Fix #1333: arch Linux build fails

[aryoda]

I have fixed all known issues on our side an try to contact the package maintainer now asking

• to fix the missing dependencies
• to reconsider setting LC_ALL=C for makeing

On my arch VM the build & unit tests are working now (after manually installing the required dependencies)...

Fixes #1333

[buhtz]

I cloned directly from "aryoda:issue/1333_Arch_build_fails".

Testing on Debian 11 (stable) via ./configure && make && make install && make test runs fine (even in an SSH terminal without X11 or Wayland involved).

I also used my Manjaro VM for testing. When I'm on a temrinal emulator in the desktop environment make test is fine.
But when I log into that same machine via SSH in a real terminal (without X11/Wayland involved) I got an error about QPainter which polutes the output and make a test fail.

I was wondering why Qt is involved here. But I saw that Qt is imported in diagnostics.py.

user@Alf ~/sfd/backintime/common $ make test
python3 -m unittest  -b test/test_applicationinstance.py
...................
----------------------------------------------------------------------
Ran 19 tests in 0.048s

OK
python3 -m unittest  -b test/test_argparser.py
.....................
----------------------------------------------------------------------
Ran 21 tests in 3.355s

OK
python3 -m unittest  -b test/test_backintime.py
WARNING: Failed to connect to Udev serviceHelper daemon via D-Bus: org.freedesktop.DBus.Error.Spawn.ChildExited
WARNING: D-Bus message: Launch helper exited with unknown return code 1
WARNING: Udev-based profiles cannot be changed or checked due to Udev serviceHelper connection failure
.Fs
======================================================================
FAIL: test_local_snapshot_is_successful (test.test_backintime.TestBackInTime)
end to end test - from BIT initialization through snapshot
----------------------------------------------------------------------
Traceback (most recent call last):
  File "/media/sf_Downloads/backintime/common/test/test_backintime.py", line 146, in test_local_snapshot_is_successful
    self.assertRegex(filtered_log_output, re.compile(r'''INFO: Lock
AssertionError: Regex didn't match: 'INFO: Lock\nINFO: Take a new snapshot. Profile: 1 Main profile\nINFO: Call rsync to take the snapshot\nINFO: Save config file\nINFO: Save permissions\nINFO: Create info file\nINFO: Unlock\n' not found in 'INFO: Lock\nINFO: Take a new snapshot. Profile: 1 Main profile\nINFO: Call rsync to take the snapshot\nINFO: Save config file\nINFO: Save permissions\nINFO: Create info file\nQPainter::begin: Paint device returned engine == 0, type: 2\nQWidget::render: Cannot render with an inactive painter\nINFO: Unlock\n'

----------------------------------------------------------------------
Ran 3 tests in 4.166s

FAILED (failures=1, skipped=1)
make: *** [Makefile:570: unittest] Fehler 1

[aryoda]

Very good test, this looks like a new bug.

But when I log into that same machine via SSH in a real terminal (without X11/Wayland involved)
I got an error about QPainter which polutes the output and make a test fail.

Did I introduce this with my PR or does it happen with master too?

I can't see a line number for this QPainter message so I need a way to reproduce this.
Do you have a minimal setup to reproduce this (eg. without ssh?).
Meanwhile I will try to boot Linux into a console instead of a desktop environment, perhaps this is showing the message too...

I was wondering why Qt is involved here. But I saw that Qt is imported in diagnostics.py

Yes, diagnostics does show the PytQt5 version number, but only when "--diagnostics" is called.
And this is the most plausible reason since some unit tests do use "--diagnostics".

backintime/common/diagnostics.py

Lines 121 to 130 in 2121c60

	# Qt
	try:
	import PyQt5.QtCore
	except ImportError:
	qt = '(Can not import PyQt5)'
	else:
	qt = 'PyQt {} / Qt {}'.format(PyQt5.QtCore.PYQT_VERSION_STR,
	PyQt5.QtCore.QT_VERSION_STR)
	finally:
	result['python-setup']['qt'] = qt

Perhaps it would be better to use PyQt5 only if --diagnostics was started from backintime-qt
(I saw a "flag" to recognize this in the code - I am searching...)

Could you please do a quick check via the console that printed the "QPainter" message:

python3 -c "import PyQt5.QtCore; print(PyQt5.QtCore.PYQT_VERSION_STR)"

If this also writes the QPainter message we would have found the culprit.

[aryoda]

When I boot into a Manjaro console via run-level 3 in grub the same unit test fails but with another actual/expected difference:
One line too much in the output:

WARNING: Inhibit Suspend failed.

If I ignore this warning by filtering it like the DBus warnings (test_backintime staring line 147) the unit tests succeed.

Edit: This warning is shown when the DBus communication with a service does fail (I bet the service is only running when a desktop environment is running). I will add this warning to the filter...

I could not yet reproduce the above mentioned QPainter messages (I am seaching the web for the same message ATM):

QPainter::begin: Paint device returned engine == 0, type: 2
QWidget::render: Cannot render with an inactive painter

[buhtz]

I'm still confused and not sure if "my problem" is relevant here or should block the merge.
I tend to merge that PR and see what happens in the wildlife out there. 🐅

I agree, but give me one more day to try to reproduce the issue in a new VM since it may
hurt using BiT in CLI-only mode on "embedded" devices.

The error happens in a unittest where --diagnostics isn't called. That confuse me, too.

Yes, you are right. --diagnostics was the only place where I could see the import PyQt5.
But I have a new suspect: The pluginmanager! It seems to load all plugins (even from the qt/plugins folder)

How I reproduce
pytest test/test_backintime.py::TestBackInTime::test_local_snapshot_is_successful
It is Manjaro on a tty terminal, on a ssh terminal and on a tmux terminal.

I am setting up a new console-only VM to test this. It looks like a plain backintime check-config could suffice to trigger the QPainter output...

Do you get the QPainter message also when running BiT master instead of my PR version?

[buhtz]

Btw: It seems that you edited my own comment instead of quote and answer it. ;)

Do you get the QPainter message also when running BiT master instead of my PR version?

Yes I can reproduce this with master also. This indicates to me that it isn't a serious problem because it was there before but no one noticed or reported about it.

[aryoda]

Btw: It seems that you edited my own comment instead of quote and answer it. ;)

Damn it, I shouldn't do two things in parallel :-)

Do you get the QPainter message also when running BiT master instead of my PR version?

Yes I can reproduce this with master also. This indicates to me that it isn't a serious problem
because it was there before but no one noticed or reported about it.

OK, then I suggest to merge now and I do the analysis until EoW.