tigerquoll commented on PR #1061:
URL: https://github.com/apache/yunikorn-k8shim/pull/1061#issuecomment-5466451327

   Rebased onto master and pushed. Two things changed since the last review 
round.
   
   **Three shim-owned exemptions are gone.** YUNIKORN-3367 (#1074) landed on 25 
Aug — it closes the shim stop channel and always releases the dispatcher and 
placeholder manager. That kills the exemptions for 
`PlaceholderManager.Start.func1`, `dispatcher.Start.func1` and 
`wait.BackoffUntilWithContext` (YUNIKORN-3368 was closed as a duplicate of 
3367). Shim-owned carve-outs go from four to one; the survivor is 
YUNIKORN-3369, still open.
   
   @wilfred-s — this is the burn-down, one cycle in. The exemptions this PR 
opened with are being deleted by the fixes they documented, and the deletion is 
mechanical precisely because each entry names its JIRA.
   
   **Two core exemptions had silently stopped matching.** The yunikorn-core 
bump on master renamed `EventSystemImpl.StartServiceWithPublisher.func2` to 
`.func1` and unexported the publisher (`EventPublisher.StartService` → 
`eventPublisher.start`), so both entries had become no-ops and `pkg/shim` was 
red. Realigned, and I recorded a caveat in `leakcheck.go` that a dependency 
bump can rename a frame out from under an exemption — now demonstrated rather 
than hypothetical.
   
   I also re-probed every remaining exemption by deleting it and confirming the 
check then fails, so none is dead weight: the DRA tracker entry 
(YUNIKORN-3371), the AssumePod entry (YUNIKORN-3369), and all 17 inherited 
core-service entries.
   
   Verification on the rebased branch:
   
   - `go test ./pkg/... -tags deadlock`: 18/18 ok
   - `go test ./pkg/... -race -tags deadlock`: 18/18 ok, no data races
   - `go test ./pkg/shim/ -tags deadlock -count=3`: ok
   - golangci-lint, `make license-check`, `go vet`, `gofmt`: clean
   - `go mod tidy`: no-op, so no `go.sum` change is needed on the shim side
   
   @manirajv06 — your review is still marked changes-requested. On mirroring 
the core PR: as in my earlier reply, YUNIKORN-3357 spans both repos and the 
per-component defects live in their own linked JIRAs, and the explanatory 
essays have already been stripped from both PRs in favour of `See 
YUNIKORN-####` references. Is there something further you'd like changed here, 
or can this go back for another look?
   


-- 
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.

To unsubscribe, e-mail: [email protected]

For queries about this service, please contact Infrastructure at:
[email protected]

Reply via email to