IGNORE -- crash from spr#2492
Closed
jucor wants to merge 3 commits into
Closed
Conversation
This was referenced Mar 30, 2026
jucor
force-pushed
the
spr/edge/0c448343
branch
2 times, most recently
from
9264716 to
c51da68
Compare
jucor
force-pushed
the
spr/edge/d39cf65d
branch
from
59482f0 to
e811512
Compare
When spr force-pushes all stack branches at once, each branch triggers a new E2E run while the previous one is still queued. The concurrency group (keyed on branch ref) cancels the stale run, cutting the number of concurrent E2E jobs roughly in half. commit-id:1a8eb714
## Summary Add `SKIP_GOLDEN=1` environment variable to disable golden snapshot regression tests. During stacked PR development, golden snapshots become stale as computation changes cascade through the stack. Rather than re-recording snapshots at every rebase (which causes conflict cascades in jj/git), we skip them until the stack is merged into `edge`. ### Changes - **`test_regression.py`**: Add `@_skip_golden` decorator to `test_conversation_regression` and `test_conversation_stages_individually` — the only two tests that compare against golden snapshots. Other dataset-using tests (Clojure comparison, smoke tests) are unaffected. - **`python-ci.yml`**: Set `SKIP_GOLDEN=1` in CI so the stacked PRs don't fail on stale snapshots. ### Usage ```bash SKIP_GOLDEN=1 pytest tests/ # skip golden snapshot tests pytest tests/ # run everything (default) ``` ## Test plan - [x] `SKIP_GOLDEN=1 pytest tests/test_regression.py -v`: 4 skipped, 5 passed - [x] `pytest tests/test_regression.py -v`: all 9 collected (golden tests run normally) commit-id:d39cf65d
## Summary Replace `pip install` with `uv pip install` in the delphi Dockerfile for faster dependency installation. `uv pip` is a drop-in replacement — same `requirements.lock`, same `pyproject.toml`, same installed packages in `site-packages`. ### Changes - **`delphi/Dockerfile`**: - Copy `uv` binary from `ghcr.io/astral-sh/uv:0.11.2` (pinned version, single static binary) - Place in `/opt/uv/` in builder to avoid leaking into production image via `COPY --from=builder /usr/local/bin` - Set `UV_SYSTEM_PYTHON=1` (install into system Python, not a venv) - Replace all `pip install` with `uv pip install` in builder and test stages - Update BuildKit cache mount targets from `/root/.cache/pip` to `/root/.cache/uv` - Test stage copies `uv` from builder (single source of truth) ### What's NOT changed - **Makefile** — untouched, `make rebuild-delphi` works as before - **docker-compose.yml / docker-compose.test.yml** — untouched - **pyproject.toml / requirements.lock** — untouched, same format - **pip-compile workflow** — untouched, still used for lock file generation - **Final/production image** — no `uv` added, stays lean ### CI Benchmark (GitHub Actions, `ubuntu-latest`, `--no-cache`) 5 pip runs (Mar 27 stack push) vs 3 uv pip runs (this PR): | Step | pip (n=5) | uv pip (n=3) | Speedup | |------|-----------|--------------|---------| | **Docker build** | **264s** (sd=6) | **169s** (sd=2) | **1.56x (-94s)** | | Pytest run | 223s (sd=4) | 227s (sd=4) | ~same | **~94 seconds saved per CI run** on the Docker build step. Pytest runtime is unchanged (same packages, same tests). Low variance in both groups confirms this is a real improvement, not noise. ### Local Benchmark (Apple M1 Max, 64GB, Docker Desktop, `--no-cache`) | Step | pip | uv pip | Speedup | |------|-----|--------|---------| | **Dependencies install** | **149.3s** | **80.4s** | **1.9x** | | Dev deps install | 10.0s | 2.2s | **4.5x** | ## Test plan - [x] `docker compose -f docker-compose.test.yml build --no-cache delphi` succeeds locally - [x] Built image has all expected packages (`pip show` diagnostic passes in build log) - [x] CI passes (3 successful runs) - [x] `make rebuild-delphi` works (Makefile untouched) commit-id:0c448343
jucor
force-pushed
the
spr/edge/0c448343
branch
from
c51da68 to
e4353af
Compare
Collaborator
Author
|
PR created by a crashed |
Delphi Coverage Report
|
jucor
changed the title
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
Replace
pip installwithuv pip installin the delphi Dockerfile for faster dependency installation.uv pipis a drop-in replacement — samerequirements.lock, samepyproject.toml, same installed packages insite-packages.Changes
delphi/Dockerfile:uvbinary fromghcr.io/astral-sh/uv:0.11.2(pinned version, single static binary)/opt/uv/in builder to avoid leaking into production image viaCOPY --from=builder /usr/local/binUV_SYSTEM_PYTHON=1(install into system Python, not a venv)pip installwithuv pip installin builder and test stages/root/.cache/pipto/root/.cache/uvuvfrom builder (single source of truth)What's NOT changed
make rebuild-delphiworks as beforeuvadded, stays leanCI Benchmark (GitHub Actions,
ubuntu-latest,--no-cache)5 pip runs (Mar 27 stack push) vs 3 uv pip runs (this PR):
~94 seconds saved per CI run on the Docker build step. Pytest runtime is unchanged (same packages, same tests). Low variance in both groups confirms this is a real improvement, not noise.
Local Benchmark (Apple M1 Max, 64GB, Docker Desktop,
--no-cache)Test plan
docker compose -f docker-compose.test.yml build --no-cache delphisucceeds locallypip showdiagnostic passes in build log)make rebuild-delphiworks (Makefile untouched)commit-id:0c448343
Stack: