On 03/23/2014 02:35 PM, Paul Gevers wrote: > On 23-03-14 17:31, Christian Marillat wrote: >> David Smith <sidic...@gmail.com> writes: >> >> [...] >> >>> I created a more detailed way to reproduce the problem here: >>> https://sourceforge.net/p/liferea/bugs/1142/#f50c >>> >>> But the potential for it to always segfault on startup seems like a >>> pretty nasty bug. Even though technically it might not have happened on >>> previous versions due to flaws in previous versions not even being able >>> to add this feed at all. When I run liferea 1.10.4, the feed disappears >>> and I can't seem to add it no matter what I do. 1.10.7 seems to allow >>> it to be added if you try hard enough, but then just crashes later. >> Adding this feed doesn't crash i386 or amd64 liferea. I think you have >> miscompiled liferea. > Hi David, > > I also tried to reproduce your test case. For what it is worth, after > adding intltoolize to the override_dh_autoreconf target and compiling > the package from the Alioth git (targeting wheezy-amd64) also I can't > reproduce your test case. > > Paul > > Hmm.. Have you tried moving ~/.config/liferea out of the way to reset your configuration to the defaults and then try loading up liferea again? For me, It's crashing when it's trying to save an Etag and it says the Etag is out of bounds. The ETag comes from the HTTP header (which you can check with "wget -S --spider url").. So it really sounded to me like it's something wrong with specific feeds.. Looking at it further, liferea uses libsoup to grab the Etag and according to libsoup's documentation the value will always be either null or the etag. Liferea checks for null and it's not null.. It's just "out-of-bounds" when it tries to store it internally.
Upstream has suggested that startup crashes like this might be a race condition.. Where it's going out and getting the feed's HTTP header in a separate thread and then returning it back to the application so fast that the object that stores it, inside of the application, hasn't even been initialized yet. I guess that would require a very fast Internet connection and a very slow CPU to crash there if that was the case. I've got the fast internet connection 1000Mbit but I didn't think my CPU was particularly slow. On another note, I believe I've got 1.10.7-1 in the git repo all set for a review for upload into unstable. Let me know if you find any problems. Respectfully, -David -- To UNSUBSCRIBE, email to debian-bugs-dist-requ...@lists.debian.org with a subject of "unsubscribe". Trouble? Contact listmas...@lists.debian.org
