed binary. No live GitHub issue writes were used for qualification, and the original shorthand behavior remains unchanged. BlueboxC |
|
This PR entered the Gas City checkpoint pipeline. We first verify the problem it solves on current |
An initialized GitHub tracker returned the bare issue number for any github.com issue URL, so a ref to another repository's issue addressed the configured repository's issue with the same number (#7781). ExtractIdentifier now resolves URLs through the tracker's RefScope (configured host and owner/repo, case-insensitive) and returns "" for anything else. The github:<number> shorthand keeps its unchanged arm, ahead of the scope check, and an uninitialized tracker keeps the legacy parse. TestExtractIdentifierScopedToRepository is the contributor's test from #7794, with its five cases kept and extended with REST, host-case, foreign-host, malformed-path, issue-zero, empty and shorthand cases plus a GitHub Enterprise subtest. AC: AC-5 Derived-From: pr/83fd4db238088def5cf66d17f5741830d46164ae, opus/0acc1e8ff66474a8d5b6e77c5a7b1002b4d67d72, glm/f9c901971b1e11d13b3a899ddbfe5d18c8f475b9 Co-authored-by: Nahum Litvin <14837912+NahumLitvin@users.noreply.github.com> Agent-Signature: claude-code-claude-opus-5-5-unknown-reasoning on behalf of Julian Knutsen Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
|
Merged as part of #8068, with credit. Thank you! |
…gastownhall#7794) (gastownhall#8068) * fix(github): pull creates a bead for another repository's same-numbered issue Engine regressions for gastownhall#7781: a bulk or selective pull of repository B's issue 1313 creates a new bead instead of falling back onto the bead linked to repository A's issue 1313, selecting A's bead by ID skips it without fetching B's issue, and incremental pre-link hydration no longer fetches B's issue into A's bead. Document the isolation and the relink step after a repository rename or transfer. AC: AC-1 Co-authored-by: NahumLitvin <NahumLitvin@users.noreply.github.com> Agent-Signature: claude-code-claude-opus-5-5-unknown-reasoning on behalf of Julian Knutsen Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> (cherry picked from commit 068dc3b) Ported-From: opus/068dc3b6ef61899122d6025e7b76885ab7b978bd AC: AC-1 * fix(github): scope ExtractIdentifier to the configured repository An initialized GitHub tracker returned the bare issue number for any github.com issue URL, so a ref to another repository's issue addressed the configured repository's issue with the same number (gastownhall#7781). ExtractIdentifier now resolves URLs through the tracker's RefScope (configured host and owner/repo, case-insensitive) and returns "" for anything else. The github:<number> shorthand keeps its unchanged arm, ahead of the scope check, and an uninitialized tracker keeps the legacy parse. TestExtractIdentifierScopedToRepository is the contributor's test from gastownhall#7794, with its five cases kept and extended with REST, host-case, foreign-host, malformed-path, issue-zero, empty and shorthand cases plus a GitHub Enterprise subtest. AC: AC-5 Derived-From: pr/83fd4db238088def5cf66d17f5741830d46164ae, opus/0acc1e8ff66474a8d5b6e77c5a7b1002b4d67d72, glm/f9c901971b1e11d13b3a899ddbfe5d18c8f475b9 Co-authored-by: Nahum Litvin <14837912+NahumLitvin@users.noreply.github.com> Agent-Signature: claude-code-claude-opus-5-5-unknown-reasoning on behalf of Julian Knutsen Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * test(github): same-repository respellings still match on pull A bead linked to repository B's issue through a case variant, the REST URL or the github:<number> shorthand is updated and canonicalized by a pull of B's canonical URL, in a bulk pull and a selective pull by number, while repository A's same-numbered bead (sorted last, so a bare-number index would pick it) is left alone and no duplicate is created. The canonical exact-ref case is the control that rejects a fix disabling all matching. AC: AC-2 Derived-From: opus/b9afcafdcf59f7b7ea8ae10cb643ebba1f399c67, glm/01b8a8ecee7d8104a34e917692955181d5eea056 Agent-Signature: claude-code-claude-opus-5-5-unknown-reasoning on behalf of Julian Knutsen Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * fix(tracker): dry-run push skips linked issues the tracker cannot address A real push skips a linked issue whose external_ref yields no identifier, but the --dry-run preview counted it as an update. Skip it in the preview too, so pushing to GitHub repository B reports repository A's same-numbered bead as skipped in both modes and sends no request for it. The GitHub regression covers, in a real run and a preview: A's bead alone (no request at all), A's bead beside an own-repository update and an unlinked create (exact GET/PATCH/POST in the real run, no request in the preview), and a "keep local" sync with no conflict to force; plus a real own-repository conflict on issue 99 whose forced push updates issue 99 and still skips A's bead without addressing issue 1313. AC: AC-3 Derived-From: opus/96c6b166be3842a25016768dc6b58b1296ad3913, glm/8c4e70cebc6644fbef3dea6be63cb06e0f248f27 Agent-Signature: claude-code-claude-opus-5-5-unknown-reasoning on behalf of Julian Knutsen Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * test(github): conflict detection ignores another repository's bead A bead linked to repository A and edited since the last sync is not compared against repository B's same-numbered issue, also edited since then: conflict detection sends no request at all, reports no conflict and writes nothing to A's bead. AC: AC-4 Derived-From: opus/c571c8d337782a71335f790a369283f2c71def88, glm/90499672c68e5f3ebee0f124e2cc67ff0b377f1e Agent-Signature: claude-code-claude-opus-5-5-unknown-reasoning on behalf of Julian Knutsen Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Co-authored-by: Nahum Litvin <14837912+NahumLitvin@users.noreply.github.com> * fix(github): keep syncing beads linked under a renamed repository's name A config that still names a renamed or transferred repository keeps working against GitHub, which serves the old name by redirect, but every issue it returns has an html_url under the current name, and pull links beads to those URLs. Scoping ExtractIdentifier to the configured owner/repo made the tracker refuse the refs it wrote itself: push and conflict detection skipped those beads without a warning, so the next remote edit overwrote an unpushed local edit. Init now reads the repository once (GET /repos/{owner}/{repo}, which follows the redirect), and RefScope also accepts its current full_name, for ExtractIdentifier and relationship sync alike. A failed lookup keeps the configured name as the only one. A pull that selects a bead by ID now says when the bead's ref is not an issue the tracker can address, instead of claiming it has no external ref. Agent-Signature: claude-code-claude-opus-5-5-unknown-reasoning on behalf of Julian Knutsen Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com> * fix(github): keep syncing beads linked under a repository's former name Scoping refs to the configured repository made a config that names a renamed or transferred repository's new name refuse the beads still linked under its old name: a pull imported their issues again as second beads, and push and conflict detection skipped their local edits without a warning. That is the config a user gets by updating github.owner and github.repo after the rename, and the one GITHUB_REPOSITORY gives in CI without any update. Main matched those beads by issue number. Init now also reads each other repository that a local ref names on the configured host, and accepts its name when GitHub resolves it to the configured repository's ID: GitHub redirects a former name to the repository. Those beads keep syncing, and a pull relinks them to the current URL. Another repository answers with its own ID, a deleted or inaccessible one not at all, and both stay foreign. Names are checked against GitHub's character set and path-escaped before the lookup, and a failed lookup adds nothing. The same lookup covers a name the repository had between two renames, which the config-name/current-name pair missed. README and CHANGELOG drop the relink-before-pull step for the normal case and keep it for a failed lookup or a reclaimed old name. Agent-Signature: claude-code-claude-opus-5-5-max on behalf of Julian Knutsen Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com> --------- Co-authored-by: Julian Knutsen <julianknutsen@users.noreply.github.com> Co-authored-by: Nahum Litvin <14837912+NahumLitvin@users.noreply.github.com> Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com> Co-authored-by: gascity-eddie[bot] <338826946+gascity-eddie[bot]@users.noreply.github.com>
…gastownhall#7794) (gastownhall#8068) GitHub pull matched a pulled issue to a local bead by issue number alone, so pulling issue 1313 from a second repository overwrote the bead linked to the first repository's issue 1313 (gastownhall#7781). An initialized tracker's ExtractIdentifier now resolves issue URLs through RefScope: a URL yields a number only when it names the configured host and repository (or another name GitHub resolves to it), and any other URL yields an empty identifier, which sync leaves alone. The github:<number> shorthand and the uninitialized parse are unchanged. Push, conflict detection, selective pull and pre-link hydration address issues through that identifier, so they no longer reach another repository's same-numbered issue either, and --dry-run push reports those beads as skipped, as a real push does. A pull that selects a bead by ID now says when the bead's ref is not an issue the tracker can address. Renamed or transferred repositories keep syncing: Init reads the configured repository (GET /repos/{owner}/{repo}, following redirects) to learn its current full_name, and reads each other repository that a local ref names on the configured host, accepting that name only when GitHub answers with the configured repository's ID (a former name redirects to it; another repository answers with its own ID; a deleted one 404s). Fixes gastownhall#7781 [hotfix/1.3.2 backport] Adapted for v1.3.x, which lacks the GitHub relationship sync from gastownhall#5971 (internal/github/links.go, LinkResolver, review_fixes_test.go). Instead of patching links.go, the minimal prerequisite was lifted from main's links.go into a new internal/github/refscope.go: the RefScope type, NewRefScope, hostFromURL, IssueNumberFromRef, splitIssueURLPath, plus gastownhall#8068's withNames/otherRepository/issueURL/namesRepository. The sub-issue and blocked-by link code, DependencyLink, PushLinks and the legacy SubIssueLinkFromParentChild/BlockedByLinkFromBeadsDependency wrappers are omitted, and the RefScope doc comment was reworded to describe issue sync rather than the relationship endpoints. Tracker.RefScope() is added here (on main it came with gastownhall#5971); the LinkResolver scope wiring is omitted because LinkResolver does not exist on v1.3.x. In engine.go the selective pull helper and the dry-run guard are the same code; the dry-run guard sits directly after the create branch because v1.3.x has no --create-only preview branch (gastownhall#6337). Tests: tracker_cross_repo_test.go and the tracker_test.go hunk are taken verbatim except that TestInit_ResolvesRenamedRepositoryName checks tr.RefScope() instead of tr.LinkResolver().Scope(); the RefScope tests (including gastownhall#5971's TestRefScopeRejectsForeignHostAndRepository) live in refscope_test.go instead of review_fixes_test.go. The README "GitHub Issues Sync" section is included; CHANGELOG.md and BUILD.bazel changes are dropped. (cherry picked from commit 0e7b758) Co-authored-by: Nahum Litvin <14837912+NahumLitvin@users.noreply.github.com> Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Backport tracker, CLI, MCP and storage corrections from original main 126b30a, retaining the v1.3.1 API, Continuity features, qualified engine tuple and Graph Preview11/generation13. Preserve local endpoint/credential authority, unknown-write non-replay, native restore rehearsal, history and stronger fork admission/log bounds. Update Go build/compiler and compatible compression/Python locks together. Add contributor regressions and preservation controls; align public contracts. Adapted source commits (original attribution retained by reference): 9fdd8f6 Julian Knutsen - test(tracker): add real UOW proxy conformance fixture (gastownhall#6445) 8446d09 Harry Miller - fix(tracker): honor --create-only in push --dry-run preview (gastownhall#6337) (gastownhall#6712) 721e42e Wilhelm Uschtrin - fix(notion): make the query page bound configurable and honour Retry-After (gastownhall#6006) 22e9178 gascity-eddie[bot] - fix(tracker): hydrate pre-linked issues from overlap-only bulk rows (supersedes gastownhall#6807) (gastownhall#8005) 0e7b758 gascity-eddie[bot] - fix(github): scope issue sync to the configured repository (supersedes gastownhall#7794) (gastownhall#8068) 0a004ca gascity-eddie[bot] - test(tracker): pin the pulls that must leave the push hash alone (supersedes gastownhall#7755) (gastownhall#8004) 5579171 gascity-eddie[bot] - fix(tracker): gastownhall#8004 follow-up — a pull keeps a local label edit while GitHub's labels are unchanged (gastownhall#8409) 8373677 Wenjie F. - fix(security): stop option injection via sync remotes and beads-mcp argv (gastownhall#7881) 5a3b720 gascity-eddie[bot] - fix(mcp): report reopen of an issue that is not closed (gastownhall#7927) fb09d9f gascityinc-randy[bot] - fix(beads-mcp): bump pyjwt to 2.15.1 and urllib3 to 2.8.0 for security advisories (gastownhall#7466) 6442502 Julian Knutsen - fix(beads-mcp): bump joserfc to 1.7.7/1.8.0 for GHSA-jpxq-7h8p-2q76 and GHSA-9pg5-8c7x-5hfx (gastownhall#8115) 5a6da89 Julian Knutsen - fix(setup): confine recipes.toml recipe paths to bd setup's directory (gastownhall#7901) 37f9475 Julian Knutsen - fix(doctor): import issues.jsonl only into a database the repair created or found with no issues (gastownhall#7899) 882467e gascityinc-randy[bot] - fix(migrate-personal,dolt): refuse half-moves and make dry-run delete match the real one (gastownhall#7983) 7b80f4c gascity-eddie[bot] - test(cmd/bd): route bd -C <non-git project> by the selected project (gastownhall#7930) 7e465cb gascity-eddie[bot] - fix(dep): let --quiet silence the unresolved-target warning (gastownhall#7941) e4cd352 gascity-eddie[bot] - test(store_reopen): pin explicit --db spellings onto the symlinked workspace (GH#4627) (gastownhall#7949) 3fc493d gascity-eddie[bot] - fix(cli): name the skipped user-global ~/.beads in workspace refusals (gastownhall#7950) 3477429 gascity-eddie[bot] - fix(formula): order nested loops' deps and reject a step needing its own loop (gastownhall#7636) (gastownhall#7963) 2a00540 Takumi Koshii - fix(init): persist the stealth no-git-ops flag in the project config (gastownhall#7939) bedc36d gascity-eddie[bot] - fix(doctor): dial a probe-found server's remotesapi at its configured host (supersedes gastownhall#7724) (gastownhall#8002) 00adff8 gascity-eddie[bot] - fix(rules): compact previews unless --force and backs up before deleting (gastownhall#8051) 8ffd4e5 gascity-eddie[bot] - fix(doctor): stop offering doctor --fix as an unopenable-store rebuild (GH#4521) (gastownhall#8052) 78089d4 Wenjie F. - fix(security): block executable settings from .beads/.env and workspace API (gastownhall#7879) ba58464 gascity-eddie[bot] - test(hooks): pin the lock preflight on the commit's index (gastownhall#7956) 217ded3 gascity-infra-agent[bot] - fix(security): take .beads/.env values literally and refuse command and credential keys by name (gastownhall#8137) f1756b4 gascity-eddie[bot] - fix(doctor): untrack dolt-server-config.yaml and dolt-server.activity (gastownhall#6191) (gastownhall#7856) f57f987 gascity-eddie[bot] - test(cook): condition primitive's dropped leaf still cooks (supersedes gastownhall#6642) (gastownhall#8008) c938c0a gascity-eddie[bot] - fix(cli): skip store init for store-free config get keys (gastownhall#7905) f386202 gascity-eddie[bot] - fix(hooks,doctor): refuse symlinked hooks dirs and confine doctor's destructive database paths (supersedes gastownhall#8070) (gastownhall#8077) 12f387d Julian Knutsen - fix(config): replace config.yaml in one step instead of rewriting it in place (gastownhall#8199) 3a068db Julian Knutsen - fix(telemetry): keep the raw command line off the OTel resource (gastownhall#8222) 2edea9c Julian Knutsen - fix(telemetry): key the bd.storage.* metrics by db.operation alone (gastownhall#8223) df63f0b Julian Knutsen - test(doctor): spell Choice 2's sqlite path from the canonical workspace (gastownhall#8226) 473aa09 Julian Knutsen - fix(telemetry): pool gauges follow open stores, console exporters to stderr, fractional ms (gastownhall#8224) 125c620 gascity-eddie[bot] - fix(doctor): stop flagging cross-repository dependency targets as orphans (supersedes gastownhall#8154) (gastownhall#8400) 126b30a gascity-eddie[bot] - fix(telemetry): keep issue content out of bd.args (gastownhall#8505) (gastownhall#8508) c7b818e gascity-eddie[bot] - fix(dolt): redact URL credentials from transfer errors (supersedes gastownhall#7644) (gastownhall#8089) 8977ea1 gascity-eddie[bot] - test(bd): pin the reworded dolt start help in the command-tree golden (gastownhall#7904) 062fe1e gascity-eddie[bot] - docs(dolt): show the positional remote in push/pull usage and help (gastownhall#7937) 7833e4e gascity-eddie[bot] - docs(config): state no-git-ops scope; document dolt.local-only (GH#3463) (gastownhall#7908) 6839936 Julian Knutsen - fix(proxy): fail fast with a clear MySQL error when an external upstream is unreachable (gastownhall#7181) c9a36b2 Julian Knutsen - fix(dbproxy): wait out a live starter's spawn marker instead of a fixed 5s (gastownhall#7249) 2f7699c gascity-eddie[bot] - fix(storage): stop the close-gate walk reading a closed or pinned parent's stale bit (gastownhall#7924) c153e3b Julian Knutsen - fix(dolt): require proof the server serves the workspace's Dolt directory; reset-data keeps remote data (gastownhall#7902) f7b9465 gascityinc-randy[bot] - fix(config): bound setting key and value lengths in the shared role rule (gastownhall#8014) 7f8046b gascity-eddie[bot] - fix(versioncontrolops): drain auto-resolve CALLs on the CLI-repair session (gastownhall#7960) 0275ae7 gascity-eddie[bot] - fix(sql): refuse a read that calls a Dolt function that may write under --readonly (gastownhall#7957) cbfdafc gascity-eddie[bot] - fix(versioncontrolops): fail closed on unparseable BackupRestore URLs (gastownhall#7988) 7423018 Rongjun GENG - fix(uow): stop outage backoff from spinning at CPU rate after the outage window (gastownhall#7196) 5b58d9e gascity-eddie[bot] - test(dolt): pin the complete-read GetMoleculeProgress rollup (supersedes gastownhall#7055) (gastownhall#8009) f85a4a0 gascity-eddie[bot] - fix(issueops): check Rows.Err in DeleteIssuesBySourceRepoInTx (supersedes gastownhall#7054) (gastownhall#8007) a138433 gascity-eddie[bot] - fix(compact): refuse out-of-order date partitions and verify the replay (gastownhall#6645) (gastownhall#8050) 194b894 Julian Knutsen - fix(storage): generated ids are free in both planes and within a create batch (gastownhall#8057) 6f9b540 Julian Knutsen - fix(dbproxy): Shutdown waits for a start's spawn marker without competing for its lock (gastownhall#8056) 9311c24 gascity-eddie[bot] - fix(schema): keep issue and wisp updated_at across frozen migrations (gastownhall#5981) (gastownhall#8053) 93ecb95 Julian Knutsen - perf(schema): probe columns with SHOW COLUMNS in the updated_at migration guard (gastownhall#8080) d95de62 gascity-eddie[bot] - fix(delete): cascade follows only parent-child edges (GH#6301) (gastownhall#8078) 3c205e0 gascity-eddie[bot] - fix(schema): suspend dolt_transaction_commit for the migration pass (gastownhall#4470) (gastownhall#8055) 6d713a7 Julian Knutsen - test(schema): cover the schema-27 upgrade under dolt_transaction_commit (gastownhall#4137) (gastownhall#8136) f3fef33 gascity-eddie[bot] - test(dolt): settleMergeInTx carries cleanly merged tables (GH#7811) (gastownhall#7944) e4c51a2 gascity-eddie[bot] - fix(dolt): verify project identity for Path-only server-mode opens (gastownhall#7947) 710f712 gascity-eddie[bot] - fix(storage): verify Compact's rebuilt chain before swapping main (supersedes gastownhall#7296) (gastownhall#8011) d7bcee6 gascity-eddie[bot] - fix(dolt): apply metadata socket on config opens (supersedes gastownhall#7475) (gastownhall#8082) ca4d2d8 Rongjun GENG - fix(backup): validate manifest before destructive restore (gastownhall#7192) ab5b72c gascity-eddie[bot] - fix(schema): let pending ignored migrations run over dirty ignored tables (gastownhall#5816) (gastownhall#8245) 8bb4e33 gascity-eddie[bot] - fix(flatten): soft-reset main so clone-local FKs and pending writes survive (bd-c5s3mv) (supersedes gastownhall#6772) (gastownhall#8175) e0e1553 gascity-eddie[bot] - fix(uow): refuse a database whose schema is ahead of the binary on proxied and bd serve opens (supersedes gastownhall#8121) (gastownhall#8388) 3282c1e gascity-eddie[bot] - fix(schema): add the missing wisps columns before migration 0053 (supersedes gastownhall#8099) (gastownhall#8389) 035ac7b gascity-eddie[bot] - fix(schema): recognize Dolt 2.6 drift text in the aux row re-key (gastownhall#8395) 5261031 gascity-eddie[bot] - fix(backup): refuse a restore a Dolt server on another host would read unchecked (gastownhall#7192 follow-up) (gastownhall#8413) 44f45ed gascity-eddie[bot] - build(security): build with Go 1.27.4 for GO-2026-6859 (net/http) (gastownhall#8431) BlueboxC
The entry said the fix "supersedes #7791", which reads as if rjgeng's work was replaced. #8098 ported it: d76ade2 carries their Co-authored-by trailer and keeps #7791's stress harness. Use the file's wording for a ported contributor fix, as the #8191 and #7794 entries do: "(#7780, reported by @ganniterix; first fixed by @rjgeng in #7791)". Agent-Signature: claude-code-claude-opus-5-5-unknown-reasoning on behalf of Julian Knutsen Co-Authored-By: Claude <noreply@anthropic.com>
Closes #7781
I pull my issues from several repos into one beads db. Pulling issue 1313 from a second repo said "0 created, 1 updated" and silently overwrote the bead linked to issue 1313 of the first repo.
When the exact URL lookup misses, the pull engine falls back to
ExtractIdentifier, which returned the bare number for any repo's URL. Once initialized it now goes through the existingRefScope.IssueNumberFromRefand returns "" for refs from other repos.TestExtractIdentifierScopedToRepositoryfails without the fix and passes with it. I ran both in my fork's Actions, since my network can't fetch go-mysql-server:I haven't re-run the original two-repo repro with a patched binary. GitLab's
ExtractIdentifierhas the same shape; left for a separate change.Behavior change: a ref that still points at a repo's old owner/name after a transfer or rename no longer matches.
Drafted with Claude Code, reviewed by me.