[libcxx-commits] [libcxx] [libc++] Mark std::unique_ptr::release() as nodiscard (PR #110847)

[Nikolas Klauser via libcxx-commits]

philnik777 wrote:

> > @zmodem @davidben When you `release()` the pointer, do you use a custom deleter, or also the default deleter? If it's the former we could look into applying `[[nodiscard]]` only when it's the default deleter. That's why I never closed the referenced bug.
> 
> For the cases I've looked at, it's the default deleter. Like in @davidben's examples above, we call .release() because the ownership has been handed over to some other object already.
> 
> We did find at least one bug though: https://chromium-review.googlesource.com/c/chromium/src/+/4289548/13/chrome/browser/ui/views/bookmarks/saved_tab_groups/saved_tab_group_bar.cc#197
> 
> So marking it nodiscard certainly has value, and would probably have made perfect sense if it was there from the beginning. As to whether the signal-to-noise ratio is high enough now, I don't have a strong opinion. If we do decide to apply it, I would ask for a macro to make it easier to roll out.

Interesting. How does the C function call `operator delete`? I would have expected a custom wrapper for `free()` to allow interacting with C.

https://github.com/llvm/llvm-project/pull/110847