[osg-users] Re: OpenSceneGraph-3.6.5 release candidate 2 tagged, please test
Am Montag, 27. Januar 2020 10:57:23 UTC+1 schrieb Robert Osfield: > > Hi Fiabian, > > On Monday, 27 January 2020 09:41:43 UTC, Fabian Roth wrote: >> >> Hi, >> I am currently testing this RC with openmw. >> If i have the fps display or profiler open while exiting the application >> i get a crash on exit. >> I am not sure if this is due to a bug in my build, a bug in openmw or a >> real issue with osg. >> The issue seems to be related to the destruction of the default font, but >> i was not able to investigate further. >> Attached is a Backtrace of the issue i am currently observing. >> > > That stack trace looks like the automatic clean up of the ObjectCache with > the DefaultFont within it is related somehow to the crash. How the > DefaultFont is managed has changed, to address bugs ironically, and in a > general sense the clean up the stack trace looks just fine to me, it's > roughly what I'd expect to happen. However, I don't have any clear idea > why in this instance the crash has occurred. > > Given there isn't any obvious amiss we are unfortunately are left to widen > out the search for what is amiss. > > Does running an OSG example like osgtext fail? > > Do others in the OpenMMW community seen this same crash? > > Is it possible to run OpenMMW single threaded to see if there might be > some thread interaction? > > What OS/compile and OpenMMW version combination are you using? > > One possible cause of crash like this is memory corruption during the run > of the application that is only revealed on clean up. Using a memory tools > like valgrind might be spot this type of issue. > > Perhaps others might have seen something similar and can help shed some > light on the nature of the crash. > > If it's possible to recreate the crash with an standard OSG example, or a > small modification of one, then this would be really helpful for me to jump > in a start investigating the issue. > > Cheers. > Robert. > Hi, Thank you for the fast reply. My build is using static osg, static osg-plugins and link time optimization. I created an address sanitizer enabled build. It exhibits a heap-use-after-free. I will try to further investigate this week. Greetings, Fabian = ==11872==ERROR: AddressSanitizer: heap-use-after-free on address 0x603082c0 at pc 0x55b4b9659551 bp 0x7ffdf8a9c190 sp 0x7ffdf8a9c180 READ of size 8 at 0x603082c0 thread T0 #0 0x55b4b9659550 in OpenThreads::ScopedPointerLock::ScopedPointerLock(OpenThreads::Mutex*) ./openmw/extern-git/OpenSceneGraph/include/OpenThreads/ScopedLock:54 #1 0x55b4b9659550 in osg::StateAttribute::removeParent(osg::StateSet*) ./openmw/extern-git/OpenSceneGraph/src/osg/StateAttribute.cpp:38 #2 0x55b4b965a033 in osg::StateSet::clear() ./openmw/extern-git/OpenSceneGraph/src/osg/StateSet.cpp:734 #3 0x55b4b965a5ef in __base_dtor ./openmw/extern-git/OpenSceneGraph/src/osg/StateSet.cpp:285 #4 0x55b4b965a9f8 in __deleting_dtor ./openmw/extern-git/OpenSceneGraph/src/osg/StateSet.cpp:286 #5 0x55b4b9c98246 in osg::Referenced::unref() const ./openmw/extern-git/OpenSceneGraph/include/osg/Referenced:201 #6 0x55b4b9c98246 in osg::ref_ptr::~ref_ptr() ./openmw/extern-git/OpenSceneGraph/include/osg/ref_ptr:41 #7 0x55b4b9c98246 in void std::_Destroy >(osg::ref_ptr*) /usr/include/c++/7/bits/stl_construct.h:98 #8 0x55b4b9c98246 in void std::_Destroy_aux::__destroy*>(osg::ref_ptr*, osg::ref_ptr*) /usr/include/c++/7/bits/stl_construct.h:108 #9 0x55b4b9c98246 in void std::_Destroy*>(osg::ref_ptr*, osg::ref_ptr*) /usr/include/c++/7/bits/stl_construct.h:137 #10 0x55b4b9c98246 in void std::_Destroy*, osg::ref_ptr >(osg::ref_ptr*, osg::ref_ptr*, std::allocator >&) /usr/include/c++/7/bits/stl_construct.h:206 #11 0x55b4b9c98246 in std::vector, std::allocator > >::~vector() [clone .lto_priv.5218] /usr/include/c++/7/bits/stl_vector.h:434 #12 0x55b4b9da20dd in osgText::Font::~Font() ./openmw/extern-git/OpenSceneGraph/src/osgText/Font.cpp:295 #13 0x55b4b9e59ed2 in __base_dtor ./openmw/extern-git/OpenSceneGraph/src/osgText/DefaultFont.cpp:35 #14 0x55b4b9e59ed2 in __deleting_dtor ./openmw/extern-git/OpenSceneGraph/src/osgText/DefaultFont.cpp:37 #15 0x55b4b8df93d6 in osg::Referenced::unref() const ./openmw/extern-git/OpenSceneGraph/include/osg/Referenced:201 #16 0x55b4b98e7518 in osg::ref_ptr::~ref_ptr() ./openmw/extern-git/OpenSceneGraph/include/osg/ref_ptr:41 #17 0x55b4b98e7518 in std::pair, double>::~pair() /usr/include/c++/7/bits/stl_pair.h:208 #18 0x55b4b98e7518 in __base_dtor /usr/include/c++/7/bits/stl_pair.h:208 #19 0x55b4b98e7518 in void __gnu_cxx::new_allocator, std::allocator >, osg::ref_ptr > const, std::pair, double> > > >::destroy, std::allocator >, osg::ref_ptr > const, std::pair, double> > >(std::pair, std::allocator >, osg::ref_ptr > const, std::pair, double> >*)
[osg-users] Re: OpenSceneGraph-3.6.5 release candidate 2 tagged, please test
Hi Robert, As I've mentioned in the past, I'm an OpenMW (note the single M) developer. It was actually me who reported the issues with the default font, and only a subset were resolved before you went on hiatus. I've lost the thread where we were discussing it as I'd bookmarked the forum thread, which is now dead. As I recall, though, my most recent minimal things-don't-get-relased-when-they-should example was resolvable by a couple of potential changes I suggested, but you said they'd have to wait until you could investigate more thoroughly, as it was in code no one had touched for a while, so could maybe upset other applications, or not exhaustively remove all such bugs. I pretty much stopped trying to make OpenMW play nicely with 3.6.x when you disappeared as there wasn't much point if nothing would be mergeable until you had more time, and now you do, I don't have much. Anyway, on to the matter at hand, I don't get the crash, but I'm missing some commits in the 3.6 branch, and also I still have one of my proposed fixes. I'm making things more upstream-like, and I'll add another sentence once everything's rebuilt. So after waiting an age for everything to rebuild, I got reminded that the occlusion query API got changed by Julien Valentine's recent PR. He made a PR for OpenMW that was supposed to resolve that, but obviously it didn't work until I tweaked it. Once I'd built everything, I tested it, and I'm not seeing the crash here. This is with a Debug build of OpenMW and OSG, and on Windows, and I don't think anyone else is using debug builds of both, especially not on Windows. I guess that means that I've not really given any more information beyond this not being something that happens for everyone. It might be dependent on other factors, so a more detailed description of how reliable the crash is and whether it's dependent on anything is needed before anyone can do anything on OpenMW's end. One thing of note is that the OpenMW profiler doesn't use the default font (at least on my machine). It uses a truetype one. I seem to remember seeing it use the default font in the past, and it's not impossible that this is toggled via a setting I've forgotten about, but I've had a good look and can't find one. So that's what I know. Chris On Monday, 27 January 2020 09:57:23 UTC, Robert Osfield wrote: > > Hi Fiabian, > > On Monday, 27 January 2020 09:41:43 UTC, Fabian Roth wrote: >> >> Hi, >> I am currently testing this RC with openmw. >> If i have the fps display or profiler open while exiting the application >> i get a crash on exit. >> I am not sure if this is due to a bug in my build, a bug in openmw or a >> real issue with osg. >> The issue seems to be related to the destruction of the default font, but >> i was not able to investigate further. >> Attached is a Backtrace of the issue i am currently observing. >> > > That stack trace looks like the automatic clean up of the ObjectCache with > the DefaultFont within it is related somehow to the crash. How the > DefaultFont is managed has changed, to address bugs ironically, and in a > general sense the clean up the stack trace looks just fine to me, it's > roughly what I'd expect to happen. However, I don't have any clear idea > why in this instance the crash has occurred. > > Given there isn't any obvious amiss we are unfortunately are left to widen > out the search for what is amiss. > > Does running an OSG example like osgtext fail? > > Do others in the OpenMMW community seen this same crash? > > Is it possible to run OpenMMW single threaded to see if there might be > some thread interaction? > > What OS/compile and OpenMMW version combination are you using? > > One possible cause of crash like this is memory corruption during the run > of the application that is only revealed on clean up. Using a memory tools > like valgrind might be spot this type of issue. > > Perhaps others might have seen something similar and can help shed some > light on the nature of the crash. > > If it's possible to recreate the crash with an standard OSG example, or a > small modification of one, then this would be really helpful for me to jump > in a start investigating the issue. > > Cheers. > Robert. > > > > > > > > > -- You received this message because you are subscribed to the Google Groups "OpenSceneGraph Users" group. To unsubscribe from this group and stop receiving emails from it, send an email to osg-users+unsubscr...@googlegroups.com. To view this discussion on the web visit https://groups.google.com/d/msgid/osg-users/ff578130-867c-4c93-9fb0-fd573b20ab91%40googlegroups.com.
[osg-users] Re: OpenSceneGraph-3.6.5 release candidate 2 tagged, please test
Hi Fiabian, On Monday, 27 January 2020 09:41:43 UTC, Fabian Roth wrote: > > Hi, > I am currently testing this RC with openmw. > If i have the fps display or profiler open while exiting the application i > get a crash on exit. > I am not sure if this is due to a bug in my build, a bug in openmw or a > real issue with osg. > The issue seems to be related to the destruction of the default font, but > i was not able to investigate further. > Attached is a Backtrace of the issue i am currently observing. > That stack trace looks like the automatic clean up of the ObjectCache with the DefaultFont within it is related somehow to the crash. How the DefaultFont is managed has changed, to address bugs ironically, and in a general sense the clean up the stack trace looks just fine to me, it's roughly what I'd expect to happen. However, I don't have any clear idea why in this instance the crash has occurred. Given there isn't any obvious amiss we are unfortunately are left to widen out the search for what is amiss. Does running an OSG example like osgtext fail? Do others in the OpenMMW community seen this same crash? Is it possible to run OpenMMW single threaded to see if there might be some thread interaction? What OS/compile and OpenMMW version combination are you using? One possible cause of crash like this is memory corruption during the run of the application that is only revealed on clean up. Using a memory tools like valgrind might be spot this type of issue. Perhaps others might have seen something similar and can help shed some light on the nature of the crash. If it's possible to recreate the crash with an standard OSG example, or a small modification of one, then this would be really helpful for me to jump in a start investigating the issue. Cheers. Robert. -- You received this message because you are subscribed to the Google Groups "OpenSceneGraph Users" group. To unsubscribe from this group and stop receiving emails from it, send an email to osg-users+unsubscr...@googlegroups.com. To view this discussion on the web visit https://groups.google.com/d/msgid/osg-users/d1bd2b26-bb79-4825-9082-f7d6713480f7%40googlegroups.com.
[osg-users] Re: OpenSceneGraph-3.6.5 release candidate 2 tagged, please test
Hi,
I am currently testing this RC with openmw.
If i have the fps display or profiler open while exiting the application i
get a crash on exit.
I am not sure if this is due to a bug in my build, a bug in openmw or a
real issue with osg.
The issue seems to be related to the destruction of the default font, but i
was not able to investigate further.
Attached is a Backtrace of the issue i am currently observing.
Greetings,
Fabian
*** Fatal Error ***
Invalid permissions for mapped object (signal 11)
Address: 0x5612597e50c0
* Backtrace
#2 0x7f13a65b7f20 in () at
/lib/x86_64-linux-gnu/libc.so.6
#3 0x5612597e50c0 in ()
#4 0x56125898fdc4 in
OpenThreads::ScopedPointerLock::ScopedPointerLock(OpenThreads::Mutex*)
() at ./openmw/extern-git/OpenSceneGraph/include/OpenThreads/ScopedLock:54
m = 0x5612597e5250
this =
lock = {m_lock = 0x5612597e5250}
pitr =
#5 0x56125898fdc4 in osg::StateAttribute::removeParent(osg::StateSet*)
(this=0x5612647f58e0, object=, object@entry=0x5612647d48f0)
at ./openmw/extern-git/OpenSceneGraph/src/osg/StateAttribute.cpp:38
lock = {m_lock = 0x5612597e5250}
pitr =
#6 0x561258991d2c in osg::StateSet::clear()
(this=this@entry=0x5612647d48f0) at
./openmw/extern-git/OpenSceneGraph/src/osg/StateSet.cpp:734
itr = {_M_node = 0x5612647f5a80}
#7 0x561258991f96 in __base_dtor (this=0x5612647d48f0) at
./openmw/extern-git/OpenSceneGraph/src/osg/StateSet.cpp:285
#8 0x561258992129 in __deleting_dtor (this=0x5612647d48f0) at
./openmw/extern-git/OpenSceneGraph/src/osg/StateSet.cpp:286
#9 0x561258bad988 in signalObserversAndDelete (signalDelete=true,
doDelete=true, this=0x5612647d48f0) at
/usr/include/c++/7/bits/stl_construct.h:107
newRef =
needDelete =
#10 0x561258bad988 in osg::Referenced::unref() const
(this=0x5612647d48f0) at
./openmw/extern-git/OpenSceneGraph/include/osg/Referenced:201
newRef =
needDelete =
#11 0x561258bad988 in osg::ref_ptr::~ref_ptr() () at
./openmw/extern-git/OpenSceneGraph/include/osg/ref_ptr:41
#12 0x561258bad988 in std::_Destroy
>(osg::ref_ptr*) () at
/usr/include/c++/7/bits/stl_construct.h:98
#13 0x561258bad988 in
std::_Destroy_aux::__destroy*>(osg::ref_ptr*,
osg::ref_ptr*) () at
/usr/include/c++/7/bits/stl_construct.h:108
#14 0x561258bad988 in
std::_Destroy*>(osg::ref_ptr*,
osg::ref_ptr*) () at
/usr/include/c++/7/bits/stl_construct.h:137
#15 0x561258bad988 in std::_Destroy*,
osg::ref_ptr >(osg::ref_ptr*,
osg::ref_ptr*, std::allocator
>&) () at /usr/include/c++/7/bits/stl_construct.h:206
#16 0x561258bad988 in std::vector,
std::allocator > >::~vector() [clone
.lto_priv.5148] (this=0x561264714620) at
/usr/include/c++/7/bits/stl_vector.h:434
#17 0x561258bad988 in __base_dtor (this=0x5612647145c0) at
./openmw/extern-git/OpenSceneGraph/src/osgText/Font.cpp:295
#18 0x561258bada13 in __base_dtor () at
./openmw/extern-git/OpenSceneGraph/src/osgText/DefaultFont.cpp:35
this = 0x5612647145c0
#19 0x561258bada13 in __deleting_dtor (this=0x5612647145c0) at
./openmw/extern-git/OpenSceneGraph/src/osgText/DefaultFont.cpp:37
#20 0x561258a1b6b4 in signalObserversAndDelete (signalDelete=true,
doDelete=true, this=0x5612647145c0) at
./openmw/extern-git/OpenSceneGraph/src/osg/Referenced.cpp:281
newRef =
needDelete =
this = 0x561259805c38
__p = 0x56126472ce20
this = 0x561259805c38
__p = 0x56126472ce20
#21 0x561258a1b6b4 in osg::Referenced::unref() const
(this=0x5612647145c0) at
./openmw/extern-git/OpenSceneGraph/include/osg/Referenced:201
newRef =
needDelete =
this = 0x561259805c38
__p = 0x56126472ce20
this = 0x561259805c38
__p = 0x56126472ce20
#22 0x561258a1b6b4 in osg::ref_ptr::~ref_ptr() () at
./openmw/extern-git/OpenSceneGraph/include/osg/ref_ptr:41
this = 0x561259805c38
__p = 0x56126472ce20
this = 0x561259805c38
__p = 0x56126472ce20
#23 0x561258a1b6b4 in std::pair,
double>::~pair() () at /usr/include/c++/7/bits/stl_pair.h:208
this = 0x561259805c38
__p = 0x56126472ce20
this = 0x561259805c38
__p = 0x56126472ce20
#24 0x561258a1b6b4 in __base_dtor (this=0x56126472ce40) at
/usr/include/c++/7/bits/stl_pair.h:208
this = 0x561259805c38
__p = 0x56126472ce20
this = 0x561259805c38
__p = 0x56126472ce20
#25 0x561258a1b6b4 in
__gnu_cxx::new_allocator, std::allocator >, osg::ref_ptr > const, std::pair, double> > >
>::destroy, std::allocator >, osg::ref_ptr > const, std::pair, double> >
>(std::pair, std::allocator >, osg::ref_ptr > const, std::pair, double> >*) () at
/usr/include/c++/7/ext/new_allocator.h:140
this = 0x561259805c38
__p = 0x56126472ce20
this = 0x561259805c38
__p = 0x56126472ce20
#26 0x561258a1b6b4 in
[osg-users] Re: OpenSceneGraph-3.6.5 release candidate 2 tagged, please test
Hi Robert, OSG 3.6.5-rc2 built fine on Windows with VC2019 (so VC2017 is probably OK) and appears to be running properly in our application. If there are more RCs I can try to test those and I should be able to get 3.6.5 binaries posted soon after it is released. Other than the usual mods I make to build with VC using GNU make the FBX plugin support has some issues that I worked around, wrt CMake usage and support for the newer VC and FBX versions. I can share my fixed versions if you would like and you tell me the best way to do that now. Cheers, Stuart -- You received this message because you are subscribed to the Google Groups "OpenSceneGraph Users" group. To unsubscribe from this group and stop receiving emails from it, send an email to osg-users+unsubscr...@googlegroups.com. To view this discussion on the web visit https://groups.google.com/d/msgid/osg-users/19f2d491-8ffc-45fc-b7fe-5126c48ea797%40googlegroups.com.
[osg-users] Re: OpenSceneGraph-3.6.5 release candidate 2 tagged, please test
HI All, Still waiting on feedback on how well 3.6.5-rc2 is working OK. I'm ready to tag 3.6.5 at my end as there are no Issue reported yet that I can look into resolving. If there are no Issue's raised by Monday I'll go ahead and tag 3.6.5 stable release. Cheers, Robert. -- You received this message because you are subscribed to the Google Groups "OpenSceneGraph Users" group. To unsubscribe from this group and stop receiving emails from it, send an email to osg-users+unsubscr...@googlegroups.com. To view this discussion on the web visit https://groups.google.com/d/msgid/osg-users/756a9af1-c0d5-4795-83fc-773a3d99130b%40googlegroups.com.
