[02:53:22] --- Russ has left: Disconnected
[07:26:07] --- jaltman has left: Disconnected
[08:39:34] --- jaltman/FrogsLeap has become available
[10:17:32] --- Simon Wilkinson has become available
[11:43:36] --- Simon Wilkinson has left
[12:04:10] --- Russ has become available
[13:15:09] --- steven.jenkins has left
[13:42:21] --- asedeno has left
[14:05:01] --- asedeno has become available
[14:06:00] --- Simon Wilkinson has become available
[14:06:51] <Simon Wilkinson> So, a) a compiler bug (if you have the stomach for it, look at the generated assembler and see if it is obviously wrong), or b) a race condition that is more likely to occur with non-optimized code. These days, I would vote for b).
[14:11:06] <kaduk@mit.edu/barnowl> There wasn't anything *obviously* wrong in the generated assembly (but
mind that the faulting instruction was 441 instructions into the
function and we only looked nearby).  But of course, my obvious is not
your obvious.
[14:13:05] <Simon Wilkinson> I'd look at stuff that gets reordered by the optimiser. It's possible that there is an ordering fault in the unoptimised code.
[14:14:09] <kaduk@mit.edu/barnowl> Or I could decide that clang is the way of the future and not spend
time tracking down gcc-4.2 bugs that may have already been fixed.
Though, I suppose I should actually try a clang-built module before
making such claims ...
[14:15:22] <Simon Wilkinson> Thing is, normal optimiser problems show up when you enable optimisation, not when you disable it.
[14:15:29] <kaduk@mit.edu/barnowl> Exactly.
[14:15:50] <kaduk@mit.edu/barnowl> (Also, I should probably fix the packaging bug that was reported to me
on friday before I go pouring unbounded time into other things.)
[14:17:04] <Simon Wilkinson> Can you reproduce with gcc-4.0 ?
[14:18:27] <kaduk@mit.edu/barnowl> Hmm, we don't seem to have that in ports, so it would be a fair bit of
work to test.
[14:18:55] <Simon Wilkinson> And clang? Does an unoptimised clang compile have problems?
[14:21:09] <kaduk@mit.edu/barnowl> Will give it a try soon.  (My test machine is currently my laptop,
so.)
[14:35:09] --- jaltman/FrogsLeap has left: Disconnected
[14:42:47] --- jaltman/FrogsLeap has become available
[15:11:53] <kaduk@mit.edu/barnowl> (Looking at this packaging bug) do we actually support mounting AFS
anywhere other than /afs ?
[15:14:05] <Simon Wilkinson> Yes
[16:31:04] --- steven.jenkins has become available
[22:58:22] --- Russ has left: Disconnected