[darcs-users] TortoiseDarcs and Darcs 2 Hashed Repos

[David Roundy]

On Thu, Oct 16, 2008 at 11:30 AM, zooko <zooko at zooko.com> wrote:
> On Oct 14, 2008, at 8:00 AM, David Roundy wrote:
>
>> Why would  we need more than 2 seconds? (Not disbelieving, just
>> curious.) Does windows have a time stamp  granularity of greater than
>> 1 second?
>
> FAT's granularity is 2 seconds, but I think I saw rumours on the web
> archives of various mailing lists that on some filesystems, operating
> systems, virtual machines, etc. that the effective granularity of mtimes is
> sometimes 3 or 4 seconds.
>
> Still not knowing what we are actually talking about, I wonder if "sleep 2"
> or "sleep 4" couldn't be replaced with something like
>
> touch a_temp_file
> while mtime(thefile) == mtime(a_temp_file):
>    sleep 0.2
>    touch a_temp_file
>
> If that could be made to work, it would be more reliable and faster than
> "sleep 2".

I don't think it's worth this effort, partly because of the infinite
loop possibility that Dan mentions (which one could certainly work
around), and partly because in the worst case this issue could lead to
a false negative, and odds are that even with a granularity of four
seconds, we'll see a failure (if there's a bug) after just a few runs
of the test.

David