ci: fail loudly when pnpm install exhausts its retries #480

Merged

gmackie merged 1 commit from ci/fail-loud-on-install-failure into main

2026-08-27 00:40:15 +00:00

gmackie commented

2026-08-26 22:22:09 +00:00

Owner

Both pnpm install --frozen-lockfile retry loops passed the step even when all three attempts failed. A for loop's exit status is its last command — here the trailing sleep 10 — so a total install failure reported success and the next step ran against an empty node_modules.

The symptom is a lie

#340 has failed three times with:

sh: 1: storybook: not found

which says nothing about the cause. Storybook builds fine locally on both that branch and main, and pnpm install --frozen-lockfile succeeds on both. So the install was failing in CI for network reasons — exactly what the retry loop exists for — and then hiding it behind a downstream error that sent me looking at storybook, lockfiles, and workspace globs instead.

The fix is already in this file

The gitleaks download above tracks an ok flag and exits 1 when retries are exhausted, with a comment that puts it better than I can:

the one outcome worse than a noisy gate is one that skips quietly

The deploy health checks guard themselves too. Only these two loops did not.

Verified

Ran the shapes directly: the old form exits 0 when every attempt fails, the new form exits 1, and still exits 0 when the first attempt succeeds.

Both `pnpm install --frozen-lockfile` retry loops **passed the step even when all three attempts failed**. A `for` loop's exit status is its last command — here the trailing `sleep 10` — so a total install failure reported success and the next step ran against an empty `node_modules`. ## The symptom is a lie #340 has failed three times with: ``` sh: 1: storybook: not found ``` which says nothing about the cause. Storybook builds fine locally on **both** that branch and main, and `pnpm install --frozen-lockfile` succeeds on both. So the install was failing in CI for network reasons — exactly what the retry loop exists for — and then hiding it behind a downstream error that sent me looking at storybook, lockfiles, and workspace globs instead. ## The fix is already in this file The gitleaks download above tracks an `ok` flag and exits 1 when retries are exhausted, with a comment that puts it better than I can: > the one outcome worse than a noisy gate is one that skips quietly The deploy health checks guard themselves too. Only these two loops did not. ## Verified Ran the shapes directly: the old form exits **0** when every attempt fails, the new form exits **1**, and still exits 0 when the first attempt succeeds.

gmackie added 1 commit

2026-08-26 22:22:09 +00:00

ci: fail loudly when pnpm install exhausts its retries

CI / gitleaks (pull_request) Successful in 6s

Details

CI / storybook (pull_request) Successful in 1m39s

Details

forgegraph/ci CI passed

CI / ci (pull_request) Successful in 9m19s

Details

8e12714ff5

Both `pnpm install --frozen-lockfile` retry loops passed the step even when
all three attempts failed. A `for` loop's exit status is its last command —
here the trailing `sleep 10` — so a total install failure reported success
and the next step ran against an empty node_modules.

The symptom is a lie. #340 has failed three times with

    sh: 1: storybook: not found

which says nothing about the cause. Storybook builds fine locally on both
that branch and main, and `pnpm install --frozen-lockfile` succeeds on both,
so the install was failing in CI for network reasons — exactly what the retry
loop exists for — and then hiding it behind a downstream error.

The correct pattern is already in this file: the gitleaks download above
tracks an `ok` flag and exits 1 when the retries are exhausted, with a
comment that puts it better than I can — "the one outcome worse than a noisy
gate is one that skips quietly." The deploy health checks guard themselves
too. Only these two loops did not.

Verified the shapes directly: the old form exits 0 when every attempt fails,
the new form exits 1, and still exits 0 when the first attempt succeeds.

gmackie merged commit 2b7fd917aa into main

2026-08-27 00:40:15 +00:00

gmackie referenced this pull request from a commit

2026-08-27 00:40:16 +00:00

ci: fail loudly when pnpm install exhausts its retries (#480)

gmackie referenced this pull request

2026-08-27 00:56:49 +00:00

feat(web): web run reports — consume @preflight/runreport #340

gmackie referenced this pull request

2026-08-27 01:07:18 +00:00

docs(field-reports): the self-hosted runners never restore the pnpm cache #481

gmackie referenced this pull request from a commit