On Sun, 13 Mar 2005 03:55:37 -0500, Gavriel State gav@transgaming.com wrote:
Actually, we first learned about the issue in the November-December timeframe, and mentioned it in our 4.2 release notes in December. In general, we still recommend that people use 2.4 kernels, since the scheduling changes can cause performance issues. We started to have a look at the problem, but by then Linus was already involved and the issue seemed well in hand. Cedega's signal handling code is certainly close enough to Wine at the lowest levels to still be affected by the same kind of issues with ptrace.
While we've tested the 2.6.11 ptrace fixes on x86, we had not done so on x86-64. We haven't recieved any reports from users that it's still broken, but if the equivalent x86-64 ptrace patch didn't get applied to 2.6.11, the it presumably could still be broken. Though I don't know how the 64-bit kernel deals with 32-bit code in this respect - is it possible that the x86 32 bit pthreads code is being used for 32-bit processes even on a 64-bit kernel?
Take care, -Gav
-- Gavriel State, Founder & CTO TransGaming Technologies Inc. gav@transgaming.com http://www.transgaming.com Let the games begin!
Ok, I understand now. I'm really looking for a better info under what conditions it fails. If wine and cedega are similar enough here, then they actually should be both exhibiting trouble on x86-64. Also, Andi Kleen has made up x86-64 patches, but hasn't merged them because they are untested. I offered to have them tested. So perhaps I'll get access to an amd system or maybe Thomas Zeeman find out more. I really haven't seen an amd system in action here.
Jesse