Skip to content

Stop the boot when an app the boot publishes fails (#219) - #220

Merged
StefanMaron merged 2 commits into
masterfrom
fail-boot-on-publish-failure-219
Nov 7, 2026
Merged

StefanMaron merged 2 commits into
masterfrom
fail-boot-on-publish-failure-219

Conversation

@StefanMaron

Copy link
Copy Markdown
Collaborator

Fixes #219.

Without libSkiaSharp every dev-endpoint publish failed on BC 29.0, also for apps with no report. The boot only echoed each HTTP code, wrote /tmp/bc-ready, and the healthcheck passed on a tier with no Test Runner, Library Assert or Any.

Changes

  • The keep-set install, the default test toolkit publish and the Test Runner Extension publish now record a failure, and print the response body. A 422 that says the app is already deployed still counts as success, using the same rule as publish-app.sh.
  • If any of them failed, the entrypoint lists them, stops NST and exits with rc 1. The publish code runs in a background subshell, so it marks the failure and stops NST, and the supervisor loop exits the container. wait-for-bc-healthy.sh and docker compose up --wait then fail right away, not after the 30-minute healthcheck.
  • BC_ALLOW_PUBLISH_FAILURES=1 finishes the boot anyway.
  • The SkiaSharp warning now says app publishing can fail too.
  • /tmp/bc-ready is removed at the start of the entrypoint, so a docker restart cannot pass the healthcheck early on the previous run's marker.

Testing

Image built from this branch, local boots:

Boot Result
BC 27.5, normal healthy, all 8 toolkit apps and the Test Runner Extension HTTP 200
BC 29.0, normal healthy, all 7 toolkit apps and the Test Runner Extension HTTP 200
BC 29.0, SkiaSharp native removed, NuGet unreachable the SkiaSharp error from the issue, then the container exited with rc 1 after 91 s, listing the 8 apps that failed
BC 27.5 (SkiaSharp 3.119.0), SkiaSharp native removed, NuGet unreachable healthy, every app published

The last row is why the warning says "can fail": the reported failure is on BC 29.0 with SkiaSharp 4.152.1, and BC 27.5 with SkiaSharp 3.119.0 published fine without the native. CI covers the keep-set path on every version.

https://claude.ai/code/session_01B5GPNVNEoh5CSGzgTPEthD

Without libSkiaSharp every dev-endpoint publish fails, also for apps
with no report. The test toolkit publish only echoed the HTTP code, so
the boot wrote /tmp/bc-ready and the healthcheck passed on a tier with
no Test Runner, Library Assert or Any.

The keep-set install, the default toolkit publish and the Test Runner
Extension publish now record failures, with the response body. If any
failed, the entrypoint stops NST and the container exits with rc 1.
BC_ALLOW_PUBLISH_FAILURES=1 continues instead. The SkiaSharp warning now
says that publishing fails too.

Claude-Session: https://claude.ai/code/session_01B5GPNVNEoh5CSGzgTPEthD
@StefanMaron
StefanMaron merged commit 3321941 into master Nov 7, 2026
22 checks passed
@StefanMaron
StefanMaron deleted the fail-boot-on-publish-failure-219 branch November 7, 2026 20:52
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

Without libSkiaSharp, every app publish fails, and the container still goes healthy

1 participant