Jan Ph. H.

> error recovery Error recovery sounds very much like a component test (kill and restart a component) and not for unit tests, similar to what a panic test would do....

~~Probably the documenation (#1596) should be finished first.~~ Just curious: Is identifying manually the header files, which dont depend on libc, that difficult?

> print a header before error messages > collect these errors into a machine printable form 1. Do you have any opinion how the errors should look like? 2. What...

This proposal is lacking in some details, for example left and right shift can be both arithmetical and logical. However, the main problem is that this proposal does not clarify...

@RogierBrussee see [here](https://en.cppreference.com/w/c/chrono/timespec). POSIX does not specify, if [time_t](https://en.cppreference.com/w/c/chrono/time_t) is signed or unsigned. [Some history](https://unix.stackexchange.com/a/25367/273935). Being signed would have the advantage to represent events before midnight UTC of January 1,...

Should be fixed in PR #15501. Can be closed once Windows 10 SP1 is mandatory/WIndows 10 support is dropped with known shortcoming for non-ntfs systems.

I'd like to work on this, but without a pile of merge conflicts due to #14152.

> Take a situation where 2 threads simultaneously spawn a child process: you don't want one messing with fds to affect the other ==> you need to do all the...

scandum wrote another sorting algorithm with a nice overview table: https://github.com/scandum/crumsort/issues/1 You might want to take a lookt at the bottom tables: fluxsort and crumsort outperform pdqsort in all cases....

lithdrew dropped a comment of google getting 1GB/s with SIMD-accelerated quicksort https://opensource.googleblog.com/2022/06/Vectorized%20and%20performance%20portable%20Quicksort.html However, this depends on https://github.com/google/highway to unify common SIMD routines across platforms. **UPDATE** 20240418: There is [a company...