Szelethus added inline comments.
================ Comment at: clang/lib/StaticAnalyzer/Checkers/CheckerDocumentation.cpp:199 + /// + /// This callback should be used to construct the checker's fields. + /// ---------------- Szelethus wrote: > Hmmm, what use case do you have in mind? Do we actually want this to happen? > By the time checkers are constructed, the AST is already built, alongside all > the other data structures that will be inspected by the analyzer. Since > checkers are constructed once for each clang invocation (not for every > `BeginAnalysis` -- `EndAnalysis` cycle), I would imagine that the checker > registry functions are more appropriate for such initializations. > > I'm only aware of one non-debug checker that uses `EndAnalysis` (maybe > `RetailCount`?), that clears some sort of a locally stored map that it > shouldn't really have anyways (since checkers are supposed to be stateless), > and I struggle to think of a realistic, supported use case for this callback > other than debug checkers. Turns out `deadcode.UnreachableCode` uses `EndAnalysis` quite cleverly as well. Repository: rC Clang CHANGES SINCE LAST ACTION https://reviews.llvm.org/D69745/new/ https://reviews.llvm.org/D69745 _______________________________________________ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
