Repository navigation
sampleplatform CI ("CI - linux" / "CI - windows" checks) is broken across all open PRs — stale baseline + infra failures #2305
Description
Problem
The CI - linux and CI - windows checks (powered by sampleplatform.ccextractor.org) are failing on essentially every open PR right now, regardless of what the PR actually changes.
Evidence
Checked 6 currently-open PRs (#2294, #2298, #2301, #2302, #2304, and one more) — none touching related code to each other and all show CI - linux and/or CI - windows as failing:
| PR | CI - linux | CI - windows |
|---|---|---|
| #2301 (EIA-608 encoding) | fail "Not all tests completed successfully" | fail same |
| #2302 (libpng ARM/CMake) | fail same | fail same |
| #2298 (EIA-608 CC3/CC4) | fail "VM creation failed (ZONE_RESOURCE_POOL_EXHAUSTED_WITH_DETAILS). Please contact the administrator." | fail "Not all tests completed successfully" |
| #2294 (Rust network framing) | fail "Could not complete test, time limit exceeded" | pending |
| #2304 (DVB subtitle fix, mine) | fail same as #2301/#2302 | fail/pending |
Three distinct failure modes visible:
- *Stale comparison baseline: the "compared to test for commit ..." reference used across multiple different PR runs is 2feb09a (dated **2026-04-24), while current master HEAD is ba4a839 (dated **2026-07-26) a *3-month gap. Sample outputs/expected results have likely drifted in that window, producing failures unrelated to any given PR's actual diff (e.g. failures in Broken, CEA-708, General, Options, Teletext, WTV, XDS categories none of which relate to PR Fix silent success return for unhandled DVB string-coded subtitle obj… #2304's DVB-subtitle-only change).
- Cloud VM provisioning failures: ZONE_RESOURCE_POOL_EXHAUSTED_WITH_DETAILS (PR [FIX] --out report: probe both EIA-608 fields so CC3/CC4 are reported accurately (#2177) #2298) the underlying cloud runner pool appears to be exhausted.
- Timeouts: "time limit exceeded" (PR fix(rust): harden network block framing #2294).
Suggested fix direction
- Update/automate the baseline comparison commit so it tracks current master instead of a fixed, aging reference.
- Investigate the sampleplatform cloud runner pool exhaustion / timeout issues (may be capacity/quota related).
Activity
pantha704 commentedon Oct 14, 2026
Hi! I'd like to take this one. I've done CI upkeep before (disk cleanup work in rustc_codegen_cranelift), so this is my kind of task. I'll dig in and post questions or a PR as I go.
pantha704 commentedon Oct 15, 2026
Follow-up after digging in: the 'CI - linux' / 'CI - windows' checks are emitted by the sampleplatform GitHub App (check-runs, not this repo's Actions workflows). The failures (GCP operation timeout / incomplete test run) point at the platform side, not repo code. I'll step off this issue and continue with repo-level work (picking up #2367). If there's a repo-side angle I'm missing, happy to jump back in.