Null pointer deref in mlt_deque_allocate (self=self@entry=0x0) at mlt_deque.c:92

Brought to you by: ddennedy, lilo_booter

#181 Null pointer deref in mlt_deque_allocate (self=self@entry=0x0) at mlt_deque.c:92

Milestone: head

Status: closed

Owner: nobody

Labels: None

Priority: 5

Updated: 2013-02-03

Created: 2012-12-19

Creator: Marcin Owsiany

Private: No

This is a 100% reproducible crash with 0.8.6-1
It happened to me first when using mlt embedded in openshot but I managed to reproduce it with standalone melt.

Attached:

xml file (exported from openshot) that melt crashes on. Unfortunately the referenced input files are on the order of 1GB in total, but I hope the project files gives you some idea of the transforms involved, etc.
typescript made with "-silent -debug"
stack traces from the core file (Thread 1 seems to be the one segfaulting)

I'm more than happy to do additional debugging if someone gives me some hints - I know C and can use gdb, but have no experience with melt nor video codecs.

1 Attachments

irlandia-zachodnia.xml

Discussion

Marcin Owsiany - 2012-12-19

typescript made with "-silent -debug"

typescript2.bz2

If you would like to refer to this comment somewhere else in this project, copy and paste the following link:

Marcin Owsiany - 2012-12-19

stack traces from the core file (Thread 1 seems to be the one segfaulting)

gdb2.txt

If you would like to refer to this comment somewhere else in this project, copy and paste the following link:

Marcin Owsiany - 2012-12-19

FTR, the original report against openshot is at https://bugs.launchpad.net/openshot/+bug/1085727

If you would like to refer to this comment somewhere else in this project, copy and paste the following link:

Dan Dennedy - 2012-12-22

status: open --> pending
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:

Dan Dennedy - 2012-12-22

I am quite sure this was due to a regression introduced in version 0.8.2. I have traced it down, and the change was not critical - an erroneous attempt to plug a minor, no-longer-reproducible memory leak. This is now applied to git.

Here is a link to a openshot build script that will build MLT, and OpenShot from the head of their repositories if you want to try to verify it:
https://raw.github.com/mltframework/mlt-scripts/master/build/build-openshot.sh

Honestly, I did not try to reproduce it and confirm it with your XML. I just took a look at the backtrace and analyzed the code and changes, and the problem quickly jumped out at me.

Last edit: Dan Dennedy 2012-12-22

If you would like to refer to this comment somewhere else in this project, copy and paste the following link:

Marcin Owsiany - 2012-12-22

Thanks for the fast reply.
Is there a particular git commit that is supposed to fix it, that I can try patching in, rather than building from HEAD?

If you would like to refer to this comment somewhere else in this project, copy and paste the following link:

Dan Dennedy - 2012-12-22

yes, there are 2:

decb908499e2499bf845107a35bed3b5ed2802fb

b959dc0bb82bd019601eb45b2f87474c4e812250

If you would like to refer to this comment somewhere else in this project, copy and paste the following link:

Dan Dennedy - 2012-12-22

milestone: feature_request --> head
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:

Marcin Owsiany - 2012-12-25

I can confirm that with these two patches my video renders without a crash.
Many thanks for the fast turnaround.

If you would like to refer to this comment somewhere else in this project, copy and paste the following link:

Dan Dennedy - 2012-12-26

status: pending --> accepted
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:

Dan Dennedy - 2013-02-03

status: accepted --> closed
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:

Log in to post a comment.