GNOME Bugzilla – Bug 792799
Site crashes after swipe gesture
Last modified: 2018-01-24 01:26:53 UTC
Leaving epiphany with a gesture (swipe from left to right to open overview, swipe from bottom to open osk) and re-entering causes a crash of the current page. As epiphany doesn't crash (only the current page) i do not have a backtrace yet. Always reproducible.
You do need to get a backtrace for us to have a chance of solving the bug. The easiest way of getting a trace for the web process is to run 'coredumpctl gdb'. You might need to look up instructions for enabling coredumpctl in your distribution if it doesn't have it running out of the box. Alternatively, you can manually connect to the web process using 'gdb -p'; finding the pid of the web process should be easy as long as you only have one tab open. Most likely, once we get a stack trace, I'll wind up directing you to WebKit Bugzilla, or possibly reassign the bug to GNOME Bugzilla, because Epiphany doesn't have any gesture-related code. But getting the trace is the first step here. Now, for the canned reply: Thanks for taking the time to report this. Without a stack trace from the crash it's very hard to determine what caused it. Can you get us a stack trace? Please see https://wiki.gnome.org/Community/GettingInTouch/Bugzilla/GettingTraces for more information on how to do so. When pasting a stack trace in this bug report, please reset the status of this bug report from NEEDINFO to its previous status. Thanks in advance!
(In reply to Michael Catanzaro from comment #1) > Most likely, once we get a stack trace, I'll wind up directing you to WebKit > Bugzilla, or possibly reassign the bug to GNOME Bugzilla I meant "reassign to GTK+" if it looks like the problem lies in GTK+ gesture management. A problem in WebKit is probably more likely, though....
Sorry i should have been a little bit clearer: Webkit process receives a SIGKILL so there is no backtrace available. coredumpctl gdb just delivers the last known SIGSEGV in _create_touch_event ()(which have been fixed in mutter and is already installed on my tablet). This issue is not related to the bug fix as it is nagging me for month.
Just noted this information within the shell before SIGKILL: Received an invalid message "WebPageProxy.DidReceiveEvent" from the web process.
WebkitGTK kills the process after receiving an invalid message instead of just skipping it... What's the recommended way to analyze this issue? Which message is invalid?
Additional information: Only occures using wayland not with x
Wow, SIGKILL! This is going to be a tougher one.... (In reply to Jan-Michael Brummer from comment #5) > WebkitGTK kills the process after receiving an invalid message instead of > just skipping it... Yeah, the web process should never be sending invalid messages. It's actually been quite a while since I've seen a problem like this. > What's the recommended way to analyze this issue? Which message is invalid? WebPageProxy.DidReceiveEvent... it told you that. :D This is actually going to be quite hard to debug unless we can reproduce, because it's not clear to me why the message is invalid. And unfortunately I don't have a touchscreen, and I'm not sure if any of the other WebKit developers do either. Maybe it will be possible to use GTK+'s simulated touch mode. Anyway, the problem is obviously in WebKit... so let's move this bug there. Canned reply will follow.
Thanks for taking the time to report this. This issue is most likely a bug in WebKit. Please report the bug to https://bugs.webkit.org/ including a link to this bug report and noting the version of WebKitGTK+ that you have installed. When reporting the WebKit bug, be sure to include the prefix '[GTK]' in the bug summary and select the 'WebKit Gtk' component.
I think it might be a gtk bug within gdk wayland and not webkit...
Probably not, because it seems quite unlikely that a GTK+ bug could cause WebKit to send an invalid IPC message. I know it's suspicious that the problem only occurs under Wayland, but WebKit has a bunch of Wayland-specific code, too. > What's the recommended way to analyze this issue? It will be to manually add printfs to the code. :/ I'm not actually sure where, though, because the IPC code is not simple. First step is to move to the right Bugzilla.