utime assigns arbitrary subsecond timestamps

[bhaible]

The utime() function, specified by previous POSIX at ​https://pubs.opengroup.org/onlinepubs/9699919799/functions/utime.html, ought to set the access time and modification time of a file to the specified values.

On Haiku (I'm using a July 2024 snapshot) the subsecond part is ignored, and instead an subsecond value is used. If this arbitrary value was the same for all files, it would be OK. But no, the arbitrary values are different!

This has the effect that files that should have the same modification time suddenly have different modification times. And thus, when I have a Makefile rule that generates one of the files from the other one, this Makefile rule will trigger, although it shouldn't.

How to reproduce: Save fu.c (attached).

$ gcc -Wall fu.c
$ ./a.out
$ ls -lrt --full-time dummy*
-rw-r--r-- 1 user root 0 2024-10-30 16:48:30.023855104 +0100 dummy3
-rw-r--r-- 1 user root 0 2024-10-30 16:48:30.023855104 +0100 dummy2
-rw-r--r-- 1 user root 0 2024-10-30 16:48:30.023855104 +0100 dummy1
-rw-r--r-- 1 user root 0 2024-10-30 16:48:30.023855104 +0100 dummy0
-rw-r--r-- 1 user root 0 2024-10-30 16:48:30.024117248 +0100 dummy8
-rw-r--r-- 1 user root 0 2024-10-30 16:48:30.024117248 +0100 dummy7
-rw-r--r-- 1 user root 0 2024-10-30 16:48:30.024117248 +0100 dummy6
-rw-r--r-- 1 user root 0 2024-10-30 16:48:30.024117248 +0100 dummy5
-rw-r--r-- 1 user root 0 2024-10-30 16:48:30.024117248 +0100 dummy4
-rw-r--r-- 1 user root 0 2024-10-30 16:48:30.024379392 +0100 dummy9

You can see that dummy0 and dummy9, for example, have different time stamps.

[bhaible]

test case fu.c

[waddlesplash]

This invokes the same syscall as #19213 (_kern_write_stat).