Re: allocate_string_data memory corruption

[Chong Yidong]

Stefan Monnier <address@hidden> writes:

>> In this function, data->string is set to s, and nbytes is set to
>> nbytes.  If check_sblock is a no-op, there should be no change.
>> However, we get an abort on the second debugging check:
>
> BTW, it's possible that -DSYNC_INPUT fixes the bug.

I just got a reply from Friedrich Friedrichs:

   CFLAGS="-O0 -DSYNC_INPUT -g" ./configure
   make clean
   make

   Then see if you can crash Emacs again.  (The point here is to see
   if the -DSYNC_INPUT compilation flag helps remove the problem.)

  I can't crash it with -DSYNC_INPUT enabled. (At least it doesn't
  crash during the usual operations and I was able to re-publish my
  whole planner-muse-wiki without problems.)

So it looks like signal handling during string/cons allocation is
indeed the culprit.