higher-performance

@CastleOnTheHill, if this was as simple as merging your PR, we would have done so. I think you're under the (I believe mistaken) impression that your workaround actually fixes the...

> I didn't expect it to have any drawbacks. @CastleOnTheHill the drawback is now it hides a problem that still exists, and the lack of an error would potentially mislead...

Hi @joka921, thanks for the suggestions. I did take a look at this earlier actually, but it seems that ensuring exception safety for the class (not just this function) is...

> I think this could be interesting. Could you add a description of the benefit of this change? I know the linked issue explains it a bit, but it's good...

Anybody know why my tests keep getting canceled? I'm seeing errors like: > Canceling since a higher priority waiting request for 'Build and Test libc++-112751' exists Also, while I'm asking...

By the way - I keep getting [this confusing error](https://github.com/llvm/llvm-project/actions/runs/11472747270/job/31925816201?pr=112751#step:3:4068) in CI, and I'm not sure if I'm decorating something incorrectly. If anyone has an idea, please let me know:...

Ah thanks so much! I didn't realize I was adding this to `operator[]` on iterators too, that makes sense.

I seem to be encountering a strange situation where this code gives a dangling reference error when I annotate `std::vector::operator[]` with `[[clang::lifetimebound]]`: ```cpp #include #include std::string_view foo() { std::vector v(1);...

Ah wow, thanks for isolating that. It feels like the ` [[gsl::Pointer]]`/` [[gsl::Owner]]` analyzer strips off a layer of references/pointers when it shouldn't. This would be a blocker for us...

@Xazax-hun nope, thanks for asking!