Bug reporting and feature requests are managed with GNOME's GitLab. You need to create an account to file issues and comment on them. Take a quick look at the existing list of bugs and feature requests to see if your problem has already been reported.
To report a bug/problem in the software, create an issue. Ideally you can describe exactly the steps anyone can follow to reproduce the bug. The more details, the better.
Providing debugging information
Sharing sample files, projects, and “scenarios”
To be able to reproduce an issue, we might ask you to share sample media files with us. If the file is too large to attach to the GitLab issue, you can use for example Dropbox, Google Drive, MEGA or other service to share such media.
You can also share in a similar way a project archive containing the project and all the media is uses:
- Use the “Select unused clips” feature to easily remove unused media from your project, this will help you save space (and upload time).
- Click the menu button top-right and choose the “Export project as
tarball...” menu item. Save the
.xges_tarfile somewhere. It will contain your project file and its associated media.
- Upload it as described above.
In addition to the project archive, it is extremely helpful to provide “scenario” files. These are automatically generated each time you use a project and contain the operations you made. Combined with the project archive, these allow us to perform exactly the actions that have occurred to trigger the bug. This makes reproducing the issue on our machines a very easy and reliable process, which saves you a ton of time! Here's how to provide scenario files to facilitate the process:
- Save your project, right before triggering the bug.
- Trigger the bug (make Pitivi crash or freeze).
- Get the last/newest scenario file from
- Upload it as described above, so we can reproduce your issue and integrate it into our test suite so that it does not happen again in the future!
Back traces for crashes and deadlocks
When reporting a crash (application window disappears) or a deadlock (application is frozen), we can't do much without a back trace.
First try to see if you can locate a coredump file created by your system automatically when a crash takes place. For example:
$ coredumpctl list | tail Wed 2019-08-28 23:02:20 CEST 31783 1000 100 11 present /usr/bin/python3.7 $ coredumpctl info 31783 PID: 31783 (python3) Storage: /var/lib/systemd/coredump/core.python3.1000.e907bb24f9c14aafb3ec0c900ee5bc4a.31783.1567026134000000.lz4 $ lz4 -d /var/lib/systemd/coredump/core.python3.1000.e907bb24f9c14aafb3ec0c900ee5bc4a.31783.1567026134000000.lz4 ~/coredump
A coredump can be investigated using gdb. Look below for the proper way
to start gdb, but at the end instead of
gdb python3 -ex ... simply run
gdb python3 ~/coredump.
Alternatively, if you are missing a coredump, start Pitivi in gdb as described below, then try to reproduce the crash.
Finally, in gdb run
bt full to get the back trace for the crash.
Tip: To avoid the need to press Enter to “scroll” in gdb, run
set pagination 0.
For a deadlock, start Pitivi in gdb as described below, press Ctrl+Z
thread apply all bt to get the backtraces for all the threads.
When running in the development environment
Install the GNOME SDK Debug symbols and update them, see below.
Enter the sandbox:
- Start Pitivi inside gdb:
gdb python3 -ex "run $PITIVI_REPO_DIR/bin/pitivi"
When running with Flatpak
- Install the GNOME SDK and its Debug symbols and update them:
flatpak --user install flathub org.gnome.Sdk/x86_64/42 flatpak --user install flathub org.gnome.Sdk.Debug/x86_64/42 flatpak --user update org.gnome.Sdk/x86_64/42 flatpak --user update org.gnome.Sdk.Debug/x86_64/42
- Start a shell in the Pitivi flatpak sandbox:
flatpak run -d --command=bash org.pitivi.Pitivi
- Start Pitivi inside gdb:
gdb python3 -ex "run /app/bin/pitivi"
When running from the packages of your Linux distro
GNOME's Getting Stack Traces has excellent documentation and tips on the subject, including how to install the relevant debug packages. Below is a quick reminder for those already familiar with the process.
When you want to “attach” to an existing Python process (useful for deadlocks, where the application will be hung instead of crashed):
gdb python3 THE_PITIVI_PROCESS_NUMBER
When you want to run Pitivi entirely in gdb from the start:
gdb python3 -ex "run $(which pitivi)"
When you need to know what’s going on inside Pitivi, you can launch it with a debug level. In loggable.py, there are six levels: ( ERROR, WARN, FIXME, INFO, DEBUG, LOG ) = range(1, 7). As such, if you want to see errors and warnings only, you launch
...and if you want to see everything you do
If that's “too much” and you want to focus on particular parts of the
code, you can do so. For example, you can get output from the
MediaLibraryWidget classes only:
Here are various examples of commands you can use to generate detailed debug logs that include not only Pitivi's debug output, but also GStreamer's:
A basic log can be obtained by running:
PITIVI_DEBUG=*:5 GST_DEBUG=2 pitivi > debug.log 2>&1
To get debugging information from Non-Linear Engine, you could use:
PITIVI_DEBUG=5 GST_DEBUG=3,nle*:5,python:5 pitivi > debug.log 2>&1
The information most likely to be useful would probably be the debug info from GES in addition to Pitivi's:
PITIVI_DEBUG=5 GST_DEBUG=ges:5 pitivi > debug.log 2>&1;
When using GST_DEBUG, the resulting logs will most likely be too big to be attached to a bug report directly. Instead, compress them (in gzip, bzip2 or lzma format) before attaching them to a bug report.
Python performance profiling
In the rare cases where a performance problem is caused by our UI code,
you can profile Pitivi itself, with this command (and yes,
JUMP_THROUGH_HOOPS is needed for this case, it is an environment
JUMP_THROUGH_HOOPS=1 python3 -m cProfile -s time -o pitivi_performance.profile bin/pitivi
pitivi_performance.profile file can then be processed
to create a visual representation of where the most time was spent and
which functions were called the most often in the code. See also Jeff's
blog posts on profiling.
The results of the search are