Marc Haber <[EMAIL PROTECTED]> writes: > On Mon, Apr 28, 2008 at 09:26:19AM +0200, Marc Haber wrote: >> I now see the same behavior (gnutls-cli -s -p 25 localhost crashing >> during TLS handshake) with one of my own hosts, running a current >> Debian sid. Unfortunately, this does not seem to be reproducible, >> since a freshly installed Debian sid in a virtual machine does not >> exhibit this behavior. > > Valgrind output attached. ... > *** Starting TLS handshake > *** Fatal error: Internal error in memory allocation. > *** Handshake has failed > ==4534== > ==4534== Invalid free() / delete / delete[] > ==4534== at 0x402265C: free (vg_replace_malloc.c:323) > ==4534== by 0x804D684: (within /mnt/usr/usr/bin/gnutls-cli) > ==4534== by 0x804EA4E: (within /mnt/usr/usr/bin/gnutls-cli) > ==4534== by 0x414444F: (below main) (in /lib/libc-2.7.so) > ==4534== Address 0x430ce18 is 0 bytes inside a block of size 10 free'd > ==4534== at 0x402265C: free (vg_replace_malloc.c:323) > ==4534== by 0x804D684: (within /mnt/usr/usr/bin/gnutls-cli) > ==4534== by 0x804EA43: (within /mnt/usr/usr/bin/gnutls-cli) > ==4534== by 0x414444F: (below main) (in /lib/libc-2.7.so)
There isn't much to go on here, could you put --memcheck:num-callers=50 in your ~/.valgrindrc? Does the server by any chance have a very large set of known CAs? Could you run gnutls-cli with -d 4711 or something and post the output too? If the packets are very large, this may be the case. Is it possible to test against this host remotely? /Simon -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]