Describe your environment

OS: GitHub Actions (ubuntu-latest)
Python version: N/A (CI infrastructure issue)
Package version: N/A

What happened?

The Core Contrib Test CI check triggered from opentelemetry-python PRs fails intermittently with:

failed to find branch, tag, or commit `<SHA>`
fatal: ambiguous argument '<SHA>^0': unknown revision or path not in the working tree.

Re-running the job fixes it. This happens ~5–20% of the time on fresh commits.

Steps to Reproduce

1. Open or push a commit to a PR in open-telemetry/opentelemetry-python
2. The Core Contrib Test workflow is triggered with CORE_REPO_SHA set to the new commit SHA
3. Observe the job failing immediately on uv pip install 'opentelemetry-api @ git+https://github.com/open-telemetry/opentelemetry-python.git@<SHA>...'
4. Re-run the job a minute later — it passes

Example failure: open-telemetry/opentelemetry-python#4898

Expected Result

CI passes deterministically on the first run.

Actual Result

Flaky failure requiring a manual re-run. The failure is not related to the code under test.

Root Cause

Each Core Contrib Test job performs two separate git operations against the core repo:

Step A — actions/checkout@v4 checks out core at path: opentelemetry-python using GitHub's internal API. This always works.

Step B — tox installs core packages via a fresh raw git clone:

uv pip install 'opentelemetry-api @ git+https://github.com/open-telemetry/opentelemetry-python.git@<SHA>#egg=opentelemetry-api&subdirectory=opentelemetry-api'

Step B hits GitHub's public git CDN directly. On fresh commits, the SHA hasn't propagated to all CDN nodes yet, causing the failure. Step A succeeds because it uses a different (internal) path.

Proposed Fix

Eliminate step B by pointing tox at the already-checked-out local copy from step A instead of doing a second network clone.

The fix involves:

1. Adding per-package env vars to tox.ini (CORE_REPO_API, CORE_REPO_SDK, etc.) with fallbacks to the existing git URL behavior — fully backward compatible for local dev
2. Adding a CI step in the Jinja2 workflow template that sets those vars to local paths under $GITHUB_WORKSPACE/opentelemetry-python/
3. Regenerating core_contrib_test_0.yml via tox -e generate-workflows

Both package @ /abs/path with tox-uv and nested tox substitution ({env:VAR:{env:OTHER}\#fragment}) have been validated locally.

Additional context

No existing issues or PRs track this in either repo. The fix is entirely in opentelemetry-python-contrib — no changes needed in opentelemetry-python.

Would you like to implement a fix?

Yes