On Sat, Nov 19, 2011 at 5:09 AM, Chris Siebenmann <cks@...> wrote:
> To condense a long story, it appears that there is an internal bug
> in some versions of GIO's networking code where if an in-progress
> HTTPS connection is cancelled or otherwise closed down, the code
> will segfault. This manifests as NULL pointer dereference crashes
> in g_tls_client_connection_gnutls_finish_handshake(). Affected
> glib-networking versions include at least the version currently in
> Fedora 15 (glib-networking version 126.96.36.199 according to RPM version
> numbers), but it looks like the glib-networking people may have fixed
> the bug in early August.
> (This has surfaced before as Debian bug #628068,
> I don't know if this warrants any sort of note in the release notes or
> the FAQ for Liferea 1.8, but since I just finally diagnosed all of this
> for myself I figured I should mention it just in case.
Thanks for the hint! There are occusional crash tickets on the
SF tracker. I'll check wether we have some related to this. If it
becomes more frequent I'll add hints in FAQ and blog.