Re: Unique transaction names and revision number (was Re: [PATCH] Remove APR ICONV dependency on Windows)

On Mon, 09 Jul 2007, D.J. Heap wrote:

> On 7/9/07, Mark Phippard <markphip@gmail.com> wrote:
> [snip]
> >Are we sure this is even the problem? 10 milliseconds is a pretty
> >small window. Could the test be changed to sleep for 10 milliseconds
> >to prove this is it?
>
>
> I don't know why APR isn't using the high-resolution timers on
> Windows, but Ivan's explanation matches the behavior I'm seeing. To
> test it out, I added a Sleep(100) call to the failing tests where-ever
> they were creating new revs/txns and the tests consistently pass now.

Adding a serial transaction number to FSFS seems like a higher
priority than digging into this, since it should really be added
regardless. I'd rather see time spent implementing it...