Decodetalkers wrote: > > > > I'm not convinced that simply removing these flags would solve the > > > > problem. Clangd should be ignoring the flags, so they are noise on top > > > > of the behavior of clangd. As such, I would like to see confirmation > > > > first that this works without this change. > > > > > > > > > I have debugged many times. And have compared the command that clang++ > > > used with that gcc is using. Then I found these options are redundant. In > > > fact, clang++ also removes the mapfile in its commands, for example > > > @CMakeFiles/.../abc.modmap. So what I have changed is meaningful > > > > > > It looks like we can just create `.clangd` config to remove these flags. > > I.e. > > ``` > > CompileFlags: > > Remove: [-fmodule*, -fdeps-*] > > ``` > > > > > > > > > > > > > > > > > > > > > > > > And after that everything seems working fine > > However, I also found that seems we don't forward command mangler right to > > modules dependency scanner, because I see some errors in logs: > > ``` > > E[10:34:07.505] Scanning modules dependencies for .../module-import/Use.cc > > failed: error: unknown argument: '-fmodules-ts' > > error: unknown argument: > > '-fmodule-mapper=CMakeFiles/Use.dir/Use.cc.o.modmap' > > error: unknown argument: '-fdeps-format=p1689r5' > > > > E[10:34:07.506] The command line the scanning tool use is: /usr/bin/g++-14 > > --driver-mode=g++ -O3 -DNDEBUG -std=gnu++20 -fmodules-ts > > -fmodule-mapper=CMakeFiles/Use.dir/Use.cc.o.modmap -MD -fdeps-format=p1689r > > ``` > > Seems to be a regression bug as we have #122606
Sorry for that, I think the database has already addressed with mangle in https://github.com/llvm/llvm-project/blob/main/clang-tools-extra/clangd/ClangdLSPServer.cpp#L566-L567, so I just ask why we need mangle again in this place. https://github.com/llvm/llvm-project/pull/200001 _______________________________________________ cfe-commits mailing list [email protected] https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
