Repository navigation
test: take real-clock bounds out of the MAF test hosts - #223
Merged
Merged
Conversation
Row6 (open-run bound) now fires its bound from a manual clock after the attempt is captured; the shared five-line host, the semantic-retrieval host and the verifier, lessons and holdout hosts run on a clock whose timers never fire, so a loaded machine cannot time out a retrieval or an indexing call that would have finished. SingleCallParity's one test that needs a real bound opts back into the system clock. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01GPXz56QPQAzDSyJTxbQTTy
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Three flaky tests, each a real-clock bound racing work on a loaded CI runner:
ExperienceCaptureTests.Row6_...: a real 150 msMaxOpenRunDurationclosed the run before its attempt was recorded. Now a manual clock fires the bound after the attempt is captured.AgentExperienceFiveLineSetupTests.A_faulted_model_call_...: one wiring hit the real 500 ms retrieval timeout. The shared host now defaults to a clock whose timers never fire (a test that sets its own clock still wins).AgentExperienceSemanticRetrievalTests.A_generator_that_throws_...: same host pattern, same default.The same default is applied to three more full-pipeline hosts (verifiers, lessons, provider holdout). Short bounds that guard work that never completes are left alone. Test-only.
Verified: the touched classes plus the parity and setup suites, 30 of 30 runs under 8 CPU burners, 0 failures; MicrosoftAgentFramework.Tests 1439/1439; Core.Tests 1348/1348.
🤖 Generated with Claude Code
https://claude.ai/code/session_01GPXz56QPQAzDSyJTxbQTTy