Pádraig Brady <[email protected]> writes:
> On 08/11/2025 06:41, Collin Funk wrote:
>> On GNU/Hurd tests/tail/tail-c fails because the following command times
>> out:
>> timeout --verbose 1 tail -c 4096 /dev/urandom >/dev/null 2>err
>
> Ah right. TBH Hurd is perfectly entitled to behave like that.
> I.e. tail resorts to reading all data as non seekable.
> The patch looks fine. Please push.
I thought it was fine too, because urandom and random aren't
standardized. I prefer the behavior on GNU/Linux though.
On GNU/Linux the offset is always zero and lseek is a noop that
succeeds:
$ cat main.c
#include <stdio.h>
#include <stdlib.h>
#include <fcntl.h>
#include <unistd.h>
#include <inttypes.h>
int
main (void)
{
int fd = open ("/dev/urandom", O_RDONLY);
if (fd < 0)
abort ();
printf ("%jd\n", (intmax_t) lseek (fd, 0, SEEK_CUR));
printf ("%jd\n", (intmax_t) lseek (fd, 1, SEEK_CUR));
unsigned char dummy;
if (read (fd, &dummy, sizeof dummy) <= 0)
abort ();
printf ("%jd\n", (intmax_t) lseek (fd, 0, SEEK_CUR));
printf ("%jd\n", (intmax_t) lseek (fd, 1, SEEK_CUR));
return 0;
}
gcc main.c && ./a.out
0
0
0
0
> Note also the 1 second timeout is too aggressive.
> For any test that we consider to have failed after a timeout
> the timeout should be 10 seconds to cater for slow/loaded systems
> and avoid intermittent false failures.
> I've just pushed a patch to fix that.
Cool, thanks.
Collin
