| Age | Commit message (Collapse) | Author |
|---|
|
|
|
|
|
The previous "fix" attempt with QT_GRAPHICSSYSTEM=native apparently
wasn't fixing the issue consistently.
This extra env var works 100% of the times by my tests though.
- Resolves: #8028
|
|
- Resolves: #8028
|
|
|
|
|
|
according to [1] the backend should raise a Runtime Error, instead what happens is that the process is spawned again and again but never runs actual code.
[1] https://docs.python.org/2.7/library/multiprocessing.html#multiprocessing.freeze_support
|
|
the original implementation causes problems in win32 (see #7240) possibly because pyinstaller cannot resolve the lambda correctly. using the documented multiprocessing.Process approach to pass parameters works here
|
|
- Related: #7512
|
|
|
|
this is a bug in macholib, there's a missing letter in the plugins path, so
unless we fix this they cannot be loaded from the bundle.
See:
https://bitbucket.org/pydica/pyside-setup/commits/4b8be97e5a00b577fe30ce9aa7e5723ff2a66f94
Quoting from http://code.activestate.com/lists/pythonmac-sig/23278/:
"""
The problem might be this line:
@rpath/Contents/mageformats/libqtiff.dylib
The "i" from "imageformats" is missing!
This _might_ be related to the unusual case that "libqtiff" has no path
at all, or something else
is funny, and we end up with a name that will not be found at all.
Then the loader finds the plugin in the installed Qt, which causes it to
load everything
again from there.
"""
|
|
due to zmq, we are hitting the limit and getting app crashes.
- Resolves: #7319
|
|
if the lsb_release utility is present on system, log part of its output
so we can have more information about the platform bitmask is running
on.
- Resolves: #7162
- Releases: 0.9.0
|
|
Rotate bitmask.log file on each start.
|
|
API to get the children has changed on latest psutil, this takes care
of it.
|
|
Replace logging.getLogger with custom get_logger.
Remove some unneeded dependencies and reorder other.
|
|
- Add a new handler for a zmq/thread-safe log send between components.
- Update the log window to use this new handler.
- Remove old custom handler
We have implemented a new handler that uses logbook, so this one is no
longer needed.
- Adapt log silencer to new handler
- Use log file always as default
- Related: #6733
|
|
If the events server is initialized in a different process than the backend,
the txzmq socket raises an "zmq.error.ZMQError: Interrupted system call"
exception during the events server initialization. Despite that, communication
seems to work flawlessly after the initialization.
Moving the events server initialization to the same process as the backend
causes the exception to not be raised during events server intialization.
|
|
- Related: #6359
|
|
- Resolves: #6426
- Resolves: #6681
|
|
|
|
Fix typo for signal disconnection.
The backend is the one who always creates the certificates. Either if it
is run separately or in a process in the same app as the frontend.
|
|
Add the 'check_online' method to check whether the backend is accessible
or not.
Reduce the wait for running threads timeout on quit.
Add retry feature to the backend requests send.
|
|
|
|
|
|
Stop the backend if the frontend process does not exist any more and
backend is not a daemon.
|
|
|
|
|
|
Running the GUI in a child process gives problems on OSX.
Also, change signal handling since we have less processes.
|
|
|
|
Add serialize/deserialize to dict helper.
|
|
|
|
|
|
|
|
This prepares the scenario to run the frontend and the backend in
different processes.
|
|
because the fork happens after the registration of the signal handler,
all children processes were inheriting the handler (and being passed a
reference to the mainwindow object, ugh...)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
Refactor logic from backend to the vpnprocess.
|
|
SIGTERM is caught by twisted, so we have to use addSystemEventTrigger()
to pass a custom handler to twisted reactor.
See https://leap.se/code/issues/5672 for more.
|
|
|
|
|
|
Closes #4990.
|
