bug#10913: 24.0.94; Crash in redisplay code

[Jan Djärv]

29 feb 2012 kl. 19:34 skrev Eli Zaretskii:

>> From: Hannu Koivisto <azure@iki.fi>
>> Date: Wed, 29 Feb 2012 18:12:35 +0200
>> 
>> Random crash.  I don't know how to reproduce this.  Backtrace included
>> below.  Some string variable contents have been changed from what
>> they really were.
> 
> This is an optimized build, right?  Otherwise, some variables shown
> here make no sense at all.  E.g., this:
> 
>> #3  0x010873f3 in fast_looking_at (regexp=6, pos=676026, pos_byte=96427,
>>    limit=89596416, limit_byte=285891, string=1) at search.c:588
> 
> `regexp' is a Lisp string, so it cannot have this value.  And `pos'
> and `limit' are likewise garbled.
> 
> Anyway, if I'm to believe this backtrace (which I don't, since the
> build is evidently optimized), the crash happened here:
> 
>         switch (SWITCH_ENUM_CAST ((re_opcode_t) *pat++))
>           {
>           case on_failure_keep_string_jump:
>             assert (str == NULL);
>             goto continue_failure_jump;
> 
>           case on_failure_jump_nastyloop:
>             assert ((re_opcode_t)pat[-2] == no_op);
>             PUSH_FAILURE_POINT (pat - 2, str);
>             /* Fallthrough */
>           ...
>           default:
>             abort (); <<<<<<<<<<<<<<<<<<<<<
>           }
> 
> which means what? that compile_pattern produced an invalid result?
> I'd be surprised.


Remember that optimization may join aborts together, so they look like they 
happened there, but it can be just about any abort in that file.

        Jan D.