Alex pointed me to the fact that a.mtime == b.mtime compares values to a unit smaller than a second, while it apparently stops at the second so far with MRI. So we stopped doing mtime equality because it was a bad idea, but you might want to look into this, if that change was not intended.
On Wed, May 26, 2010 at 15:14, Charles Oliver Nutter <head...@headius.com>wrote: > On Wed, May 26, 2010 at 5:06 PM, Antoine Toulme <anto...@lunar-ocean.com> > wrote: > > > > > > On Wed, May 26, 2010 at 15:02, Charles Oliver Nutter < > head...@headius.com> > > wrote: > >> > >> I have not looked at the issues...they don't happen with RJB? > > > > No idea, because win7 and RJB kinda suck. > >> > >> It could > >> be a problem with the way we implement mtime on Windows. Any update on > >> this in the past few days? (I was out of town). > > > > We changed the way we test mtime. Instead of doing ==, we now use < and > > to > > avoid races. > > We still have a few failing specs re permissions. No headway on those. > > > >> > >> We're looking to push JRuby 1.5.1 very soon, so if there's something > >> to fix we need to get it in. > > > > So far nothing to report. > > If you can narrow it down to something specific broken in JRuby, we'll > gladly fix it. > > - Charlie >