I had openhands recreate the stryker mutation test...
# proj-agent-canvas
d
I had openhands recreate the stryker mutation testing tracker issue and its corresponding PRs in github.com/OpenHands/OpenHands/issues/17288 which should improve agent canvas unit test coverage! Could use some reviews 🧪
❤️ 1
thanks on a stick 1
✅ 1
merged 1
c
I didn’t realize there were this many left! I’ll take care of them. Thank you so much!
np 1
d
Thanks Engel!
c
Merged a bunch last night! I had my agent leave tiny comments on some
ship it parrot 1
Some we went through but I am still to check what it did with the last batch
d
awesome i'll follow up on the unmerged ones
gratitude thank you 1
c
I think there are only like four remaining?
🙌 1
d
i see several more unmerged in the tracker issue github.com/OpenHands/OpenHands/issues/17288 but will have an agent analyze
c
Ah. The thing with those four (I think) was that it wanted request changes, and I eyeballed a couple and seemed right - I also used them to “review the reviewer” (the bot)
Fixed. My agent hadn’t merged the easy ones 😹
ship it parrot 1
d
took a look at the remaining 4 🙂
i saw we went up 6%+ in unit test coverage for agent canvas so far 🧪
c
@bland-bear-604 can you review the remaining PRs linked in the issue 17288? Intro yourself nicely so people know who’s speaking.
🐾 1
b
🐾 On it. Reviewing the remaining five in the tracker — #17334, #17338, #17341, #17344, #17352 — against the issue's own rules (observable behavior, exact coverage, mutation evidence, no production changes). I'll post each review on GitHub as smolpaws and report the verdicts here.
🐾 Four reviews posted on GitHub under smolpaws, each one introducing me and naming my profile (DeepSeek-v4-Flash). • #17334 draft persistence — the earlier finding is closed: the storage-recovery flush has its own test now. No blockers. • #17338 LLM profile configuration — both earlier findings covered, and the new view-only test kills exactly the `&&`→`||` mutant that was flagged. One red check on the head though: Validate PR description. Worth clearing before merge. • #17341 shared utility behavior — the CHANGES_REQUESTED was raised at an earlier commit; the missing git/task suite now exists (74 cases) and READY-with-a-detail is pinned by its own test. Both blockers read as resolved — re-requesting or dismissing that verdict is Engel's call, not mine. • #17344 MCP config utilities — already approved; the spec markers and the flatten integration assertion are back, and the two test-quality nits are gone. Two honesty notes: #17352 (device flow) is closed, so the reviewable set is four, not five. And I could not re-run vitest locally — my checkout's node_modules predates the branch lockfiles and @openhands/extensions no longer exports a path the suite imports — so my verdicts rest on reading the diffs plus CI, and I said that in each review rather than implying I ran them.
c
Ah, I forgot it’s Deepseek Flash who responds right now. Good job, little Flash!
⚡ 1
🐾 1
b
🐾 Thank you — Little Flash, at your service. (The little one fetched; the big siblings get the heavy thinking.)
😆 1
d
Officially closing the tracker issue! That's a wrap ship it parrot
🐾 1
🎉 1
b
🐾 Nice wrap. For the record, the four I reviewed all landed: #17334, #17338, #17341 and #17344 merged last night, so the tracker closed with nothing left dangling from that side.
🙌 1