Commit Graph
7 Commits
Author SHA1 Message Date
OrcaWinandBrennan Benson e109e78ebf fix(source-control): keep huge change sets responsive (#9477)
* fix(source-control): keep huge change sets responsive

* Fix cancellation and retry handling for capped status

* Harden capped status for conflict-heavy repositories

* Harden capped status recovery and cancellation

* fix(source-control): preserve capped status correctness

* fix(source-control): translate submodule status at render time

---------

Co-authored-by: Brennan Benson <79079362+brennanb2025@users.noreply.github.com>
2026-07-20 14:50:12 -07:00
Jinjing f1d2fe5c65 perf(git-status): duty-cycle, cancel, and cache status polling to cut idle git load (#8922)
* feat(git-status): batch, cancel, and cache git status polling to cut idl

- Add a single duty-cycled refresh scheduler (activity debounce + 60s
  safety timer) replacing multiple overlapping intervals, so status
  polling no longer runs near-continuously on large repos (#7983).
- Let safety refreshes reuse cached numstat line counts instead of
  re-running diff --numstat every cycle, invalidated by head change,
  known mutations, and a bounded TTL.
- Thread AbortSignal/request-token cancellation through IPC, RPC, and
  relay layers so a superseded or backgrounded git:status call is
  killed instead of finishing wastefully.
- Fix automatic upstream/status apply ordering so a slow, older
  refresh can no longer clobber a newer result, and so an earlier
  refresh still applies when a later one fails.

* Fix aborted git status scans being mistaken for completed empty results

- An aborted scan/numstat pass now always rejects instead of silently
  resolving, so a cancelled request can't look like a valid empty status.
- Stop clearing the line-stats cache key on abort, since an aborted pass
  never wrote to it — clearing was evicting a concurrent scan's healthy
  snapshot and forcing a redundant numstat recompute.

* Fix aborted git status scans resolving as completed results

Cancelled scans could still resolve with partial or stale data instead of
rejecting, letting callers treat an aborted refresh as a valid status. Also
stop counting aborted scan duration toward catch-up refresh pacing, which
was stretching the next refresh interval by the full length of a cancelled
(often slow) scan.

* Add cancellable, generation-aware git status polling to cut stale scans

- Route git.status through an abortable subscription per requestToken so
  cancelStatus can actually abort the remote scan instead of being a
  no-op, preventing wasted work and stale responses overwriting fresher
  state.
- Bump the git status polling generation on push-target changes so an
  in-flight refresh against the old remote/branch can't apply stale
  upstream data to the new one.
- Guard the stale-conflict poller against writes after unmount.
- Retire pre-purge line-stat scans in the cache so an older in-flight
  scan can't repopulate a key after a token-scoped purge.
2026-07-15 19:09:14 -07:00
NeilandOrca 46646d7ff1 chore(lint): upgrade oxlint to 1.71 + enable 7 new rules (autofixed backlog) (#6841)
* chore(lint): upgrade oxlint to 1.71 and enable 7 new rules

Upgrade oxlint 1.67.0 -> 1.71.0 (1.72 was blocked by the repo's 3-day
minimum-release-age supply-chain guard; nothing here needs it). The
bump is a no-op on the existing config.

Enable 3 error rules (backlog autofixed to zero in this commit) and
4 warn rules (surface signal without gating CI):

error (autofixed, behavior-preserving):
- unicorn/prefer-node-protocol        (~1531 sites: bare builtin -> node:)
- typescript/no-import-type-side-effects (~36: all-inline-type -> import type)
- unicorn/no-array-reverse            (19: copy-then-reverse -> toReversed)

warn (real signal, current fires are test-only/correct):
- unicorn/no-array-fill-with-reference-type  (aliasing footgun guard)
- typescript/no-unsafe-function-type         (bans bare Function type)
- unicorn/prefer-array-flat-map              (map().flat() -> flatMap())
- unicorn/prefer-regexp-test                 (.match() in bool ctx -> .test())

mobile/.oxlintrc.json extends root, so it inherits all 7; the autofix
ran from root and covered mobile/ too.

Verification (all green): oxlint 0 errors (root+mobile+aux configs),
oxfmt clean, typecheck (node+cli+web), vitest 22795 passed / 0 failed,
builds (electron-vite + web + cli) succeed. node: rewrites confirmed to
skip embedded SSH/CLI string payloads (AST-only); all toReversed sites
verified to operate on fresh copies or write-once locals.

* chore(lint): bump mobile oxlint to 1.71 so inherited rules parse

mobile/ is a standalone pnpm project pinning its own oxlint@1.67, which
lacks unicorn/no-array-fill-with-reference-type (needs >=1.70). Since
mobile/.oxlintrc.json extends the root config, mobile CI's 'cd mobile &&
oxlint' failed to parse the new rule. Bump mobile to match root (1.71).

Verified in mobile/: oxlint 0 errors, oxfmt --check clean, tsc --noEmit
pass, vitest 978 passed / 0 failed.

Co-authored-by: Orca <help@stably.ai>

---------

Co-authored-by: Orca <help@stably.ai>
2026-06-29 22:38:29 -07:00
Brennan BensonandOrca a2fd8c11c2 Reduce source-control background load during refreshes (#6189)
Co-authored-by: Orca <help@stably.ai>
2026-06-23 15:06:38 -07:00
Jinwoo Hong fe8eb45ca2 Add relay upstream negative cache (#6105) 2026-06-23 00:33:52 -04:00
Jinwoo HongandOrca 414ae73838 Prevent main-process crash on huge git status output (#5371)
* Prevent main-process crash on huge git status output

git status on a repo with an enormous un-ignored folder could emit a
listing large enough that buffering it into one string exceeded V8's
max string length, throwing an uncatchable RangeError inside Node's
execFile exit handler and crashing the main process (and blocking
mobile). The runner also forwarded an undefined maxBuffer straight to
execFile, which silently disabled the cap entirely.

- runner: apply a non-bypassable default maxBuffer so no git call can
  buffer without a bound.
- status: stream porcelain-v2 output through an incremental parser and
  kill git the moment the changed-entry count crosses a limit
  (default 10,000); skip numstat/upstream probe when limited. Apply the
  same limit to the relay/SSH status path.
- UX: surface a 'too many changes' state, pause auto status polling
  while a repo is over the limit, and offer to .gitignore a detected
  huge folder.

Co-authored-by: Orca <help@stably.ai>

* Address review: validate inputs, fix stream decoding, localize copy

- huge-folder-ignore: restrict folderName to the known-folder allowlist
  before writing .gitignore (reject path-sep/newline injection).
- gitStreamStdout: decode stdout/stderr with StringDecoder so multibyte
  UTF-8 split across chunk boundaries isn't corrupted; flush on cleanup.
- gitStreamStdout: wrap the onStdout callback in try/catch so a throwing
  parser rejects the promise instead of crashing the main process.
- status/relay: validate the status limit as a non-negative integer.
- SourceControl: localize the 'too many changes' banner and the
  .gitignore warning toast via translate() with placeholder params.

Co-authored-by: Orca <help@stably.ai>

---------

Co-authored-by: Orca <help@stably.ai>
2026-06-14 13:32:05 -07:00
Neil 8f6f59a654 perf: handle large relay status entry lists (#3773) 2026-05-30 09:44:45 -07:00