Summary: | Assertion failure in NetscapePlugin::platformDestroy when running plugins/mouse-events.html in WebKit2 on Windows | ||||||
---|---|---|---|---|---|---|---|
Product: | WebKit | Reporter: | Adam Roben (:aroben) <aroben> | ||||
Component: | Plug-ins | Assignee: | Nobody <webkit-unassigned> | ||||
Status: | RESOLVED FIXED | ||||||
Severity: | Normal | CC: | abarth, andersca, eric, jhoneycutt, webkit.review.bot | ||||
Priority: | P2 | Keywords: | InRadar, LayoutTestFailure, PlatformOnly | ||||
Version: | 528+ (Nightly build) | ||||||
Hardware: | PC | ||||||
OS: | Windows XP | ||||||
Attachments: |
|
Description
Adam Roben (:aroben)
2010-09-27 16:46:56 PDT
The plugin starts out windowed (as all plugins do on Windows), but then sets itself to be windowless after the test starts running. We could fix the assertion by making platformDestroy destroy m_window if it exists regardless of the value of m_isWindowed. But maybe we should also be destroying the window when the plugin tells us it's becoming windowless? (In reply to comment #2) > But maybe we should also be destroying the window when the plugin tells us it's becoming windowless? It looks like Firefox and Chrome don't do this, so we shouldn't either. The windowless mode is being changed via plg.windowedPlugin = true. This was added in r66408 [1]. It seems like changing the windowed mode this was was a mistake, as on Windows you aren't allowed to change the windowed mode after NPP_New is called. A better solution would have been to add a windowless or wmode attribute to the plugin element. I've filed bug 48448 to cover this issue. 1. http://trac.webkit.org/changeset/66408 Created attachment 72064 [details]
Patch
Committed r70685: <http://trac.webkit.org/changeset/70685> http://trac.webkit.org/changeset/70685 might have broken GTK Linux 64-bit Debug |