finish: close a referenced issue when its last open dependency closes #805

New issue

Closed

opened 2026-10-19 02:54:18 +00:00 by hexajon · 1 comment

hexajon commented

2026-10-19 02:54:18 +00:00

Owner

Today:

ClassifyBranchIssues gives ShouldClose: false when an issue has any open dependency (internal/service/workflow/classify.go:71-76).
The classification runs only inside finish. No later step checks a skipped close again.

What I want: when madtea closes an issue, it checks each issue that depends on it. If one of those issues had its close skipped only because of open dependencies, and it now has none, madtea closes it with a note that names the PR that built it.

Acceptance:

A test: finish of issue A with an open dependency B references A. When madtea then closes B, A closes too, with a note that names A's PR.
This step never closes an issue that had no skipped close.

When I finish a branch for an issue that still has an open dependency, finish references the issue and does not close it. That part is right. But nothing closes the issue later, when its last open dependency closes. So a built issue stays open, and I must find and close each one by hand. In one repo I found 15 such issues. Today: - `ClassifyBranchIssues` gives `ShouldClose: false` when an issue has any open dependency (`internal/service/workflow/classify.go:71-76`). - The classification runs only inside finish. No later step checks a skipped close again. What I want: when madtea closes an issue, it checks each issue that depends on it. If one of those issues had its close skipped only because of open dependencies, and it now has none, madtea closes it with a note that names the PR that built it. Acceptance: - A test: finish of issue A with an open dependency B references A. When madtea then closes B, A closes too, with a note that names A's PR. - This step never closes an issue that had no skipped close.

hexajon commented

2026-10-20 01:53:07 +00:00

Author

Owner

Superseded by #825, which covers this close-later step plus a catch-up check, through a madtea/held label queue, for blockers closed outside madtea. It rests on the forge probes in internal/testforge/closeprobe.

hexajon closed this issue

2026-10-20 01:53:07 +00:00

No milestone

No project

No assignees

1 participant

Notifications

Due date

The due date is invalid or out of range. Please use the format "yyyy-mm-dd".

No due date set.

Dependencies

No dependencies set.

Reference

sixfold-space/madtea#805

No description provided.

Rows
Columns