Repository navigation
Pipeline workspace artifact exports - #626
Merged
Merged
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Completed Turns wait for output Artifact capture, whose sequential File reads amplify remote latency. Export files in batches of at most four and the declared handle limit, with one ordered tar writer. Preserve file size/mtime validation, symlink handling, all export bounds, and descriptor identity; cancel and join the batch on errors and settle handles before releasing owner references.
A seven-file real File fixture preserves byte-identical archives and all 41 RPCs, while 20/100 ms injected handler latency falls about 36.6%. Zero-delay samples show no reliable improvement; cumulative Go allocations increase by roughly 80 KiB. Readers hold at most one bounded File response each and share a 32 KiB copy buffer per batch, without whole-file buffering. Measurements precede the final error-cleanup cancellation follow-up; successful scheduling is unchanged. Live latency is not yet requalified.
Validation: focused agenthost/dispatch tests, relevant race tests ×20, exact 200 MiB streaming and advertised handle limits 1–4, failure/cancellation/late-acquisition cleanup, vet, check-names, check-docs and check-ci. Independent full-diff review found one cancellation issue; corrected with a three-line change and an asymmetric blocked-read regression (old fails, correction passes race ×20). No other substantive findings.
Need help on this PR? Tag
@codesmith-botwith what you need. Autofix is disabled.