[libcxx-commits] [libcxx] [libc++] Fix vector<const T> (PR #80711)

[Louis Dionne via libcxx-commits]

ldionne wrote:

> To provide a different perspective, I am extremely happy to learn today that this landed yesterday, and we haven't lost a day's worth of test results buried under `std::vector<const T>` issues. It seems entirely reasonable to bypass standard procedure to recover from breakage, and then to reflect afterwards on what could have gone better, and to always assume positive intent.

I wasn’t assuming bad intent. I just think it’s surprising that someone who we’ve never (to my recollection ) interacted with in the context of libc++ would have sufficient permissions to merge code into libc++ just like any regular maintainer, and bypass CI checks. I don’t think my surprise is an overreaction, I would argue that LLVM is an outlier in that respect.

In this case there’s no actual problem, the FreeBSD CI was fine to ignore and this simply returned things to green faster. My observation is a general one about how liberal we are with permissions and I do think that’s something we can improve on to make the project run smoother (even outside of security aspects).

https://github.com/llvm/llvm-project/pull/80711