Re: allocate_string_data memory corruption

[Chong Yidong]

> "Richard M. Stallman" <address@hidden> writes:
>
>> One possible solution would be to realize the mouse-highlighted version
>> of each face when we realize the face itself.  In effect, that means
>> calling realize_face earlier, so that when lookup_face is called
>> in the signal handler, it will find what it needs.
>
> All of this seems much trickier than just defining SYNC_INPUT by default.

I'm beginning to think so too.

Just by eyeballing, I can find several other places in
handle_one_xevent that can end up calling Fcons or Fmake_vector:

5884: x_handle_dnd_message -> Fmake_vector

6069: record_asynch_buffer_change -> kbd_buffer_store_event ->
  kbd_buffer_store_event_hold -> Fcons

6865: kbd_buffer_store_event_hold -> Fcons

6882, 6888: gen_help_event -> kbd_buffer_store_event ->
  kbd_buffer_store_event_hold -> Fcons