Re: balsa segfaults or doesn't die properly occasionally

[Peter Bloomfield <PeterBloomfield bellsouth net>]

On 02/29/2004 03:46:24 AM, Kacper Wysocki wrote:
> On 02/28/04 06:31:46, Kacper Wysocki wrote:
>> I have been reluctant to post about this because I have very  
>> little in terms of debug information, but the problem is  
>> persistent and has been bothering me for quite some time. Next  
>> up I will try stracing balsa in the hopes of locking in on the  
>> problem.
> 
> So I managed to get a traceback, but it's not very useful on  
> account of the missing symbols, so my questions are:
> -how do I compile cvs balsa with symbols enabled? Is it as easy  
> as setting CFLAGS to '-g' before ./configure or make? 
> 
> -are there any comilation flags that I can set to have balsa  
> spit out more debugging information than it already is?
> 
> Also, the last debug messages I got before the segfault were:
> --
> Send_Message: 233790096, 1, 250 i1T8PCHo000266 Message accepted  
> for delivery
> 
> 
> C: QUIT
> 
> S: 221 2.0.0 my.host.com closing connection
> 
> Send_Message: 233783144, 1, Disconnected
> Send_Message: 234009392, 4,
> free(): invalid pointer 0x9227e50!
> free(): invalid pointer 0x435231b0!
> "/home/comotion/" is not a core dump: Is a directory
> --
> so it looks like a classic segfault to me..

Those `invalid pointer' messages mean heap corruption, and the  
segfault was probably a consequence.  If you run with  
MALLOC_CHECK_=2 the allocation/free routines will check every  
action, so you can find the errors as they occur instead of  
later.  If you also run with --g-fatal-warnings you'll get an  
abort on any such warning, and maybe get a useful stack trace.

I admire for your efforts in hunting down this one!

Peter