D14606: KCrash: DrKonqi cancelled = able to start...

rjvbb added a comment. OK, if no one else cares :) I told you how to reproduce the problem: do whatever it takes to make `cat /proc/sys/kernel/core_pattern` return a pipe command. Without that KCrash appears to think that you won't get a coredump (wrong...) and will exit when DrKonqi

D14606: KCrash: DrKonqi cancelled = able to start...

dfaure added a comment. I never said I don't care, I said: tell me how to reproduce this problem. REPOSITORY R285 KCrash REVISION DETAIL https://phabricator.kde.org/D14606 To: rjvbb, #frameworks, sitter Cc: dfaure, kde-frameworks-devel, michaelh, ngraham, bruns

D14606: KCrash: DrKonqi cancelled = able to start...

rjvbb added a comment. I'd still like that message not to appear if I just dismissed DrKonqi myself. But if no one else care about it I'll just patch the code for myself ... REPOSITORY R285 KCrash REVISION DETAIL https://phabricator.kde.org/D14606 To: rjvbb, #frameworks, sitter Cc:

D14606: KCrash: DrKonqi cancelled = able to start...

dfaure added a comment. We only stop printing it at the 4th recursion, I guess just in case that printing itself is the reason for the recursive crash... if (crashRecursionCounter < 4) { fprintf(stderr, "Unable to start Dr. Konqi\n"); } REPOSITORY R285 KCrash REVISION

D14606: KCrash: DrKonqi cancelled = able to start...

rjvbb added a comment. > I don't see why. We try to start drkonqi even the very first time. Yes, you'd print it if you're sure that first time failed (the start function got pid==0). I don't see the point in stopping to print it if you had a crash or two in the crash handler itself

D14606: KCrash: DrKonqi cancelled = able to start...

dfaure added a comment. I don't see why. We try to start drkonqi even the very first time. REPOSITORY R285 KCrash REVISION DETAIL https://phabricator.kde.org/D14606 To: rjvbb, #frameworks, sitter Cc: dfaure, kde-frameworks-devel, michaelh, ngraham, bruns

D14606: KCrash: DrKonqi cancelled = able to start...

rjvbb added a comment. > The crash recursion counter is increased when a SEGV happens inside the crash handler itself. Yeah, that would make sense. But then wouldn't the "unable to start" message have to be printed when the recursion counter is larger than 2? REPOSITORY R285 KCrash

D14606: KCrash: DrKonqi cancelled = able to start...

dfaure added a comment. The crash recursion counter is increased when a SEGV happens inside the crash handler itself. REPOSITORY R285 KCrash REVISION DETAIL https://phabricator.kde.org/D14606 To: rjvbb, #frameworks, sitter Cc: dfaure, kde-frameworks-devel, michaelh, ngraham, bruns

D14606: KCrash: DrKonqi cancelled = able to start...

rjvbb abandoned this revision. rjvbb added a comment. I'm abandoning this for now because my assumption was wrong even though my patch had the effect I intended. I'll reopen if/when I have a real fix, because there are still 2 issues here: - The inappropriate warning. This could be

D14606: KCrash: DrKonqi cancelled = able to start...

rjvbb added a comment. Actually, it's even worse than that: I didn't double-check my assumptions about the use of DrKonqi's return/exit code. Looking at the code again there is actually no way that information is even obtained. The `startProcess` function simply starts DrKonqi and then

D14606: KCrash: DrKonqi cancelled = able to start...

dfaure added a comment. In D14606#303654 , @rjvbb wrote: > I think there is no way at this level of interaction to determine if an application crashed after start and distinguish that from something like a "user-cancelled" return code, or is

D14606: KCrash: DrKonqi cancelled = able to start...

rjvbb added a comment. > > if DrKonqi gives an exit code that means it was started > > It could have crashed though... Yeah, that was meant as a snarky remark - even if it crashed it was still started :) I think there is no way at this level of interaction to determine if

D14606: KCrash: DrKonqi cancelled = able to start...

dfaure added a comment. $ /d/kde/inst/kde_frameworks/lib64/libexec/drkonqi 23:47:18.361 drkonqi(23113) org.kde.drkonqi: Invalid pid specified $ echo $? 1 > if DrKonqi gives an exit code that means it was started It could have crashed though... REPOSITORY R285

D14606: KCrash: DrKonqi cancelled = able to start...

rjvbb added a comment. > But there's a catch, on my system I get some plasma popup on which I have to click "Report Bug" in order to get the usual drkonqi dialog. Do you get that too? Usually, yes. If you're not quick enough you have to go find DrKonqi in the taskbar or even systray.

D14606: KCrash: DrKonqi cancelled = able to start...

dfaure added a comment. Tested with your `killall -SEGV kate` testcase, no bug here. But there's a catch, on my system I get some plasma popup on which I have to click "Report Bug" in order to get the usual drkonqi dialog. Do you get that too? Yes, I confirm that drkonqi exits normally,

D14606: KCrash: DrKonqi cancelled = able to start...

broulik added a reviewer: sitter. REPOSITORY R285 KCrash REVISION DETAIL https://phabricator.kde.org/D14606 To: rjvbb, #frameworks, sitter Cc: dfaure, kde-frameworks-devel, michaelh, ngraham, bruns

D14606: KCrash: DrKonqi cancelled = able to start...

rjvbb added a comment. > When I click "Close" in drkonqi (there's no Cancel button), crashtest resumes and exits, without any error message from KCrash. On my system, without the patch: > killall -SEGV kate > KCrash: Application 'kate' crashing... KCrash: Attempting to

D14606: KCrash: DrKonqi cancelled = able to start...

rjvbb added a comment. > When I click "Close" in drkonqi (there's no Cancel button), crashtest resumes and exits, without any error message from KCrash. On my system, without the patch: > killall -SEGV kate > KCrash: Application 'kate' crashing... KCrash: Attempting to

D14606: KCrash: DrKonqi cancelled = able to start...

dfaure added a comment. I can't reproduce what you describe using bin/crashtest in the drkonqi repository (after making sure KDE_DEBUG=1 isn't set, of course). When I click "Close" in drkonqi (there's no Cancel button), crashtest resumes and exits, without any error message from KCrash.

D14606: KCrash: DrKonqi cancelled = able to start...

rjvbb created this revision. rjvbb added a reviewer: Frameworks. Restricted Application added a project: Frameworks. rjvbb requested review of this revision. REVISION SUMMARY This patch addresses a minor detail that has been irking me for a while now: the message that DrKonqi couldn't be