On 2026-08-21 09:26, Pavel Cahyna wrote:
Their commit messages hint at efficiency improvements and refactoring and API simplifications, not bugfixes. But I wanted to be sure.
Although that's mostly true, it's not entirely. 4dc85c626fb9dcc1ccacf0c1ab7c0e9f2163c06f improves a diagnostic on MS-Windows, in some cases. bad45bee3e4bda479b792162dc73a7817d5664d1 prevents mishandling of files whose inode number is out of ino_t range, and likewise for dev_t and time_t; this can matter on 32-bit platforms where EOVERFLOW is an issue. 941f62b24db232c58a41a0a621e08451f7f97ed8 fixes some errno confusion in diagnostics and fixes a bug on Haiku (where AT_FDCWD == -1 and the code mistakenly passed the -1 result of a failed open call to openat). 9371d36cc707a099e6867199f18bbaa7795b6150 fixes an unlikely race.
