qtcreator crashes stopping the application

[JonB]

@abarmotov
Need to determine if @J-Hilk's

IIRC Creator sends SIGTERM and/or SIGKILL to the application process when that button is pressed

is right. Like I just wrote, put some fprintf(stderr) or qDebug() as first statement in handler.

BTW, do you want to show what code you actually do have in your signal handler(s)?

[abarmotov]

1. where is printf in handler
2. i will show code little latter, and attach to bugreport - there is no secret, but i need to remove excess code

[JonB]

@abarmotov said in qtcreator crashes stopping the application:

where is printf in handler

What do you mean by this? Unless you are ignoring the signals you have specified a function to be executed when the signals arrive. That is a "handler" function. You have some code in that function? I am suggesting you might put a fprintf(stderr, "Caught signal!\n") or qDebug() << "Caught signal!" as the first statement in whatever you have, so that we might see whether it is ever hit (in case breakpoints aren't working right).

[abarmotov]

1. create new app in creator
2. add some little code "signalhandler"
3. run it Ctrl-R
   4.1 kill from creator
   4.2 kill from outside konsoles

#include "mainwindow.h"
#include <QApplication>
#include <unistd.h>
#include <signal.h>

void signalhandler(int sig) {
	printf("SIG = %i\n", sig);
	int ExitHelperCode = 0;
	if (sig == SIGINT || sig == SIGTERM) ExitHelperCode = 123;
	else if (sig == SIGUSR1) ExitHelperCode = 124;
	qApp->exit(ExitHelperCode);
}

int main(int argc, char *argv[])
{
	signal(SIGINT,  signalhandler);
	signal(SIGTERM, signalhandler);
	QApplication a(argc, argv);
	MainWindow w;
	w.show();
	return a.exec();
}

[JonB]

@abarmotov

• Did you ever get to see the output from the printf("SIG = %i\n", sig); which is there?

• Comment out the qApp->exit(ExitHelperCode);. Does that have any effect on the "crash" you report? And btw how do you know your app was "crashing", given that it was supposed to exit?

It all still boils down to whether that signal handler function is being hit at all when you press the "Stop" button from Creator.

You might also try

signal(SIGINT,  SIG_IGN);
signal(SIGTERM, SIG_IGN);

instead as your first two lines in main().

[abarmotov]

1. yes , i saw "SIG = 15", then kill app from from console
2. myapp was "crashing" - that qtcreator says in output window

[JonB]

@abarmotov
Yes, that's SIGTERM. I think you are saying it hit your printf("SIG = %i\n", sig); line, so we know it went into sighandler()? Then like I said try commenting out the qApp->exit(ExitHelperCode);, it may not be possible/safe to call that from within a signal handler.

[andr]

When trying to stop the application, the result is set to QProcess::CrashExit temporarily and updated whenever your process comes back with a proper exit. Your process doesn't come back with something better, so you get the "%1 crashed" response.

I don't really see this as a bug, at best the exit code should perhaps be part of the message.

[abarmotov]

hm, if qApp->exit(ExitHelperCode) is not the safe - that it should to be ?
that code works fine before qtcreator v.7 and works fine in my app as ubuntu service

[JonB]

@abarmotov
That is not the current question. The question is: does the crash, misbehaviour or whatever the issue is go away if you remove that line? By now you could have given us that answer :)

[abarmotov]

@abarmotov said in qtcreator crashes stopping the application:

if i remove qApp->exit, then:

1. SIG = 15 and SIG = 2 printed IF i send that signals from console (or from kde task manager), but app continue working
2. from creator "red" button crash my app still

i will try to check this behavior on another machine from new user with clean configs ...

[abarmotov]

fresh install on new user - same behavior
so need to know which signal creator sends to app on press "red" button - KILL, TERM, INT ...

[JonB]

@abarmotov
I would suggest likely SIGTERM under Linux, as per https://stackoverflow.com/questions/14137808/signals-sent-by-qtcreator-on-stop from 10 years ago....

UPDATE
I was going to try your issue. Very unfortunately/scarily, the VirtualBox VM I use for Ubuntu has encountered a "critical error" while it was applying the latest Ubuntu patches (which I always do when firing up), and now will not get me into Ubuntu at all, and I don't know what to do. This has never happened before, and seems to leave me with an unrecoverable box :@ I am not a happy bunny.... Will get back to you if/when I can resolve this....

[abarmotov]

how to run creator with itself verbose output to see his logs ?

[JonB]

This post is deleted!

[JonB]

@abarmotov
I have now recovered my Ubuntu VM successfully --- phew! :)

I have pasted exactly your program. Run from Creator, press red Stop button. All is well, no crash. Application Output window:

13:48:42: Starting /home/jon/QtTests/build-signals-Desktop-Debug/signals ...
SIG = 15
13:48:48: /home/jon/QtTests/build-signals-Desktop-Debug/signals exited with code 123

As you can see it received a SIGTERM, it was handled and it exited cleanly.

Purely OOI, if run under Debug and click red Stop button program exits completely, cleanly, no crash, but does not visit the signal handler function. In this case the debugger is handling the Stop and doing whatever it does to stop its inferior process (the program being debugged) without sending it the SIGTERM signal.

I am Ubuntu 20.04, Qt 5.12.5 (as supplied with Ubuntu) and Qt Creator 4.11.0. I therefore conclude that this is a Creator 7.x issue. Which does not surprise me....

[abarmotov]

@JonB yes, last creator crashing my app
good tested on 4.15.2 - "exited with code 123"
made bugreport https://bugreports.qt.io/browse/QTCREATORBUG-27619

[JonB]

@abarmotov
OK, I had a read there, and posted my own finding.

The responder there says it does not happen for him on Creator 7.0.2, which is what you have? Without them being able to repro nothing can be done. If you have not done so already, I would uninstall Creator completely and re-install. Maybe try a different version, like another 7.x or a 6.x? And if you have access to another machine that might help too.

[abarmotov]

i have 7.0.2, 6.0.2 - both crashing

[JonB]

@abarmotov
I cannot comment, but you can see the person who responded to your bug report says no problem under 7.0.2. You need someone/something reproducible. [Nonetheless: report your finding for 6.0.2. too, yet not for 4.15.2, in that bug report.]

It could be something in your general "environment". If you are able to try to repro on another (preferably vanilla) machine or VM that would help your cause greatly....