fix(website): the progress feed's requests allow 120 s, so a busy forge's pull request list no longer stops the deploy #2034
No reviewers
Labels
No labels
area/accessibility
area/audio
area/input
area/install
area/menu
area/renderer
area/saves
area/tooling
area/upscaling
beanstalk: approved
beanstalk: failed
beanstalk: needs approval
beanstalk: running
beanstalk: tested
ci: approved
good first issue
gpu/amd
gpu/igpu
gpu/intel
gpu/nvidia
help wanted
kind
bug
kind
chore
kind
docs
kind
feature
kind
performance
kind
question
needs measurements
platform/linux
platform/macos
platform/steam-deck
platform/windows
preview
priority
critical
priority
high
priority
low
regression
risk
high
risk
low
risk
medium
status
blocked
status
confirmed
status
duplicate
status
invalid
status
needs-info
status
wontfix
ci: approved
No milestone
No project
No assignees
2 participants
Notifications
Due date
No due date set.
Dependencies
No dependencies set
Reference
MagicBeanDev/Longshot!2034
Loadingβ¦
Reference in a new issue
No description provided.
Delete branch "HumanBean/Longshot:fix/website-feed-timeout"
Deleting a branch is permanent. Although the deleted branch may continue to exist for a short time before it actually gets removed, it CANNOT be undone in most cases. Continue?
Summary
The website has not been deploying. Its build collects the progress feed from the forge's API and gives each request
30 seconds. With the forge busy, listing 100 open pull requests takes 23 to 70 seconds (timed by hand on 2026-11-21),
so all three attempts time out, the feed comes out empty, and
npm run checkrefuses the deploy, as it should with anempty feed. Push deploys 45137 and
45251 failed this way; an hourly run that got a
faster answer deployed.
This gives each request 120 seconds (
REQUEST_TIMEOUT_MSinwebsite/src/lib/progress-request.ts) and says so inwebsite/PROGRESS.md. The retries, their waits and the check that refuses an empty feed are unchanged. The worst caseis three 120 s attempts on one request, well inside the job's 20 minutes.
Read
website/src/lib/progress-request.ts.Testing
Tested: game-free, nothing for the game to run. Timed the collector's own requests against git.magicbean.dev
(
/pulls?state=open&sort=recentupdate&limit=100: 23 s, and 28 to 70 s for smaller lists earlier); the failing runs'logs show three 30 s timeouts on
/pullseach. CI's website build and checks run on this PR; the existingwebsite/tests/progress-request.test.tschecks the retries and fresh timeout signals and does not depend on thevalue.
Still to test:
TimeoutErroron/pulls(by hand, from the run's log)Measurements
Nothing to measure: a build-time request timeout on the website, not the game.
Risks
A forge that is down, not slow, now takes up to six minutes per request to fail instead of 1.5, and the deploy still
fails as before. A revert is enough. Players notice nothing but the site updating again.
Checklist
Ready for review
main; the title is a Conventional Commit.mise run testpasses (any failure that also fails onmainis named). CI's run on this PR.main, or Measurements says why there is nothing to measure.third_party/untouched; nothing GPL-incompatible.When it applies
Assisted-by:/Co-Authored-By:trailers.Ready to merge
main, measured again or a comment says why the numbers still hold.π€ Generated with Claude Code
WIP: fix(website): the progress feed's requests allow 120 s, so a busy forge's pull request list no longer stops the deployto fix(website): the progress feed's requests allow 120 s, so a busy forge's pull request list no longer stops the deploy@review
The ChatGPT usage limit is reached (it resets Nov 25th, 2026 1:04 AM), so this review is paused. It runs by itself once the limit lifts (tried again every 30 minutes); nothing needs asking again.
Automated code review by BeanBot (gpt-6.1-terra), asked for by HumanBean. Nothing was run; a maintainer decides.
Recommendation: Request changes. The longer timeout covers the reported 23β70 second responses, but the fix has no regression test that would fail with the old 30-second limit. A successful collection with the changed timeout is also still unproven.
The website build gives each Forgejo progress-feed request 120 seconds instead of 30, allowing slower responses to complete. Failed requests can delay a build by up to six minutes per request.
@fixwill do (1)website/src/lib/progress-request.ts:20: There is no regression test for the longer request deadline: the existing tests pass with the old 30-second timeout because they inject immediate failures and never check when the signal expires. (more on the line)For the author (1) (a run, a measurement or a decision)
website/src/lib/progress-request.ts:9: The reported timings establish that 30 seconds is too short, but they do not show this commit completing the collector and publication check when the forge is busy. (more on the line)Checks: Still running:
CI / Website build and checks (pull_request).More: does it help, other systems, switches, what was read
Does it help: Unproven: the reported Forgejo requests took 23β70 seconds and failed runs logged three 30-second timeouts, but no successful collection or deployment with this commit is reported.
Other systems: The timeout applies to every progress-feed API request and response body, including sequential pagination and releases. The website deployment and PR preview use the same collector; the game and launcher do not. The 20-minute job limit also includes installation, other requests, checks and upload.
Switches: None.
Reviewed:
49ad8f5f8aagainste6196f0f54: Read AGENTS.md, CONTRIBUTING.md, docs/dev/standards.md, the PR template, both changed files, the commit, progress-request tests, the full collector, its JSON endpoint and the website workflow. Searched for other callers and timeout settings. docs/review-checklist.md and ../Tooling/AGENTS.md were absent. No recorded runs, translations or generated files were changed; CI results and PR-supplied run reports were reviewed.; nothing run (a code review).Checked and fine:
Testing: Read-only review; no code or tests were run. CI's website build and checks were still running in the supplied status.
Overlaps and merge order: Not checked; the network is closed.
Automated code review by BeanBot (gpt-6.1-terra), asked for by HumanBean. Nothing was run; a maintainer decides.
@ -6,1 +6,4 @@// A request's time, its JSON body included. A busy forge can take more than 30 s to list 100 open pull requests// (23 to 70 s on 2026-11-21), and an empty feed stops the deploy.const REQUEST_TIMEOUT_MS = 120000;L2 (minor): The reported timings establish that 30 seconds is too short, but they do not show this commit completing the collector and publication check when the forge is busy. Run the changed build under those conditions and record the result and duration before merging.
@ -14,3 +18,3 @@let retry = false;try {response = await fetcher(url, { headers, signal: AbortSignal.timeout(30000), redirect: 'error' });response = await fetcher(url, { headers, signal: AbortSignal.timeout(REQUEST_TIMEOUT_MS), redirect: 'error' });L1: There is no regression test for the longer request deadline: the existing tests pass with the old 30-second timeout because they inject immediate failures and never check when the signal expires. Add a game-free test that simulates a response taking more than 30 seconds but less than 120 seconds, including body reading, and fails if the old deadline is restored.