Repository navigation
Check the output budget when a CAD worker exits before polling - #25
Draft
bhupendrafire-ai wants to merge 1 commit into
Draft
bhupendrafire-ai wants to merge 1 commit into
bhupendrafire-ai wants to merge 1 commit into
Conversation
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
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
What doesn't work
A CAD worker can write more than
output_limitand finish before the first polling interval. Both process-wait paths return its exit code without checking the completed log, so a successful fast exit bypasses the configured output budget.The regression waits for a short-lived worker to complete before the polling loop observes it. With 32 output bytes and a 16-byte budget, the original implementation returns success instead of raising the output-limit error.
The change
Extract the existing log-size check and also run it after process completion on Windows and POSIX. Cancellation and deadline checks keep their existing timing; the completion check only enforces the output budget.
Tests
python -m unittest tests.test_cad_process tests.test_worker_failure tests.test_cad_strategyWindows: 20 tests ran with no failures (one Linux-only skip). The added regression covers over-budget output, an exact-budget result, and
output_limit=None. The over-budget case failed before the fix. POSIX execution was not available on this machine.The documented Windows test setup also needed
keyringfor the existing test-vault guard; that setup correction is separate from this PR.