Re: 64-bit emacs crashes a lot

[Ken Brown <kbrown@cornell.edu>]

On 8/16/2013 9:31 AM, Eli Zaretskii wrote:
>> Date: Fri, 16 Aug 2013 07:41:44 -0400
>> From: Ryan Johnson <ryan.johnson@cs.utoronto.ca>
>>
>> On 16/08/2013 4:58 AM, Eli Zaretskii wrote:
>>> Again, please move this discussion to emacs-devel.
>> Back pointer: http://cygwin.com/ml/cygwin/2013-08/msg00273.html
>>
>>>> Date: Thu, 15 Aug 2013 22:35:54 -0400
>>>> From: Ken Brown <kbrown@cornell.edu>
>>>>
>>>> 1. Invoke 'emacs-nox -Q' in mintty.
>>>>
>>>> 2. M-x compile C-a C-k ls RET
>>>>
>>>> 3. C-x o
>>>>
>>>> 4. Hit 'g' repeatedly.
>>>>
>>>> I got it to abort with Fatal error 6 after slightly over 100 repetitions.
>>>>
>>>> I then tried the same thing with emacs-X11 (running under X, not in
>>>> mintty).  I hit 'g' 200 times without a problem.  I repeated this with
>>>> emacs-w32, again 200 times without a problem.
>>>>
>>>> So there's a bug somewhere.  But if it's an emacs bug, it's strange that
>>>> it only occurs with emacs-nox and not with either of the GUI versions of
>>>> emacs.
>>> I suspect that buffer relocation might be the reason.  Can you show a
>>> backtrace from the fatal error in an unoptimized build, with the above
>>> recipe?
>> I'll try to get one. Meanwhile, is there a quick way to disable buffer
>> relocation as a sanity test?
>
> I don't think so, but maybe it's already disabled in your build.  What
> do you see on these two lines when configure finishes:
>
>    Should Emacs use a relocating allocator for buffers?    yes
>    Should Emacs use mmap(2) for buffer allocation?         no

   Should Emacs use a relocating allocator for buffers?    no
   Should Emacs use mmap(2) for buffer allocation?         yes

> In any case, you could configure with --enable-checking=all, it might
> catch the problem earlier.
>
>> If the optimized build still crashes reliably without it, the real
>> problem is probably something else.
>
> Maybe, but not necessarily.

FWIW, I just tried this on the trunk, both with and without 
optimization.  The bug is still there in the optimized build, but I 
couldn't trigger it in the unoptimized build.  I've also checked that 
the bug occurs only in 64-bit Cygwin, not 32-bit Cygwin.

Here's some speculation.  On 64-bit Cygwin, the space allocated to the 
heap and mmap goes from 0x6:00000000 to 0x700:00000000, with the heap 
starting at the bottom and mmaps allocated from the top down.  (I've 
added the colons for readability).  Is it possible that the optimized 
build takes shortcuts that are somehow incompatible with this model?

I'm grasping at straws here, but I'm just trying to think of things that 
differ between the 32-bit and 64-bit systems.

Ken