qiongsiwu wrote: > On a second thought, what's the current intent for build system integration? > I'm curious why we're emitting a diagnostic here rather than just passing the > list of paths back to the build system and letting it decide how to report > these.
This is a good point. Additionally, the current API design requires a worker calling it, since it needs the worker's VFS (or whoever owns the "current " VFS) to do the stat again. So this will report a diagnostic per worker, which may not be ideal. I can change the API such that we have two new methods instead of one. The first method collects the erroneously negatively cached paths, and the second reports it (in a form that the build system likes). The collection has to be done per worker, but the reporting can now be done once when the scanning service is destroyed. Does this sound reasonable? https://github.com/llvm/llvm-project/pull/135703 _______________________________________________ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
