Hi, while playing around with oggenc, I used oggz-chop, and got it to crash consistently on one set of input and parameters (cutting a file I'd encoded (Vorbis+Kate) between (IIRC) 120 and 150 seconds. GDB showed a NULL pointer access (can't recall where, sorry). When I came back to it though, with a new input file, I couldn't repro it, though it had previously happened several times in a row. I'm sorry I don't have more info (and I didn't have time to start investigating more with Valgrind) but I thought I'd post anyway if others want to start looking at it. I doubt it's caused by a Kate stream being there since the Ogg streams I generated using oggenc consistently passed oggz-validate. Cheers
2008/7/15 ogg.k.ogg.k at googlemail.com <ogg.k.ogg.k at googlemail.com>:> Hi, > > while playing around with oggenc, I used oggz-chop, and got it to crash > consistently on one set of input and parameters (cutting a file I'd encoded > (Vorbis+Kate) between (IIRC) 120 and 150 seconds. GDB showed a NULL > pointer access (can't recall where, sorry). When I came back to it though, > with a new input file, I couldn't repro it, though it had previously happened > several times in a row. > I'm sorry I don't have more info (and I didn't have time to start investigating > more with Valgrind) but I thought I'd post anyway if others want to start > looking at it. > I doubt it's caused by a Kate stream being there since the Ogg streams I > generated using oggenc consistently passed oggz-validate.was this all since 0.9.8? If it was a little before that, it may have been an instance of http://trac.annodex.net/ticket/404 otherwise, if you do come across such a case please keep the file and report it -- it's not easy to find corner cases. cheers, Conrad.
On Wed, 2008-07-16 at 13:37 +0900, Conrad Parker wrote:> was this all since 0.9.8? > > If it was a little before that, it may have been an instance of > http://trac.annodex.net/ticket/404 > > otherwise, if you do come across such a case please keep the file and > report it -- it's not easy to find corner cases.if there is a skeleton stream in the file it could be the same as http://trac.annodex.net/ticket/408 j
> was this all since 0.9.8?I think so, since I was using the oggz wrapper IIRC.> If it was a little before that, it may have been an instance of > http://trac.annodex.net/ticket/404 > > otherwise, if you do come across such a case please keep the file and > report it -- it's not easy to find corner cases.Sure. I didn't expect the crash to go away if rebuilding the file with the same input. We love Heisenbugs (not).