NoQ added a comment. In https://reviews.llvm.org/D46187#1089087, @lebedev.ri wrote:
> > I believe we could also benefit from a method of extracting the analyzer's > > `clang -cc1` run-line from clang-tidy. This would allow arbitrary debugging > > over a single analyzer invocation. > > I'm not really following, sorry. In https://reviews.llvm.org/D46187#1091309, @alexfh wrote: > In https://reviews.llvm.org/D46187#1088722, @NoQ wrote: > > > I believe we could also benefit from a method of extracting the analyzer's > > `clang -cc1` run-line from clang-tidy. > > > It should be possible now using `clang-tidy -extra-arg=-v some/file.cpp` (or > `clang-check -extra-arg=-v ...`). Nice! So essentially @lebedev.ri could do `clang-check -analyze -extra-arg=-v ../llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp`, then copy the printed `clang Invocation:` and append `-analyzer-checker debug.ViewCallGraph` to it. Or maybe even something as horrible as clang-check -analyze -extra-arg=-Xclang -extra-arg=-analyzer-checker -extra-arg=-Xclang -extra-arg=debug.ViewCallGraph ../llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp And it sounds like the roughly right level of verbosity for discriminating between a user-facing feature and a definitely-not-user-facing feature. I don't mind having a separate flag specifically for enabling debug checkers, but when it comes to actual debugging of the analyzer, debug checkers are usually not sufficient. Playing with other flags and `-analyzer-config` options is very often necessary, so if i was to debug the analyzer from clang-tidy, most of the time i'd have preferred to do the `-v` trick. Repository: rC Clang https://reviews.llvm.org/D46187 _______________________________________________ cfe-commits mailing list cfe-commits@lists.llvm.org http://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
