fix: cancel per-attempt gRPC contexts promptly

[zjuzhongwen]

Description:

_Execute previously called defer cancel() directly inside its retry loop. Since the defer belonged to the outer function scope, cancellation for completed attempts was delayed until _Execute returned.

This could keep timeout-related resources alive longer than necessary when an RPC completed before its configured deadline.

The RPC invocation is now scoped to a small closure, allowing defer cancel() to run at the end of each attempt.

Related issue(s):

Fixes #1797

Notes for reviewer:

Checklist

• Documented (Code comments, README, etc.)
• Tested (unit, integration, etc.)

[lfdt-bot]

✅ Snyk checks have passed. No issues have been found so far.

Status	Scan Engine	Critical	High	Medium	Low	Total (0)
✅	Open Source Security	0	0	0	0	0 issues

💻 Catch issues earlier using the plugins for VS Code, JetBrains IDEs, Visual Studio, and Eclipse.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.

Files with missing lines	Coverage Δ
sdk/executable.go	77.18% <100.00%> (+1.20%)	⬆️

... and 13 files with indirect coverage changes

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[Dosik13]

Thank you for the contribution — the fix is correct, and extracting the attempt into its own function is a clean way to scope the defer cancel(). Summing up my comments:

1. Add a one-line doc comment on the helper so the per-attempt scope doesn't get inlined away later.
2. Deduplicate the two branches so the marshal + return appears once.
3. Please move the test into the existing mock_test.go and drive it through _Execute (BUSY → OK) — that checks the real retry behaviour and also covers the transaction path.

One more small thing: please change the issue reference to Fixes #1797 so it auto-closes. Happy to approve after that!

[zjuzhongwen]

Thank you for the contribution — the fix is correct, and extracting the attempt into its own function is a clean way to scope the defer cancel(). Summing up my comments:

1. Add a one-line doc comment on the helper so the per-attempt scope doesn't get inlined away later.
2. Deduplicate the two branches so the marshal + return appears once.
3. Please move the test into the existing mock_test.go and drive it through _Execute (BUSY → OK) — that checks the real retry behaviour and also covers the transaction path.

One more small thing: please change the issue reference to Fixes #1797 so it auto-closes. Happy to approve after that!

@Dosik13 Thanks for the review!

I’ve addressed all the comments:

• Added a doc comment explaining the per-attempt scope of _ExecuteRequest.
• Deduplicated the query/transaction branches so marshaling and returning happen once.
• Moved the regression test into mock_test.go and now drive it through _Execute with a transaction retry (BUSY → OK).

The test verifies that the first attempt’s context is already canceled when the second attempt begins.

Thanks again!

[Dosik13]

Thank you very much!