Description
Coverage data uploads consistently return HTTP 500 from Codecov's server, while test results uploads for the same commit, same repo, same auth method succeed. This happens on both the new ingest endpoint and the legacy V4 endpoint.
Environment
- codecov-action: v6.0.0 (
57e3a136b779b570ffcdbf80b3bdc90e7fab3de2)
- Codecov CLI: v11.2.8 (downloaded at runtime)
- OS: ubuntu-latest (GitHub Actions)
- Auth: OIDC (also tested with repository token — same result)
- Coverage format: LCOV (generated by Bun test runner)
- Repo:
archgate/cli (public)
Reproduction
Workflow config (simplified):
# Coverage upload — FAILS with 500
- uses: codecov/codecov-action@57e3a136b779b570ffcdbf80b3bdc90e7fab3de2
with:
use_oidc: true
slug: archgate/cli
files: coverage/lcov.info
disable_search: true
# Test results upload — SUCCEEDS
- uses: codecov/codecov-action@57e3a136b779b570ffcdbf80b3bdc90e7fab3de2
with:
use_oidc: true
slug: archgate/cli
files: coverage/junit.xml
disable_search: true
report_type: test_results
Verbose logs (from CI run)
Coverage upload (fails)
The CLI goes through upload-coverage which hits the ingest endpoint directly:
info -- Found 1 coverage files to report
debug -- Found 395 network files to report, (395 without filtering)
warning -- Response status code was 500. --- {"retry": 0}
warning -- Request failed. Retrying --- {"retry": 0}
warning -- Response status code was 500. --- {"retry": 1}
warning -- Request failed. Retrying --- {"retry": 1}
warning -- Response status code was 500. --- {"retry": 2}
warning -- Request failed. Retrying --- {"retry": 2}
Exception: Request failed after too many retries. URL: https://ingest.codecov.io/upload/github/archgate::::cli/upload-coverage
With use_legacy_upload_endpoint: true (also fails)
The legacy path successfully creates the commit and report, but the actual data upload still 500s:
debug -- Starting create commit process
info -- Commit creating complete
debug -- Commit creating result --- {"result": "RequestResult(error=None, warnings=[], status_code=202, ...)"}
debug -- Starting create report process
info -- Report creating complete
debug -- Report creating result --- {"result": "RequestResult(error=None, warnings=[], status_code=202, ...)"}
debug -- Selected uploader to use: <class 'codecov_cli.services.upload.legacy_upload_sender.LegacyUploadSender'>
warning -- Response status code was 500. --- {"retry": 0}
warning -- Response status code was 500. --- {"retry": 1}
warning -- Response status code was 500. --- {"retry": 2}
Exception: Request failed after too many retries. URL: https://codecov.io/upload/v4
Test results upload (succeeds — same commit, same auth)
debug -- Starting create commit process
info -- Commit creating complete (202)
debug -- Starting upload processing
debug -- Upload request to Codecov complete. --- {"response": {"raw_upload_location": "https://storage.googleapis.com/codecov-production/test_results/v1/raw/..."}}
info -- Sending upload (32700 bytes) to storage
info -- Upload queued for processing complete (200)
Key observation
The test results path uses UploadSender which gets a signed GCS URL and uploads directly to Google Cloud Storage — this works.
The coverage path uses either the streaming ingest endpoint or LegacyUploadSender — both return 500 from Codecov's server.
The create-commit and create-report API calls succeed (202) for both paths, so authentication and repo resolution are working. The 500 is specifically in the coverage data upload/processing step.
CI runs with full logs
Codecov UI message
After the failed upload, the Codecov commit page shows:
No coverage data is available due to incomplete uploads on the first attempt.
What we've tried
- ✅ Repository token (not global/org token)
- ✅ OIDC authentication
- ✅ Explicit
slug: archgate/cli
- ✅
disable_search: true
- ✅
use_legacy_upload_endpoint: true
- ✅ New commits (Codecov requires fresh commits after failed uploads)
- ❌ All produce the same 500 on coverage upload; test results always succeed
Description
Coverage data uploads consistently return HTTP 500 from Codecov's server, while test results uploads for the same commit, same repo, same auth method succeed. This happens on both the new ingest endpoint and the legacy V4 endpoint.
Environment
57e3a136b779b570ffcdbf80b3bdc90e7fab3de2)archgate/cli(public)Reproduction
Workflow config (simplified):
Verbose logs (from CI run)
Coverage upload (fails)
The CLI goes through
upload-coveragewhich hits the ingest endpoint directly:With
use_legacy_upload_endpoint: true(also fails)The legacy path successfully creates the commit and report, but the actual data upload still 500s:
Test results upload (succeeds — same commit, same auth)
Key observation
The test results path uses
UploadSenderwhich gets a signed GCS URL and uploads directly to Google Cloud Storage — this works.The coverage path uses either the streaming ingest endpoint or
LegacyUploadSender— both return 500 from Codecov's server.The
create-commitandcreate-reportAPI calls succeed (202) for both paths, so authentication and repo resolution are working. The 500 is specifically in the coverage data upload/processing step.CI runs with full logs
Codecov UI message
After the failed upload, the Codecov commit page shows:
What we've tried
slug: archgate/clidisable_search: trueuse_legacy_upload_endpoint: true