sporadic CRASH on exit in QNetworkConfigurationManagerPrivate::pollEngines() on unixes (corrupt stack?)
Sorry I can't make the description more accurate. This has been occuring to me a fair number of times (but not every time) I close musescore on my ArchLinuxARMv7A C201, and I've caught it in QtCreator debug while running latest e666061, so I'm reporting it now. I close MScore normally and get a Segmentation Fault (Signal SIGSEGV) inside QNetworkConfigurationManagerPrivate::pollEngines() at the final line this dissasembler (sorry I need to insall Qt code on my machine!!):
0xb3fec708 f0 4f 2d e9 push {r4, r5, r6, r7, r8, r9, r10, r11, lr} 0xb3fec70c <+0x0004> 10 80 90 e2 adds r8, r0, #16 0xb3fec710 <+0x0008> d4 d0 4d e2 sub sp, sp, #212 ; 0xd4 0xb3fec714 <+0x000c> 00 40 a0 e1 mov r4, r0 0xb3fec718 <+0x0010> 02 00 00 0a beq 0xb3fec728 <_ZN35QNetworkConfigurationManagerPrivate11pollEnginesEv+32> 0xb3fec71c <+0x0014> 08 00 a0 e1 mov r0, r8 0xb3fec720 <+0x0018> 01 80 88 e3 orr r8, r8, #1 0xb3fec724 <+0x001c> c8 46 fe eb bl 0xb3f7e24c <_ZN6QMutex4lockEv@plt> 0xb3fec728 <+0x0020> c0 30 8d e2 add r3, sp, #192 ; 0xc0 0xb3fec72c <+0x0024> 14 10 84 e2 add r1, r4, #20 0xb3fec730 <+0x0028> 5c 30 8d e5 str r3, [sp, #92] ; 0x5c 0xb3fec734 <+0x002c> 03 00 a0 e1 mov r0, r3 0xb3fec738 <+0x0030> 3c e4 ff eb bl 0xb3fe5830 0xb3fec73c <+0x0034> c0 00 9d e5 ldr r0, [sp, #192] ; 0xc0 0xb3fec740 <+0x0038> 01 20 a0 e3 mov r2, #1 0xb3fec744 <+0x003c> 10 10 80 e2 add r1, r0, #16 0xb3fec748 <+0x0040> 08 30 90 e5 ldr r3, [r0, #8] 0xb3fec74c <+0x0044> 0c 00 90 e5 ldr r0, [r0, #12] 0xb3fec750 <+0x0048> cc 20 8d e5 str r2, [sp, #204] ; 0xcc 0xb3fec754 <+0x004c> 03 31 81 e0 add r3, r1, r3, lsl #2 0xb3fec758 <+0x0050> c4 30 8d e5 str r3, [sp, #196] ; 0xc4 0xb3fec75c <+0x0054> 00 11 81 e0 add r1, r1, r0, lsl #2 0xb3fec760 <+0x0058> c8 10 8d e5 str r1, [sp, #200] ; 0xc8 0xb3fec764 <+0x005c> 01 00 53 e1 cmp r3, r1 0xb3fec768 <+0x0060> 78 00 00 0a beq 0xb3fec950 <_ZN35QNetworkConfigurationManagerPrivate11pollEnginesEv+584> 0xb3fec76c <+0x0064> 00 00 93 e5 ldr r0, [r3] 0xb3fec770 <+0x0068> 00 00 52 e3 cmp r2, #0 0xb3fec774 <+0x006c> 7c 72 9f e5 ldr r7, [pc, #636] ; 0xb3fec9f8 <_ZN35QNetworkConfigurationManagerPrivate11pollEnginesEv+752> 0xb3fec778 <+0x0070> 02 50 a0 e1 mov r5, r2 0xb3fec77c <+0x0074> 1c 60 84 e2 add r6, r4, #28 0xb3fec780 <+0x0078> 07 70 8f e0 add r7, pc, r7 0xb3fec784 <+0x007c> 64 00 8d e5 str r0, [sp, #100] ; 0x64 0xb3fec788 <+0x0080> 6a 00 00 0a beq 0xb3fec938 <_ZN35QNetworkConfigurationManagerPrivate11pollEnginesEv+560> 0xb3fec78c <+0x0084> 00 30 90 e5 ldr r3, [r0] 0xb3fec790 <+0x0088> 40 30 93 e5 ldr r3, [r3, #64] ; 0x40
I should also note that on starting ./build.release/mscore/mscore, I get this following message that I suspect may be related directly to this error, because both QNetwork:
QNetworkReplyImplPrivate::error: Internal problem, this method must only be called once.
But I'm unable to reproduce a subsequent execution, so clearly this is a transient error.
Comments
I often see the QNetworkReplyImplPrivate error in the terminal, and recently I have been getting a seg fault on exit too. I haven't seen any problems inside MuseScore though; everything appears to work fine.
ok. Yes, I don't see any errors related to that while running musescore. But a crash on exit is still a crash and a potential source or indicator of bad things, so I'm going to keep my eye out. Which is why I've set priority to normal, not critical.
what distoro and cpu are you on?
Ubuntu 15.10 on x86_64.
From what I understand, many of us runnng on Linux seeng this semi-regularly.
See the QNetworkReplyImplPrivate::error at mscore start in console?
Or see the QNetworkConfigurationManagerPrivate::pollEngines() when quit mscore?
If this is same, then we can say is not just an ArchLinuxARMv7 issue, but a linux CPU-independent issue.
I have seen the crash on exit quite a few times before on my other linux, but I never bothered to investigate, nor have caught it in QtCreator.
The crash in QtNetwork... on exit. I see it maybe 10-20% of all rns nder the debugger. Perhaps in runs not under the debugger as well, but it isn't as obvious that it crashed. I als see all sorts of things printed to the console on startup, but I assume that's normal.
OK. I'm changing title to reflect that this is a general linux issue.
Googling for QNetworkConfigurationManagerPrivate::pollEngines segfault reveals that a lot of other projects outside of MuseScore experience this same sporadic issue:
https://lists.debian.org/debian-qt-kde/2015/11/msg00221.html
Apparently something was fixed in Qt 5.4.2 https://codereview.qt-project.org/#/c/111893/ but I'm using 5.5.1 so apparently that didn't fix it completely, or it is a separate issue. Should I try contacting Qt?
regarding the QNetworkReplyImplPrivate::error on startup, I'm noticing on my x86-64 Win 8.1 machine running master that I'm also getting that:
I still have no idea whether these two errors are related in any way.
I believe the warning at startup is a problem with Qt Webkit. If you run mscore with -w to get no webview in the start center, it should disappear.
The crash I'm not sure but I do see it very very occasionally on Mac OSX in debug mode.
Thanks!
Has anyone experienced this crash on Windows?
I've experienced the crash on my i686 arch linux machine running official arch build of mscore 2.0.2. All I did was open mscore, edit preferences, close mscore, and then I see the telltale:
I've attached output of "gdb /usr/bin/mscore -coredump-" with "thread apply all bt" but I think the interesting bit is:
I've added "(corrupt stack?)" to this bug title.
I've saved this coredump and backtrace (and will save future coredumps) to: https://efjz.in/index.php/s/DllKY9veK8m2FAP
EDIT: I'll try what lasonic says to diable webkit with "-w", and I'll see if I still get any crash on exit. I'm noting that I at least no longer see the "QNetworkReplyImplPrivate::error".
Running shoogle's latest nightly i686 AppImage (without encountering crash), I notice console output:
So maybe there is some issue connecting something. (I think those connect messages are referring to singals & slots...someone correct me if mistaken)
APPENDUM: note that the first "Cannot connect" error appears when start screen first displays, but at that point no "Mastering MuseScore" visual appears, untill a couple seconds later, at which point the rest of those console messages appear all at once.
In windows 8.1 x86-64 git master I do receive on startup:
So clearly that issue is on all platforms...I'm going to try to fix that...