Re: [Help-gnunet] horrible performance with 0.6.0

[Christian Grothoff]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Both problems are actually related.  The TCP transport in 0.6.0 contains a bug 
that will (under certain circumstances, namely a blocking TCP socket during 
the processing of a shutdown message) either block gnunetd (possibly only for 
a while, but the block can cascade though the network) or even, as you 
observed here, make it die by freeing NULL. 
The issue is fixed in CVS and there will soon be an 0.6.0a with even more 
fixes (gnunet-check, gnunet-convert, working nanosleep for BSD/OSX/Solaris).

But thanks for telling (I was kind of astonished that nobody else was telling 
me about this one, I discovered it two days ago...) :-)

Christian

On Thursday 09 October 2003 10:34 am, Michael Shuey wrote:
> Minor addendum: gnunetd just died - it freed a null pointer.  Here's the
> logs since it was last started:
>
> Oct  9 10:13:10 WARNING: HELO message received invalid (expiration time
> over limit). Dropping.
> Oct  9 10:18:00 ERROR: lookup produced result already seen.  Case: 14
> Oct  9 10:18:39 ERROR: lookup produced result already seen.  Case: 14
> Oct  9 10:18:46 ERROR: lookup produced result already seen.  Case: 14
> Oct  9 10:18:56 ERROR: lookup produced result already seen.  Case: 14
> Oct  9 10:20:04 ERROR: lookup produced result already seen.  Case: 14
> Oct  9 10:25:32 ERROR: lookup produced result already seen.  Case: 14
> Oct  9 10:27:58 ERROR: lookup produced result already seen.  Case: 14
> Oct  9 10:28:42 FREE called with NULL ptr at tcp.c:929
> Oct  9 10:28:42 __BREAK__ at logging.c:220
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.0.7 (GNU/Linux)

iD8DBQE/hY/v9tNtMeXQLkIRArc9AKCnAtJPdkWsaZkGRbUw/LnAbU0+rgCaAgh1
oByTNpYQ72+vuf5FG0ikzhs=
=TiJe
-----END PGP SIGNATURE-----