On Sat, 19 Oct 2024 at 10:11, Al Viro <viro@xxxxxxxxxxxxxxxxxx> wrote: > > There'd been an example of live use of that posted upthread: > > https://sources.debian.org/src/snapd/2.65.3-1/tests/main/seccomp-statx/test-snapd-statx/bin/statx.py/?hl=71#L71 > > "" rather than NULL, but... Note that outside of a library, I'd argue that using NULL is an active bug, and should not be supported. It's purely a Linux extension, after all. That said, I do think my "what if a library implements fstat() using fstatat()" argument could equally well be "what if a library does a Linux-specific optimization and uses NULL instead of the empty string". So I guess it would be entirely valid for a Linux libc to do something like int fstatat(int dirfd, const char *restrict pathname, struct stat *restrict statbuf, int flags) { if ((flags & AT_EMPTY_PATH) && !*pathname) pathname = NULL; return SYSCALL(__NR_fstatat, dirfd, pathname, statbuf, flags); } so yeah, I guess it's however we just decide to document it. Linus
