Seen in glibc
emaste markj kevans obiwac_gmail.com imp
- Group Reviewers
- rGa294e679738b: Document scandirat(3)
rG9fb8e8eede5b: libc: add scandirat(3)
rGcb6e97f4dae6: libc: split scandir() into scandir_dirp() and proper scandir()
rGea448a0a43c5: Link scandir_b(3) to scandir(3)
rGaa5e19a9b3f3: scandir.3: Use .Fo instead of .Fn for long args
Looks good. Maybe add a 'tidy up a few scandir_b related issues while in the area' to the commit message?
Otherwise looks great.
|498 ↗||(On Diff #109721)|
I'd mention adding this link in the commit since it was missing before (eg, it's otherwise unrelated to this change)
The wording implies that in the case where dirname is absolute, dirfd may be invalid. But, that's not true right? So in the description of the absolute case, perhaps add, "The .Fa dirfd argument is unused, but must be a valid file descriptor or the special value .Dv AT_FDCWD."
Hm, I thought that dirfd can be invalid. And both from my reading of namei_setup(), and actual testing, dirfd is ignored when first character of path is '/':
openat(-10000004,"/etc/passwd",O_RDONLY,00) = 3 (0x3)