Subject: RE: Crash in embedded system

RE: Crash in embedded system

From: Daniel Stenberg <daniel_at_haxx.se>
Date: Mon, 14 Jun 2010 17:19:30 +0200 (CEST)

On Mon, 14 Jun 2010, Mark Smith wrote:

>> Any chance you can backtrack how the packet_length got so
>> weird? I suspect that is a reason for the following weirdness.
>
> Line 444 gets packet_length from the buffer, so I'm guessing it's failed
> before that line.

Exactly (even if line 444 doesn't match what's currently in git HEAD).

> I don't know what's supposed to happen or where the value is actually
> supposed to come from, whether it's later in the same buffer or different
> buffer content entirely. Isn't there supposed to be further content
> following the server ident before this point?

There's of course the whole handshake thing done before that point.

> I can add some debug message(s) in and re-run but I need to know what to
> look for. Is it possible to redirect trace messages to a callback or direct
> to a file?

libssh2_trace_sethandler() is the function you're looking for.

-- 
  / daniel.haxx.se
_______________________________________________
libssh2-devel http://cool.haxx.se/cgi-bin/mailman/listinfo/libssh2-devel
Received on 2010-06-14