Two defects in backlog-burndown "The Promotion Boundary", both in the paragraph that states the operational-model delta and the step that follows it.
Step 1 forces a promotion the delta says is discretionary. The delta paragraph says an operational repository "opens a promotion pull request only occasionally rather than per round, so confirm with the maintainer whether one is wanted at all there." Step 1 then reads "Open it whenever develop is ahead of main", unqualified, and names a count of zero as "the only case with nothing to promote". A run on an operational repository followed literally opens a promotion pull request every round, which is what the paragraph two sentences earlier says not to do. The absolute needs qualifying to the release model, the same shape #1276 identifies in this section's assembly procedure.
The direct-push closing set rests on a convention nothing requires. The delta says a direct push's "issues are recoverable only from the commit message itself", and instructs the assembler to "Read both, the pull requests for this run's work and the commit messages for the direct pushes". Neither git-commit-conventions nor operational-vs-release-workflow requires a commit message to name an issue at all, and neither defines a structured closing marker for one. So the read is unreliable in both directions: a valid direct commit that fixes an issue can name none and leave it open, and an incidental issue mention in a subject line can be promoted to a Fixes #N the author never intended. The feature-pull-request half of the same procedure has an explicit declaration for this reason, the mandated Closes on promotion: line, and the direct-push half has no equivalent.
This is distinct from #1276, which is that the assembly steps drop the direct-push hop the section body mandates. This one is that the hop is unsound even when performed.
Raised by qodo-code-review on the develop -> main promotion pull request #1304, findings 3 and 9, and deferred there under the maintainer's instruction to merge the promotion and take the findings up afterwards.
Two defects in
backlog-burndown"The Promotion Boundary", both in the paragraph that states the operational-model delta and the step that follows it.Step 1 forces a promotion the delta says is discretionary. The delta paragraph says an operational repository "opens a promotion pull request only occasionally rather than per round, so confirm with the maintainer whether one is wanted at all there." Step 1 then reads "Open it whenever develop is ahead of main", unqualified, and names a count of zero as "the only case with nothing to promote". A run on an operational repository followed literally opens a promotion pull request every round, which is what the paragraph two sentences earlier says not to do. The absolute needs qualifying to the release model, the same shape #1276 identifies in this section's assembly procedure.
The direct-push closing set rests on a convention nothing requires. The delta says a direct push's "issues are recoverable only from the commit message itself", and instructs the assembler to "Read both, the pull requests for this run's work and the commit messages for the direct pushes". Neither
git-commit-conventionsnoroperational-vs-release-workflowrequires a commit message to name an issue at all, and neither defines a structured closing marker for one. So the read is unreliable in both directions: a valid direct commit that fixes an issue can name none and leave it open, and an incidental issue mention in a subject line can be promoted to aFixes #Nthe author never intended. The feature-pull-request half of the same procedure has an explicit declaration for this reason, the mandatedCloses on promotion:line, and the direct-push half has no equivalent.This is distinct from #1276, which is that the assembly steps drop the direct-push hop the section body mandates. This one is that the hop is unsound even when performed.
Raised by qodo-code-review on the develop -> main promotion pull request #1304, findings 3 and 9, and deferred there under the maintainer's instruction to merge the promotion and take the findings up afterwards.