On Tue, Jan 24, 2012 at 10:03:05PM -0800, Kevin Layer wrote: >Larry Hall (Cygwin) wrote: > >>> > This problem is killing me. I'm currently looking msysgit + GnuWin32 >>> > because I just can't take the crashes of bash.exe and git.exe anymore. >>> > In my testing, so far, I've never seen msysgit or the bash that comes >>> > with it crash. Why is it that cygwin has this problem but msysgit >>> > does not? It's an honest question and I'm not trying to be >>> > provocative. I've been a cygwin user since before Red Hat acquired >>> > them, and the above statement makes me really sad. >>> >>> Have you tried running rebaseall? > >Absolutely. After updating cygwin, I reboot and run rebaseall -v >first thing.
FYI, as far as I can tell the stack trace that you provided did not seem to come from the 20120123 snapshot. >>> If not, install the rebase package and >>> read its README to get the proper procedure for running rebaseall. This >>> is a classic error message indicating colliding DLL addresses. Rebaseall >>> (and sometimes peflags) are the prescribed solution in these cases. >>> >>> If that doesn't solve the problem, a complete problem report would be >>> helpful. > >I have no idea how to make a reproducible test case of my system, >composed of 50+ repos, is large and not open source. We have shell >scripts that we use to apply git commands to each repo. > >One thing I've mentioned before: the problem became much worse when we >switched development to a 16-core machine. It's running Server 2008 >R2. > >Does anyone at Red Hat run on such a large-core machine? Why does that matter? This is a free software project staffed by one Red Hat person and a lot of people from other institutions. >The machine has been memtested, btw, and msysgit on the exact same >repos operates flawlessly, in my tests so far. All other non-cygwin >software on the machine works perfectly, too. > >If you think a bug report without a reproducible test case would be >useful, let me know what info I can provide. Hmm. Can you actually conceive of a situation where, when reporting a bug, a reproducible test case is NOT useful? Barring a reproducible test case you could provide some of the information that I asked for in the thread that you're responding to. And, we always want to see cygcheck output with the additional details asked for. cgf -- Problem reports: http://cygwin.com/problems.html FAQ: http://cygwin.com/faq/ Documentation: http://cygwin.com/docs.html Unsubscribe info: http://cygwin.com/ml/#unsubscribe-simple
