On Sun, Jun 05, 2005 at 11:25:15PM +0200, Jérôme Marant wrote: > > > > (1) works for Kurt, and (2) works for me. I've still not found a > > definitive cause. Because the breakage is timing-dependent, it's not > > possible to debug, because this slows down the build sufficiently that > > it always succeeds. > > Timing-dependent? How so?
If it isn't 100% reproducible, it's almost always a race condition. What I see is that when running "make -d", all my problems go away. Also, heaving a heavier load on my box makes it go away, or produce a slightly different build log. Kurt
