Wireshark-bugs: [Wireshark-bugs] [Bug 9920] Buildbot crash output: fuzz-2014-03-22-14025.pcap

Date: Tue, 01 Apr 2014 18:28:23 +0000

Comment # 9 on bug 9920 from
(In reply to comment #8)
> without thinking about this too hard, could RTP simply empty the hash table
> rather than destroy it and create a new one? this would avoid the crashes,
> but I don't know if it would actually give the right dissection...

It could, but that could cause memory leaks if SDP thinks they're free'd; it
would allocate them again later... which will happen if the media for that
channel changes back/forth using dynamic payload later (like in an SDP change
mid-call).

Then again, memory leaks appear to already be happening to some degree - I
don't see anything in this code that destroys the hashtables at the end of
calls, or even at the close of a capture file. The conversation stuff is
wmem-created, but the hashtable itself isn't, and there's nothing hooked into
wmem_register_cleanup_callback to destroy the hashtable.

So the wmem-managed struct has a pointer to the hashtable, but I assume wmem
doesn't somehow magically go call g_hash_table_destroy() on the member
hashtable right? :)


You are receiving this mail because:
  • You are watching all bug changes.