Let's look at the history. Thanks very much for all that research and explanations, Bruno. Likewise, Jacob.
https://lists.gnu.org/archive/html/bug-automake/2010-10/msg00000.html Aha, the explanation for some of the $sleep commands scattered throughout the Automake tests! I had no idea. I'll add a comment at some point. (I myself added more, in vain attempts to work around the timing problems, but I don't think they should be necessary, in theory. Not that I plan to remove them.) This patch is conservative; IMO safe for a 1.17 release. I like the idea of a conservative patch, for sure. I see that what you're doing here is simply omitting the 1 second trial for the sake of saving that time, in exchange for accepting a 2s delay in the Automake tests on 1s filesystems. That seems fine to me. This patch is risky. I think it's safe to commit only if followed by a 1-month testing period, I installed your first patch. where we can see if people report failures like they were reported in September 2010. I don't see any reason why the problem would have gone away. The problem with this patch is that it reverts the change 5) in the Perhaps the AC_CONFIG_COMMANDS_POST as proposed by Nick and Jacob would ameliorate that. It doesn't incur any delay on reasonable systems. I'll reply separately about that. --thanks, karl.