Problem
The benchmark matrix’s include: entries add build/next, build-cache/next, and run/run unconditionally. A filtered manual dispatch such as registry-only therefore still runs unrelated task-runner jobs and can wait roughly 40 minutes on build-cache.
That defeats the tight feedback loop added by #143 for partial dispatches.
Proposed direction
Generate the final matrix JSON in a setup job that honors the requested fixtures and variations. Add the special include pairs only when their variation is requested, or for scheduled/default runs.
Acceptance criteria
Source
Performance investigation: item 4 — matrix include pairs ignore dispatch filters
Problem
The benchmark matrix’s
include:entries addbuild/next,build-cache/next, andrun/rununconditionally. A filtered manual dispatch such as registry-only therefore still runs unrelated task-runner jobs and can wait roughly 40 minutes onbuild-cache.That defeats the tight feedback loop added by #143 for partial dispatches.
Proposed direction
Generate the final matrix JSON in a setup job that honors the requested fixtures and variations. Add the special include pairs only when their variation is requested, or for scheduled/default runs.
Acceptance criteria
build,build-cache, andrunpairs are included for scheduled/default runs.Source
Performance investigation: item 4 — matrix include pairs ignore dispatch filters