THREAD curiosity

[Bron Gondwana]

On Wed, Oct 08, 2008 at 12:06:02PM +0100, David Carter wrote:
> If I import the attached Unix mailbox (culled from some real messages 
> from an ITEF list) into Cyrus, then the index_thread_compare() call 
> started by:
>
>   . THREAD REFERENCES US-ASCII ALL
>
> segfaults with a NULL pointer deference:
>
>   Program received signal SIGSEGV, Segmentation fault.
>   [Switching to Thread 0xb79916d0 (LWP 30378)]
>   0x0807b0fe in index_thread_compare (t1=0x817ff60, t2=0x817ffc0,
>     call_data=0xbfcc0e28) at index.c:4417
>   4417      md2 = t2->msgdata ? t2->msgdata : t2->child->msgdata;
>
> This is because t2->msgdata and t2->child are both NULL.
>
> I infer that the large DKIM-Signature: header blocks are causing a buffer 
> overrun somewhere although I can't see where right now. This is true of  
> both 2.3.11 (which I am currently running) and a vanilla 2.3.13rc2.

Interesting.  Putting on my "don't let anything derail the release
plans" hat, it's not a regression :)

I don't like things that can segfault Cyrus though... makes me
uncomfortable!  Any chance you can bisect the list until you find
the message which causes the crash?

Bron.