mirror of
https://github.com/QwenLM/qwen-code.git
synced 2026-11-07 06:19:43 -10:00
main
8
Commits
| Author | SHA1 | Message | Date | |
|---|---|---|---|---|
 易良
|
32b1f12f75 | fix(release): bound summary concurrency and refresh release PRs (#15500) | ||
|
|
d5c22d336b |
ci(sdk-java): guard Flyway version uniqueness and watch red main runs (#12965)
* ci(sdk-java): guard Flyway version uniqueness and watch red main runs
Two in-flight PRs each added a V16 managed-agent-server migration and
merged 23 minutes apart; each branch was green in isolation, and Flyway
refused to boot the merged result, leaving main red for two hours while
three more merges stacked on it (#12940).
A new flyway-migrations job runs scripts/check-flyway-migrations.js on
every SDK Java pull request and push: a filename scan of the two
migration locations that share one version sequence, failing in seconds
and naming both files when two claim one version. Versions compare the
way Flyway compares them — numerically per segment — so V016 collides
with V16, and a location rename that would empty the guard fails instead
of passing vacuously.
The main-CI failure watcher now watches the SDK Java workflow, whose
path-filtered push run on main is the only signal for a merge result
neither PR could fail, and the failure-signature helper reads Surefire's
per-test error lines, so a red Java lane files one deduped issue naming
the failing tests instead of a per-commit stub.
Refs #12940
* ci(sdk-java): harden the Flyway guard and dedupe its main-red issues
Review on #12965:
- compare versions across every module of one invocation — Flyway resolves
classpath:db/migration across jars, so a version claimed in two modules
collides exactly like two files in one module
- fail when the SQL location yields no migration even if the Java location
still does, so renaming one location away can no longer pass vacuously
- escape %/CR/LF in module names and file paths before emitting ::error::
(the runner parses workflow commands from stderr too, and the job scans
contributor-controlled filenames on pull_request)
- match the migration suffix case-insensitively, as Flyway does
- keep the fully qualified Surefire id as the dedupe identity so same-named
tests in different packages no longer merge into one issue; the cosmetic
package strip moves to shortenForTitle
- give the guard's ::error:: line a stable flyway:<module>:<version>
identity, so a standing red dedupes to one issue instead of one per
stacked merge, and render the failed jobs in the test-keyed body
- pin the workflow_run watch entry to sdk-java.yml's name key
Co-authored-by: Qwen-Coder <[email protected]>
* ci(sdk-java): fix the Flyway dedupe on runner-rendered logs and close review findings
The Actions runner consumes ::error:: workflow commands and writes
##[error] into the downloadable job log, so the dedupe pattern anchored
on ^::error:: never matched production input and every merge stacked on
a standing red would have opened a fresh autofix issue. The anchor now
accepts both forms, the guard's two non-collision failure modes (moved
migration location, missing module directory) get their own stable
module+mode identity, and the synthesized flyway id is written to be
read since it doubles as the issue title and body bullet.
The issue body's "- Failed jobs:" block is now machine-owned like
"## Also failing": stripped and re-emitted from the current run on every
merge instead of freezing the lanes recorded at filing time, with the
empty-list case pinned against a dangling heading.
The guard itself now reports a multi-module collision once after all
modules are scanned, naming every claimant (a second line would file a
second issue), never prints a clean summary for a claimant, and only
requires the SQL location of a module that owns a db/migration*
directory — which lets the workflow invocation cover runtime-broker, the
module that shares the classpath:db/migration namespace without carrying
migrations today.
escapeWorkflowCommand moves to scripts/release-script-utils.js (the
byte-identical copy would have drifted from the tested original), and
the filename-forgery tests gain a runner-decoding oracle and a
capability probe so the Windows lane skips instead of failing on
filenames Win32 cannot hold.
Co-authored-by: Qwen-Coder <[email protected]>
* ci(sdk-java): close the guard's blind spots and validate its dedupe captures
Review and sandboxed verification on #12965 found the moved-location
probe keying on the two exact directory names: a rename outside the
migration* family (db/migration -> db/changelog) passed silently with
the whole sequence inside it, a renamed Java location printed "all
versions unique" from a scan that skipped it, and a Java-only module
red-lit as "moved" with its real collision never named. The probe is
now derived from the LOCATIONS table: a versioned-migration file
anywhere under a source root but outside the configured location fails
with the files named, the sibling check is per location and fires only
when the location directory itself is gone, and a repeated module
argument no longer fabricates a collision of a file with itself.
The invocation now names qwencode too — managed-agent-server depends on
it at compile scope, the same namespace premise that admitted
runtime-broker — and the .size-baseline entry is re-recorded at the
file's current size.
On the consumer side, both Flyway patterns feed their captures through a
module-shape validation before they become a dedupe identity, so a
filename-borne forgery rendered by the runner's annotation decode can no
longer inject a second identity into the issue this files, and the
guard's run-level diagnosis is ordered ahead of first-seen test ids so a
mass failure in another lane cannot crowd it past the search and body
caps. A closure pin keeps the spawned guard importable in the
dependency-free helper lane.
Refs #12940
* ci(sdk-java): close the Windows-drive forgery and split the stray-file message
The drive-prefix alternative in the module validator re-opened the
forgery it closed: backslash and colon are ordinary characters in a
POSIX filename, so a migration name could supply a `C:\evil` prefix and
pass the path-shape check. No lane that runs this consumer ever
produces a backslashed module — the producer and analyze jobs are
ubuntu-latest and the suite runs only on the Linux lint_and_static
lane — so the branch is dropped, backslashes are rejected outright, and
the fixtures hand the guard a slash-bearing argument (a no-op on POSIX)
instead of teaching the consumer a shape only a forgery can produce.
The unsafe-character half of the validator is now pinned for
whitespace, backticks and backslashes across both call sites.
A populated location with a migration-shaped stray elsewhere under the
source root no longer prints "found no migration under": the guard
names the stray as the cause, and the consumer's config identity learns
the new phrase, so the standing red still files one issue. The dep-free
closure pin now walks the guard's relative-import closure instead of a
hardcoded depth-one list, so a new hop carrying an npm import cannot
slip through.
Refs #12940
* ci(sdk-java): re-record the sdk-java.yml size baseline after merging main
The rebase onto
|
||
|
|
90c166585a |
feat(release): user-facing bilingual digest for release notes (#9216)
* feat(release): user-facing bilingual digest for release notes Stable release notes read as a type-bucketed PR list, which users find hard to scan. The finalize step now asks the model to group changes into user-facing themes with short intros, mirrors highlights and themes into a Chinese digest, attaches screenshots found in merged PR bodies (host allowlist, per-release cap), and collapses the full PR list into an appendix with normalized titles. Every model failure path keeps today's v1 output byte-for-byte, and CHANGELOG.md accepts the new v2 marker. * fix(release): tighten v2 digest fallbacks and changelog skeleton (#9216) Address review round 1 findings: - usedAi only counts themes that carry content, so a release whose digest has zero model text is no longer reported as AI-generated - hasChinese is derived from what the Chinese block actually renders, not raw model output, so zh-only-on-breaking releases no longer emit an empty or English-only section - a PR repeated inside one theme is deduped instead of discarding the whole themes digest with a misleading cross-theme error - fallback titles in the v2 digest are normalized like the appendix, killing the mixed-style look in the degradation case - normalizeAppendixTitle strips only the conventional types the changelog's formatEntry strips, keeping ci/test/security prefixes - the changelog unwraps the v2 appendix at the same sibling rank as v1's Complete Change List instead of nesting it under the previous section - drop a dead summaries max_tokens scaling term and a verbatim copy of renderChangeLine's attribution rendering * fix(release): close digest image breakout and tighten fallback signals (#9216) Co-authored-by: Qwen-Coder <[email protected]> * fix(release): drop camo proxy and harden digest text validation (#9216) * fix(release): neutralize markdown breakouts in digest text and images (#9216) * fix(release): close classification, image-URL, and text-validation bypasses (#9216) --------- Co-authored-by: qwen-code-dev-bot <[email protected]> Co-authored-by: Qwen-Coder <[email protected]> Co-authored-by: qwen-code-dev-bot <[email protected]> |
||
|
易良
|
ff5c639e26 |
fix(release): raise model timeouts and shrink batch size for slow networks (#8007)
* fix(release): raise model timeouts and shrink batch size for slow networks The AI release-notes generator timed out on every batch in the v0.21.1 finalize run on a GitHub-hosted runner (US, Azure westus3) hitting a remote LLM endpoint. The 60s per-request timeout was too close to the edge: two successful requests took 54.2s and 56.7s, and every other batch hit the 60s abort. The 12-minute total budget was consumed by retries on those timeouts, and the circuit breaker opened after 3 consecutive failures, skipping highlights entirely. - timeoutMs: 60s -> 180s — give the model enough headroom to generate a full JSON response over a high-RTT cross-region connection. - totalTimeoutMs: 12min -> 30min — ~140 PRs at 8 per batch is ~18 batches; at ~90s each that's ~27 minutes of model time. - batchSize: 12 -> 8 — fewer entries per prompt means the model generates less per request, returning faster and reducing the chance of a single slow request dragging the whole batch. - workflow timeout-minutes: 15 -> 35 — match the new 30min budget plus a margin for git/npm setup. No changes to retry logic, circuit breaker, or prompt shape. * fix(release): sync timeout workflow test * test(release): assert step timeout exceeds script budget The workflow `timeout-minutes` and the script's `totalTimeoutMs` are defined in separate files with no shared constant. If they drift apart silently and the budget exceeds the step timeout, the runner SIGKILLs the step and even the fallback release notes are lost. Add a cross-file assertion so any future change to either value that breaks the invariant fails the test immediately. Suggested-by: wenshao in PR review. * test(release): parse release notes timeout expression |
||
|
易良
|
bfd4c8e519 | fix(scripts): slim release-note model prompts and log request timing (#7941) | ||
|
易良
|
923e5ab425 |
fix(scripts): harden retry classification and preserve deadline error context (#7854)
* fix(scripts): harden retry classification and preserve deadline error context Follow-up to #7535 addressing remaining review suggestions: - isRetryableModelError: content-validation errors ('Model response did not contain message content.') are now non-retryable. These are deterministic failures from our own code — retrying the same prompt reproduces the same failure. Network-level errors (no HTTP status) remain retryable. - Deadline-expired errors now include the original error message instead of the generic 'budget exhausted', so oncall can see the root cause. - Added 4 tests: HTTP 429 retry, network error retry, content-validation non-retry, deadline error preservation. * fix(release): preserve last model error across retry budget * test(release): cover retry budget expiry after backoff * refactor(release): extract deadline error helper and preserve cause Consolidate the three identical budget-exhausted throw sites into a single deadlineError() helper so they share one message format and source variable, and attach the last model error as the Error cause so its stack trace survives into CI logs. Also conform the release-notes tests added by this series to Prettier. |
||
|
|
e5df5b22e3 |
fix(scripts): retry model calls and surface degraded release notes (#7535)
* fix(scripts): retry model calls and surface degraded release notes The stable-release notes generator fell back to pull-request titles on any model error and the finalize workflow stayed green with no visible degradation. Four recent stable releases lost every summary batch and the highlights call to 60s timeouts without anyone noticing. The completer now retries each request (default 2 retries, exponential backoff with jitter) on timeouts, network errors, 429, and 5xx, while 4xx fails immediately. After three consecutive failed summary batches a circuit breaker stops paying the timeout for the rest and skips the highlights call, since the model side is down rather than slow. Fallback warnings are emitted as ::warning:: workflow commands so they render as run annotations, and a degraded-run block is appended to GITHUB_STEP_SUMMARY when any fallback occurred, distinguishing fully degraded (titles only) from partially degraded runs. Refs #7523 Signed-off-by: Yufeng He <[email protected]> * ci: retrigger after flaky fork-dispatch test failure * fix(scripts): create the step-summary parent dir before appending The ubuntu CI shard of this PR failed the summary test with ENOENT on readFileSync after appendFileSync: the append reached the runner fine, but the parent of the summary path can be missing there. Create it recursively first, which is what the function's contract (the note must land in the summary file) already implies. * fix(scripts): assert the degraded summary through the fs mock test-setup.ts no-op mocks appendFileSync for the whole scripts project, so the new test's real-file read could never see the write under test:scripts. Assert on the mock call instead. * fix(scripts): keep summary-write failures from costing the notes and make the reset test real appendDegradedStepSummary now runs through tryAppendDegradedStepSummary so a filesystem error on the auxiliary summary cannot kill the release before the notes are written, and the circuit-breaker recovery test's mock returns valid summary payloads so the reset path is actually exercised. * fix(scripts): escape model-controlled text in workflow commands Warnings can carry model output (parse errors, PR-derived fields) into ::warning:: commands. Percent-encode %/CR/LF so a forged ::error:: line cannot emit a second runner command, and collapse newlines in the step-summary markdown list so a multi-line warning does not break the list. Closes the P1 from PR review. * fix(scripts): bound release-note AI generation time * fix(scripts): log model request retries to stderr Retry attempts in createOpenAiCompleter were silent, so an oncall engineer could not distinguish slow model responses from exponential- backoff retries in the CI log. Emit a one-line stderr log on each retry with the attempt count, the triggering error, and the backoff delay. Plain stderr (not ::warning::) so transient retries that recover do not become run annotations. * fix(scripts): escape retry log workflow commands --------- Signed-off-by: Yufeng He <[email protected]> Co-authored-by: 易良 <[email protected]> Co-authored-by: yiliang114 <[email protected]> |
||
|
易良
|
ae9e7782d4 |
feat(release): generate AI-assisted release notes (#6756)
* feat(release): generate AI-assisted release notes * docs(release): remove AI release notes planning docs * fix(release): preserve complete AI release notes * refactor(release): trim AI notes generator surface * ci: Open autofix issues for main CI failures * fix(release): harden AI notes review feedback * fix(release): close AI notes review gaps |