model-writtenrun 6388 - may be wrong, links go to the source messages. Commit-by-commit LLM analysis is in the Commits dialog.
- Tom Lane proposed reducing repeated cache-clobber initdb runs by enabling the runtime setting for tests, retaining dedicated initdb coverage, and falling back to the compile-time setting on older branches in #1; Andrew Dunstan supported exploring the idea and highlighted repetitive TAP-test work in #2.
- Tom supplied companion initdb and buildfarm patches in #3, choosing a dedicated --clobber-cache option instead of the initially suggested general configuration passthrough.
- Andrew approved the buildfarm patch and suggested limiting cache-clobber initdb coverage to one locale in #4; Tom retained per-locale behavior because the client should not make that choice and existing CCA animals apparently used only one locale in #5.
- Tom proposed including the initdb option in v14 to complete runtime-controlled bootstrap testing in #6, and Andrew agreed in #7; b741f4c3ee landed the unchanged initdb component of #3, with a v14 backport, leaving the buildfarm component separate.
- After volunteering sifaka for testing in #8, Tom reported in #9 that its full schedule including TAP tests fell from 32:53:24 to 16:15:43 and confirmed that the buildfarm patch injected the compile-time setting on older branches.
- Andrew proposed renaming the buildfarm setting to use_clobber_cache_always and rejecting conflicting CPPFLAGS definitions in #10, while scheduling a client release after unrelated default-branch support; Tom accepted those changes in #11, leaving buildfarm integration and release pending at the thread's end.