Nathan Ridge

> Compatibility with CMake < 3.5 has been removed from CMake. Seems that this is new in CMake 4.0, and therefore the issue likely arose from the CMake version used...

> However, it's not entirely clear to me what determines the CMake version used for the build, and why it's being upgraded without us asking for that... I guess it...

This was fixed by https://github.com/clangd/clangd/pull/2522.

Reduced testcase for the underscore issue: ```c++ #define gl_foo() 42 int gl_frob(); int main() { int x = glf^ } ``` Completion at the `^` offers `gl_frob` but not `gl_foo`,...

(I am going to move this into the clangd repo since it's a server-side issue.)

> Also this is not about underscore at all Revised issue title to clarify that this is about using exact prefix matching vs. fuzzy matching for macro symbols. Underscores are...

I'm re-opening this to track the addition of an option to use fuzzy matching for macros. (Making it the **default** would require some data as Kadir said in [this comment](https://github.com/clangd/clangd/issues/1480#issuecomment-1411931395),...

And I marked this as a "good first issue". If someone would like to write a PR adding such an option: * [Here](https://github.com/llvm/llvm-project/blob/d72e711e864dad7e3a434d66f3febad2b1596335/clang-tools-extra/clangd/CodeComplete.cpp#L1992) is the previously referenced place in the...

I agree that such an option would be useful. There is discussion about a "strong workspace mode" in https://github.com/clangd/clangd/discussions/907 which would accomplish this. Until that is implemented, the workaround is...

cc @ChuanqiXu9