Sam McCall

Indeed, closed too early. > Goto Definition works but does not help in my case since it's often a shared_ptr Yes. This is independent of this issue as you noted....

It would be nice to support clangd on systems without build, but this is pretty far from being feasible today. Regarding your suggestions: - the compile_commands.json used on the server...

> which in our case takes ~10 minutes to generate and I'd ideally not ask users to all run that command on their machine This seems like a bigger issue....

Looks like indexing completed OK before the request. My guess is this is yet another case-sensitivity issue somewhere...

No, there's something else going on (too?). ``` // src.h extern void foo(void); extern int bar; ``` ``` // src.c void foo(void) {} int bar; ``` No compilation database, open...

`void foo(void) {}` gets a different SymbolID when parsed in C mode vs in C++ mode. In C++ the USR is `c:@F@foo#` and in C it is `c:@F@foo`. `c:` is...

Yeah, the problem is: - we need the same USR for every redecl - in C any of the redecls can be `void foo()` with no prototype given, therefore the...

I've started a discussion on extracting a library so we can have a clang-tidy check: https://discourse.llvm.org/t/rfc-lifting-include-cleaner-missing-unused-include-detection-out-of-clangd/61228. Feel free to chime in!

I'd be happy to see this in as an option if someone wants to contribute it. On the other hand I/others can't really prioritize it if no widely used editors...

Right, break within a line comment should work. For now at least breaking at the end of a line comment doesn't continue the comment (deliberately). I don't think we implemented...