cyndyishida wrote: > Does the described test case look like a build system problem? In my > understanding at step 4 we should have re-scanned B to know that we need some > B'.pcm depending on A-2.pcm. Regardless of the answer I don't want to block > the change as it is still an improvement. Just thinking if end users are even > supposed to see the new diagnostic. And if it should be actionable for the > end users.
It could be a build system bug or a dependency scanning bug (depending on whether you consider that a build system). The most recent case I investigated is a dependency scanning bug, but where a re-scan happened, the invocations carved from the scanner created tasks similar to the test case. The diagnostic is more actionable for us (clang module bug fixers), but I actually would categorize all `OutOfDate` diagnostics as that. Related to the last tidbit in: https://github.com/llvm/llvm-project/pull/136612#discussion_r2054990077 https://github.com/llvm/llvm-project/pull/136612 _______________________________________________ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
