reproducible crash when playing an animated gif

Bug #1953743 reported by jlambrecht
6
This bug affects 1 person
Affects Status Importance Assigned to Milestone
Epiphany Browser
Confirmed
Critical
epiphany-browser (Ubuntu)
New
Undecided
Unassigned

Bug Description

This bug report following request of gnome developer(s) to submit the bug reports to Ubuntu

ProblemType: Bug
DistroRelease: Ubuntu 21.10
Package: epiphany-browser 41.0-1
ProcVersionSignature: Ubuntu 5.13.0-22.22-generic 5.13.19
Uname: Linux 5.13.0-22-generic x86_64
NonfreeKernelModules: zfs zunicode zavl icp zcommon znvpair
ApportVersion: 2.20.11-0ubuntu71
Architecture: amd64
CasperMD5CheckResult: unknown
Date: Thu Dec 9 13:30:53 2021
InstallationDate: Installed on 2021-09-16 (84 days ago)
InstallationMedia: Zorin-OS 16 Core 64bit
SourcePackage: epiphany-browser
UpgradeStatus: Upgraded to impish on 2021-12-03 (6 days ago)

Revision history for this message
In , jlambrecht (commandline.be) wrote :

Created attachment 445302
backtrace of eolie 105511

This is also tracked as issue https://gitlab.gnome.org/World/eolie/-/issues/389

The issue can be reproduced by starting a web application for twitter and playing an animated gif in this session. It was not tested if the system freeze also occurs when playing video.

os= Ubuntu 21.04
CPU/GPU = AMD Ryzen 3700U

Revision history for this message
In , jlambrecht (commandline.be) wrote :

increased severity to blocker due to system freeze

Revision history for this message
In , Michael Catanzaro (mike-catanzaro) wrote :

The crash is bug #201507. We've struggled for years to find a reproducer for this. If you're able to trigger it reliably, then you're probably going to be a hero here.

The system freeze is obviously not our fault though. That said, check to make sure WebKit is not using an excessive amount of RAM. If RAM usage looks good, then I wonder if you have uncovered a bug in mesa. If not, who knows where....

Revision history for this message
In , jlambrecht (commandline.be) wrote :

Sorry to say i've likely mixed up some things here swichting between too many topics too often.

The Twitter Web App is one run in epiphany but the backtrace here is for an Eolie crash.

I've just crashed eolie again and am attaching another backtrace.

Revision history for this message
In , Michael Catanzaro (mike-catanzaro) wrote :

*** Bug 233580 has been marked as a duplicate of this bug. ***

Revision history for this message
In , Michael Catanzaro (mike-catanzaro) wrote :

(In reply to LJoris from comment #3)
> Sorry to say i've likely mixed up some things here swichting between too
> many topics too often.
>
> The Twitter Web App is one run in epiphany but the backtrace here is for an
> Eolie crash.

It's exactly the same crash, and I was going to say "the problem is your EGL is somehow broken," but... last I checked, you were using X11 and therefore getting GLX rather than EGL, right? Could you please open up webkit://gpu, copy/paste the output into a file, and attach it here?

Revision history for this message
In , jlambrecht (commandline.be) wrote :

Created attachment 445313
eolie crash on tab loading 23839

crash was triggered by enabled use of media, images and loading both

the att.com and nu.nl URL
the nu.nl URL redirects to https://myprivacy.dpgmedia.be/ or https://myprivacy.dpgmedia.nl/ depending on what content loads after which the crash is produced, now trying to reproduce

Revision history for this message
In , Michael Catanzaro (mike-catanzaro) wrote :

Also, please confirm what WebKitGTK version you have (easiest way is to check the About dialog in Epiphany).

Revision history for this message
In , Michael Catanzaro (mike-catanzaro) wrote :

(In reply to LJoris from comment #6)
> Created attachment 445313 [details]
> eolie crash on tab loading
>
> crash was triggered by enabled use of media, images and loading both

Nope, this one is an unrelated crash in WebKitWebsiteDataStore. Weird. That's definitely a WebKit bug, but it's not *this* bug. Could you report a separate bug for it please?

Revision history for this message
In , jlambrecht (commandline.be) wrote :

Created attachment 445315
webkit-gpu output from epiphany

eolie does not permit webkit://gpu, showing output from epiphany

Revision history for this message
In , Michael Catanzaro (mike-catanzaro) wrote :

(In reply to Michael Catanzaro from comment #5)
> It's exactly the same crash, and I was going to say "the problem is your EGL
> is somehow broken," but... last I checked, you were using X11 and therefore
> getting GLX rather than EGL, right?

OK, I see you've indeed switched from X11 to Wayland. So this issue is different from https://gitlab.gnome.org/GNOME/epiphany/-/issues/1642#note_1317772.

Revision history for this message
In , jlambrecht (commandline.be) wrote :

I've been using wayland for every session and every coredump after you mentioned it was considered obsolted in terms of development efforts.

Revision history for this message
In , jlambrecht (commandline.be) wrote :
Download full text (8.0 KiB)

the syslog output from eolie, paired with a backtrace should this help

Nov 29 21:32:14 mymachine systemd-coredump[17556]: Process 17461 (python3) of user 1010 dumped core.#012#012Stack trace of thread 17461:#012#0 0x00007f8d3c22efbb __GI_raise (libc.so.6 + 0x40fbb)#012#1 0x00007f8d3c214864 __GI_abort (libc.so.6 + 0x26864)#012#2 0x00007f8d38058ed2 _dbus_abort (libdbus-1.so.3 + 0xbed2)#012#3 0x00007f8d3807c680 _dbus_warn_check_failed (libdbus-1.so.3 + 0x2f680)#012#4 0x00007f8d3806b922 dbus_message_new_method_call (libdbus-1.so.3 + 0x1e922)#012#5 0x00007f8d380b0261 socket_embed_hook (libatk-bridge-2.0.so.0 + 0x11261)#012#6 0x00007f8d213b6510 _ZN6WebKit12WebPageProxy21bindAccessibilityTreeERKN3WTF6StringE (libwebkit2gtk-4.0.so.37 + 0xb8a510)#012#7 0x00007f8d20f4b54f _ZN3IPC22callMemberFunctionImplIN6WebKit12WebPageProxyEMS2_FvRKN3WTF6StringEESt5tupleIJS4_EEJLm0EEEEvPT_T0_OT1_St16integer_sequenceImJXspT2_EEE (libwebkit2gtk-4.0.so.37 + 0x71f54f)#012#8 0x00007f8d20f40430 _ZN6WebKit12WebPageProxy17didReceiveMessageERN3IPC10ConnectionERNS1_7DecoderE (libwebkit2gtk-4.0.so.37 + 0x714430)#012#9 0x00007f8d2117281b _ZN3IPC18MessageReceiverMap15dispatchMessageERNS_10ConnectionERNS_7DecoderE (libwebkit2gtk-4.0.so.37 + 0x94681b)#012#10 0x00007f8d21271517 _ZN6WebKit15WebProcessProxy17didReceiveMessageERN3IPC10ConnectionERNS1_7DecoderE (libwebkit2gtk-4.0.so.37 + 0xa45517)#012#11 0x00007f8d2116baa5 _ZN3IPC10Connection15dispatchMessageESt10unique_ptrINS_7DecoderESt14default_deleteIS2_EE (libwebkit2gtk-4.0.so.37 + 0x93faa5)#012#12 0x00007f8d2116d9f2 _ZN3IPC10Connection24dispatchIncomingMessagesEv (libwebkit2gtk-4.0.so.37 + 0x9419f2)#012#13 0x00007f8d2ac7bdad _ZNK3WTF8FunctionIFvvEEclEv (libjavascriptcoregtk-4.0.so.18 + 0x14ecdad)#012#14 0x00007f8d2accb3cd operator() (libjavascriptcoregtk-4.0.so.18 + 0x153c3cd)#012#15 0x00007f8d2accbd03 operator() (libjavascriptcoregtk-4.0.so.18 + 0x153cd03)#012#16 0x00007f8d3b5017ef g_main_context_dispatch (libglib-2.0.so.0 + 0x557ef)#012#17 0x00007f8d3b554d28 g_main_context_iterate.constprop.0 (libglib-2.0.so.0 + 0xa8d28)#012#18 0x00007f8d3b4ff023 g_main_context_iteration (libglib-2.0.so.0 + 0x53023)#012#19 0x00007f8d3b2af27d g_application_run (libgio-2.0.so.0 + 0xdc27d)#012#20 0x00007f8d3c5b9249 ffi_call_unix64 (libffi.so.8 + 0x7249)#012#21 0x00007f8d3c5b8639 ffi_call_int (libffi.so.8 + 0x6639)#012#22 0x00007f8d3b6293e2 n/a (_gi.cpython-39-x86_64-linux-gnu.so + 0x273e2)#012#23 0x00007f8d3b62ce16 n/a (_gi.cpython-39-x86_64-linux-gnu.so + 0x2ae16)#012#24 0x00000000005407b1 PyObject_Call (python3.9 + 0x1407b1)#012#25 0x00000000005181b6 _PyEval_EvalFrameDefault (python3.9 + 0x1181b6)#012#26 0x0000000000514a75 n/a (python3.9 + 0x114a75)#012#27 0x000000000052d302 _PyFunction_Vectorcall (python3.9 + 0x12d302)#012#28 0x0000000000516543 _PyEval_EvalFrameDefault (python3.9 + 0x116543)#012#29 0x0000000000514a75 n/a (python3.9 + 0x114a75)#012#30 0x000000000051480b _PyEval_EvalCodeWithName (python3.9 + 0x11480b)#012#31 0x00000000005fb257 PyEval_EvalCode (python3.9 + 0x1fb257)#012#32 0x00000000006205fb n/a (python3.9 + 0x2205fb)#012#33 0x000000000061b724 n/a (python3.9 + 0x21b724)#012#34 0x000000000061fb2d n/a (python3.9 + 0x21fb2d...

Read more...

Revision history for this message
In , jlambrecht (commandline.be) wrote :

Created attachment 445331
eolie backtrace 17461

Revision history for this message
In , Michael Catanzaro (mike-catanzaro) wrote :

(In reply to LJoris from comment #12)
> the syslog output from eolie, paired with a backtrace should this help

That's hard to read, but the backtrace is very useful. That is a *third* crash, the at-spi2-atk issue I mentioned earlier. To keep the bug reports readable, I really need to insist on a separate bug report for each crash. Currently we have:

 * This bug for the original Nicosia::GCGLLayer::makeContextCurrent issue
 * Unrelated WebKitWebsiteDataStore crash (comment #6), needs a separate WebKit bug report
 * Unrelated accessibility crash (comment #13), needs a separate WebKit bug report *and* and at-spi2-atk bug report. Definitely an at-spi2-atk bug here, *probably* also a WebKit bug. (Not sure.)
 * A fourth unrelated crash after receiving an X server error (https://gitlab.gnome.org/GNOME/epiphany/-/issues/1642#note_1317772, probably a mesa bug, no backtrace for this one yet)

It's a lot of bugs. This might be a record: you must be very unlucky!

Anyway, at least now with good backtraces you have enough to prove there are real bugs, so we are making some progress.

Revision history for this message
In , jlambrecht (commandline.be) wrote :

I'll see what i can do.

Note the i'm also learning how to distinguish the different bugs. For now i only distinguish by looking at EXE.

I'm not affected by these bugs, only whatever dares to work with Webkit is.

Revision history for this message
In , jlambrecht (commandline.be) wrote :

Since my last post i've had repeat system freezes whenever an animated gif plays in the epiphany browser.

There is no coredump generated and i don't know how i could generate any thing useful when reproducing this event. Since it is reproducible by playing an animated GIF i'm willing to spend more time on this but i need to know what to do to collect meaningful data.

Revision history for this message
In , jlambrecht (commandline.be) wrote :

checksec --proc=epiphany

* System-wide ASLR (kernel.randomize_va_space): Full (Setting: 2)

  Description - Make the addresses of mmap base, heap, stack and VDSO page randomized.
  This, among other things, implies that shared libraries will be loaded to random
  addresses. Also for PIE-linked binaries, the location of code start is randomized.

  See the kernel file 'Documentation/sysctl/kernel.txt' for more details.

* Does the CPU support NX: Yes

         COMMAND PID RELRO STACK CANARY SECCOMP NX/PaX PIE FORTIFY
        epiphany 119202 Full RELRO Canary found No Seccomp NX enabled PIE enabled No

Revision history for this message
In , Michael Catanzaro (mike-catanzaro) wrote :

So your distro builds it with hardening flags. Great. Totally irrelevant to this issue.

Revision history for this message
In , jlambrecht (commandline.be) wrote :

The reason i shared the checksec output was to show what is the context for this and the other issue reported should that matter with interpreting backtrace information.

In the meantime I've ran an exhaustive memory integrity check. No test failed, no warnings of any kind were reported.

Revision history for this message
jlambrecht (commandline.be) wrote :
Revision history for this message
jlambrecht (commandline.be) wrote :

note i'm not habituated to reporting bugs and the webkit url paired with this bug report shows this, there are a number of bug reports submitted

Changed in epiphany-browser:
importance: Unknown → Critical
status: Unknown → Confirmed
Revision history for this message
In , jlambrecht (commandline.be) wrote :

At this point i notice 'some' improvement in that the system freeze is less accute. Web/Epiphany freezes but now the numlock remains responsive which was not the case before. I'm not certain what to attribute this improvement to.

the animated gifs now appear to play well but when video plays (with sound?) the system still freezes

To post a comment you must log in.
This report contains Public information  
Everyone can see this information.

Other bug subscribers

Remote bug watches

Bug watches keep track of this bug in other bug trackers.