Skip to content

validate-may21-prs: cross-OS CI for May 18-21 PR cohort#137

Open
danielhanchen wants to merge 7 commits into
mainfrom
validate-may21-prs
Open

validate-may21-prs: cross-OS CI for May 18-21 PR cohort#137
danielhanchen wants to merge 7 commits into
mainfrom
validate-may21-prs

Conversation

@danielhanchen

Copy link
Copy Markdown
Owner

Summary

  • Trims 22 heavy / redundant workflows so each push stays well below the 5-concurrent Windows runner cap.
  • Adds a single targeted validate-may21-prs.yml that runs the touched backend tests for the May 18-21 2026 PR cohort across ubuntu-latest, macos-14, windows-latest.
  • All tests are CPU-only under the existing tests/conftest.py CUDA spoof, so they finish on vanilla GitHub runners without GPUs.

Coverage

PR Test surface
unslothai#5603 sandbox hardening studio/backend/tests/test_sandbox_hardening.py
unslothai#5582 MTP --spec-draft-n-max studio/backend/tests/test_llama_cpp_mtp_detection.py, studio/backend/tests/test_gguf_reload_inheritance.py
unslothai#5604 lockfile audit tests/security/test_lockfile_supply_chain_audit.py + python3 scripts/lockfile_supply_chain_audit.py

Test plan

  • ubuntu-latest job green
  • macos-14 job green
  • windows-latest job green
  • No Windows runner contention (cancel-in-progress confirmed in concurrency group)

Trims 22 heavy/redundant workflows and adds a single targeted
validate-may21-prs.yml that runs the touched backend tests for the
May 18-21 2026 PR cohort across ubuntu-latest, macos-14, windows-latest.

Covers:
  PR unslothai#5603 sandbox hardening      -> test_sandbox_hardening.py
  PR unslothai#5582 MTP --spec-draft-n-max -> test_llama_cpp_mtp_detection.py
                                     test_gguf_reload_inheritance.py
  PR unslothai#5604 lockfile audit         -> test_lockfile_supply_chain_audit.py
                                     + the audit script itself

All tests are CPU-only under the conftest CUDA spoof. Workflow is
push-gated on the validate-may21-prs branch with cancel-in-progress
to stay below the 5-concurrent Windows runner cap.
@gemini-code-assist

Copy link
Copy Markdown

Note

Gemini is unable to generate a review for this pull request due to the file types involved not being currently supported.

Daniel Han and others added 5 commits May 21, 2026 13:06
Each PR ran the same staged source files before, which went stale when
the upstream PR commits advanced. Refactor to one job per PR with an
actions/checkout of that PR's head ref, so cross-OS validation
always uses the latest commit:

  - PR unslothai#5603 sandbox            -> studio-sandbox-hardening
  - PR unslothai#5620 parser parity      -> studio-tools-multi-format-v2
  - PR unslothai#5696 mtp reload guards  -> followup-mtp-reload-guards (unslothai#5582 followup)
  - PR unslothai#5695 lockfile audit     -> followup-lockfile-audit-regressions (unslothai#5604 followup)

4 jobs x 3 OSes = 12 runs; Windows = 4 (below the 5-concurrent cap).
cancel-in-progress per (workflow, ref) keeps iteration cheap.

All tests stay CPU-only and rely on the CUDA spoof harness in
tests/conftest.py + tests/_zoo_aggressive_cuda_spoof.py, so no real GPU
is required on any runner.
unslothai#5620 parser tests transitively import the safetensors loop, which
needs the datasets package. unslothai#5696 route-guard tests import
routes/inference.py, which transitively imports core/training (uses
matplotlib). Add both, plus the auth deps (pyjwt/cryptography/
aiosqlite/python-multipart) needed for any test that touches the
FastAPI route module so route-level imports resolve cleanly on all
three OSes.
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.

2 participants