Bug 200534 – [GTK] Several tests crashing after r247898 "Reorganize UIScriptController into platform-specific subclasses"

• Component: WebKitGTK
• Status: RESOLVED FIXED

[Miguel Gomez]

compositing/backing/backing-store-attachment-scroll.html [ Crash ]
compositing/iframes/remove-reinsert-webview-with-iframe.html [ Crash ]
compositing/tiling/tile-coverage-on-scroll.html [ Crash ]
compositing/tiling/tiled-mask-inwindow.html [ Crash ]
compositing/tiling/tiled-reflection-inwindow.html [ Crash ]
editing/input/delete-text-in-composition.html [ Crash ]
editing/secure-input/password-input-changed-type.html [ Crash ]
editing/secure-input/password-input-focusing.html [ Crash ]
editing/secure-input/removed-password-input.html [ Crash ]
editing/secure-input/reset-state-on-navigation.html [ Crash ]
fast/events/autoscroll-when-input-is-offscreen.html [ Crash ]
fast/events/autoscroll-with-software-keyboard.html [ Crash ]
fast/media/mq-prefers-reduced-motion-matchMedia.html [ Crash ]
fast/visual-viewport/client-coordinates-relative-to-layout-viewport.html [ Crash ]
fast/visual-viewport/client-rects-relative-to-layout-viewport.html [ Crash ]
http/tests/inspector/network/ping-type.html [ Crash ]
http/tests/navigation/ping-attribute/anchor-cookie.html [ Crash ]
http/tests/navigation/ping-attribute/anchor-cross-origin-from-https.html [ Crash ]
http/tests/navigation/ping-attribute/anchor-cross-origin.html [ Crash ]
http/tests/navigation/ping-attribute/anchor-ping-and-follow-redirect-when-sending-ping.html [ Crash ]
http/tests/navigation/ping-attribute/area-cookie.html [ Crash ]
http/tests/navigation/ping-attribute/area-cross-origin-from-https.html [ Crash ]
http/tests/navigation/ping-attribute/area-cross-origin.html [ Crash ]
http/tests/navigation/ping-attribute/area-same-origin.html [ Crash ]
pageoverlay/overlay-remove-reinsert-view.html [ Crash ]

All of these tests are crashing in calls to ScriptController, with a backtrace similar to this (see all of the backtraces at https://build.webkit.org/results/GTK%20Linux%2064-bit%20Release%20(Tests)/r248403%20(11158)/results.html)

#0  0x00007f7f8f135dde in WTFCrash () from /home/slave/webkitgtk/gtk-linux-64-release/build/WebKitBuild/Release/lib/libjavascriptcoregtk-4.0.so.18
#1  0x000055641999bab9 in _ZN3WTR18UIScriptController11doAsyncTaskEPK13OpaqueJSValue ()
#2  0x000055641999ebd5 in _ZN3WTR20JSUIScriptController25doAfterPresentationUpdateEPK15OpaqueJSContextP13OpaqueJSValueS5_mPKPKS4_PS7_ ()
#3  0x00007f7f8e4d7c20 in _ZN3JSC19APICallbackFunction4callINS_18JSCallbackFunctionEEElPNS_9ExecStateE () from /home/slave/webkitgtk/gtk-linux-64-release/build/WebKitBuild/Release/lib/libjavascriptcoregtk-4.0.so.18
#4  0x00007f7ee7fff027 in ?? ()
#5  0x00007ffffc19acd0 in ?? ()
#6  0x00007f7f8ec6f761 in llint_op_call () from /home/slave/webkitgtk/gtk-linux-64-release/build/WebKitBuild/Release/lib/libjavascriptcoregtk-4.0.so.18
#7  0x0000000000000000 in ?? ()

[Miguel Gomez]

Reading the description of the changelog, the crash may mean that we are missing some functionality that is not implemented in the subclasses, but JSC is not my field.

[Tim Horton]

(In reply to Miguel Gomez from comment #1)
> Reading the description of the changelog, the crash may mean that we are
> missing some functionality that is not implemented in the subclasses, but
> JSC is not my field.

Yes, 100%. Probably there are a few low effort things to implement that get you 90% and the rest you should just skip. If you have backtraces for all the crashes I can help figure which are which.

[Tim Horton]

“doAsyncTask” is one low effort one for example, and that’s your example BT. See the Mac implementation for example.

[Miguel Gomez]

We have this one crashing on the WPE bot as well

fast/events/focus-anchor-with-tabindex-hang.html

[Miguel Gomez]

Another test failing because of the missing doAsyncTask method:
editing/pasteboard/paste-and-sanitize.html [ Crash ]

[Miguel Gomez]

Two more

fast/forms/contenteditable-font-optical-size.html [ Crash ]
legacy-animation-engine/compositing/transitions/add-remove-transition.html [ Crash ]

[Miguel Gomez]

There are several bugs that handle tests crashing because we lack an implementation for some UIScriptController. This one mentions doAsyncTask, but there are also copyText (bug 204112) and setContinuousSpellCheckingEnabled (bug 205109).

As there are even more methods failing, I'm going to move all the bugs to this one, to keep track of the whole issue:

From bug 204112 we lack copyText, and these tests crash:
editing/async-clipboard/clipboard-change-data-while-getting-type.html
editing/async-clipboard/clipboard-change-data-while-reading.html
editing/async-clipboard/clipboard-get-type-with-old-items.html

from bug 205109 we lack setContinuousSpellCheckingEnabled, and this test crashes:
editing/spelling/toggle-spellchecking.html

[Miguel Gomez]

This one fails because we lack activateDataListSuggestion
fast/forms/datalist/datalist-click-crash.html [ Crash ]

This one because we lack simulateAccessibilitySettingsChangeNotification
fast/media/mq-inverted-colors.html [ Crash ]

This one crashes because we lack removeViewFromWindow
media/no-fullscreen-when-hidden.html [ Crash ]

[Carlos Garcia Campos]

Making this GTK only for now, because WPE doesn't even have a UIScriptController impl yet. I'll submit a new bug to add UIScriptController for WPE.

[Carlos Garcia Campos]

Created attachment 385492 [details]
Patch

[Alejandro G. Castro]

Comment on attachment 385492 [details]
Patch

Great stuff! LGTM. Let's add the same to WPE later.

[Carlos Garcia Campos]

Committed r253467: <https://trac.webkit.org/changeset/253467>