[kig] [Bug 353897] New: occasional freeze of kig upon "Save as" button (losing all work)

[Maurizio Paolini]

https://bugs.kde.org/show_bug.cgi?id=353897

Bug ID: 353897
Summary: occasional freeze of kig upon "Save as" button (losing
all work)
Product: kig
Version: unspecified
Platform: Fedora RPMs
URL: http://dmf.unicatt.it/~paolini/kig/bugs/fuga.kig
OS: Linux
Status: UNCONFIRMED
Severity: critical
Priority: NOR
Component: general
Assignee: ***@computer.org
Reporter: ***@dmf.unicatt.it

Unfortunately the behaviour that I experience is occasional, and I have no way
to systematically reproduce it.
After working with kig, or even just after loading kig with an argument,
pressing the "Save As"
button (or pressing Ctrl+Shift+S) completely freezes the kig window. I could
only kill it by
pressing "Ctrl-C" on the console from which I started it.

(On my DELL laptop with Fedora 22 and a fresh installation from sources of kig)

Reproducible: Sometimes

Steps to Reproduce:
Start kig with an example:

$ kig fuga.kig [the url above points to the file given]
[<Enter> to quit the "tip of the day"]
Press: Ctrl-Shift-S


Actual Results:
Sometimes (for me a frequency like 1 every 2 or 3 tries) the kig window
completely freezes,
no button is responding (not even the "quit" button to quit the window).

When kit does work correctly, I quit kig and retry by invoking "$ kig fuga.kig"
again.


Expected Results:
A window should appear asking the file name

kig is compiled using the following cmake invocation:

$ cmake . -DCMAKE_INSTALL_PREFIX=/usr -DCMAKE_BUILD_TYPE=debugfull
-DPLUGIN_INSTALL_DIR:PATH=/usr/lib/qt5/plugins

[Rex Dieter]

https://bugs.kde.org/show_bug.cgi?id=353897

Rex Dieter <***@math.unl.edu> changed:

What |Removed |Added
----------------------------------------------------------------------------
CC| |***@math.unl.edu

[Maurizio Paolini via KDE Bugzilla]

https://bugs.kde.org/show_bug.cgi?id=353897

--- Comment #1 from Maurizio Paolini <***@dmf.unicatt.it> ---
I have news on this problem! I still do not know *what* triggers this
behaviou, however
a collegue of mine experienced the same on his workstation (fedora 22 with an
rpm installation of kig: Applications/15.04.03), whereas on my other
workstation I could never see the problem.

The "big" news is that actually it only seems to be a visualization problem:
kig does not freeze,
but behaves exactly as if the dialog window were present! So it is sufficient
to blindly write the
file name and press enter (or just press enter).
It seems something related to some missing "refresh" or some caching of
graphics primitives
that is not flushed before user input...

I tried to experiment with the kig sources, but actually I have no idea what is
the source file where to work.

[Maurizio Paolini via KDE Bugzilla]

https://bugs.kde.org/show_bug.cgi?id=353897

Maurizio Paolini <***@dmf.unicatt.it> changed:

What |Removed |Added
----------------------------------------------------------------------------
Severity|critical |major

--- Comment #2 from Maurizio Paolini <***@dmf.unicatt.it> ---
Since you don't actually lose data I decreased the severity of the bug

[Maurizio Paolini via KDE Bugzilla]

https://bugs.kde.org/show_bug.cgi?id=353897

--- Comment #3 from Maurizio Paolini <***@dmf.unicatt.it> ---
Additional info. There are now three computers where the problem is present.
Two
workstations and one laptop. All with fedora 22 installed, kig (and the whole
kde)
at version 15.04.3.

Actually, the problem seems to reside *outside* of kig, and precisely in the
QFileDialog::getSaveFileName

However I couldn't find any other kig application exhibiting the same
behaviour, so I am
not really sure whether perhaps there is a buffer overflow somewhere cousing
the problem
just as a collateral effect.

Resuming: the problem occurs occasionally (frequency about 10 to 25 percent).
In
case of problem: pressing <Ctrl><Shift>-S (save as) no window appears, however
it seems
that there is an invisible 'ghost' window that grabs keybord keystrokes (not
sure about the mouse, but the main window is apparently completely freezed),
since pressing <Alt>-C allows to regain control of the main kig window.

Same with other actions (e.g. "export").

Unfortunately I am still not able to deterministically trigger the problem.

If you are not able to reproduce the problem, do you have some suggestions on
how I
could get more info? Some "cmake" flag?

[Maurizio Paolini via KDE Bugzilla]

https://bugs.kde.org/show_bug.cgi?id=353897

--- Comment #4 from Maurizio Paolini <***@dmf.unicatt.it> ---
Additional info: I tried "xwininfo -name "Saving..." in the two situations:
when everything is fine and the Dialog window is open and when the problem
occurs, and everything seems freezed. In the latter case I get:

$ xwininfo -name "Saving..."
[...]
Map State: IsUnMapped
[...]

whereas when I see the window I get: Map State: IsViewable
So the Dialog window is there, but it is not mapped for some reason.
I suspect that the problem is not in kig itself but perhaps in qt; however kig
is the *only*
application for which I experience the problem, so I will insist posting here.

I also tried messing with qtbase sources (5.5.0), when everything freezes.
---- exerpt from qdialog.cpp --------
int QDialog::exec()
[...]
show();
printf ("I AM HERE!\n"); // <-- this is printed when the problem occurs
QPointer<QDialog> guard = this;
if (d->nativeDialogInUse) {
d->platformHelper()->exec();
} else {
QEventLoop eventLoop;
d->eventLoop = &eventLoop;
(void) eventLoop.exec(QEventLoop::DialogExec);
}

[Maurizio Paolini via KDE Bugzilla]

https://bugs.kde.org/show_bug.cgi?id=353897

--- Comment #5 from Maurizio Paolini <***@dmf.unicatt.it> ---
The problem apparently disappeared after an update of Fedora 22 from
kf5-*-5.14.* --> kf5-*-5.15.*

I don't know exactly which is the involved package because I cannot find the
'old' *-5.14.* fedora RPMs.

Perhaps we could close the bug report...

[Rex Dieter via KDE Bugzilla]

https://bugs.kde.org/show_bug.cgi?id=353897

Rex Dieter <***@math.unl.edu> changed:

What |Removed |Added
----------------------------------------------------------------------------
Resolution|--- |FIXED
Status|UNCONFIRMED |RESOLVED

--- Comment #6 from Rex Dieter <***@math.unl.edu> ---
OK, let's consider it fixed for now (closing)

feel free to reopen otherwise.