Bug 28727 – Gtk Build broken for OSX Quartz

• Component: WebKitGTK
• Status: RESOLVED WONTFIX

[John Ralls]

Building WebKit for GTK fails:
WebCore/plugins/PluginView.cpp: In member function ‘NPError WebCore::PluginView::setValue(NPPVariable, void*)’:
WebCore/plugins/PluginView.cpp:439: error: ‘m_drawingModel’ was not declared in this scope
WebCore/plugins/PluginView.cpp:463: error: ‘m_eventModel’ was not declared in this scope
WebCore/plugins/PluginView.cpp: In constructor ‘WebCore::PluginView::PluginView(WebCore::Frame*, const WebCore::IntSize&, WebCore::PluginPackage*, WebCore::Element*, const WebCore::KURL&, const WTF::Vector<WebCore::String, 0ul>&, const WTF::Vector<WebCore::String, 0ul>&, const WebCore::String&, bool)’:
WebCore/plugins/PluginView.cpp:627: error: class ‘WebCore::PluginView’ does not have any field named ‘m_drawingModel’
WebCore/plugins/PluginView.cpp:628: error: class ‘WebCore::PluginView’ does not have any field named ‘m_eventModel’
WebCore/plugins/PluginView.cpp:649: error: ‘m_npCgContext’ was not declared in this scope

(The line numbers are not correct for the current trunk; they refer to release 1.1.12. The underlying problem described below still exists.)

It does so because PluginView.h contains the lines:
#if PLATFORM(GTK) || defined(Q_WS_X11)
        void setNPWindowIfNeeded();
#elif defined(XP_MACOSX)
        NP_CGContext m_npCgContext;
        OwnPtr<Timer<PluginView> > m_nullEventTimer;
        NPDrawingModel m_drawingModel;
        NPEventModel m_eventModel;

        void setNPWindowIfNeeded();
        void nullEventTimerFired(Timer<PluginView>*);
        Point globalMousePosForPlugin() const;
#endif

 and PluginView.cpp has:

#if defined(XP_MACOSX)
    case NPPVpluginDrawingModel: {
        // Can only set drawing model inside NPP_New()
        if (this != currentPluginView())
           return NPERR_GENERIC_ERROR;

        NPDrawingModel newDrawingModel = NPDrawingModel(uintptr_t(value));
        switch (newDrawingModel) {
        case NPDrawingModelCoreGraphics:
            m_drawingModel = newDrawingModel;
            return NPERR_NO_ERROR;

...

#if defined(XP_MACOSX)
    , m_drawingModel(NPDrawingModel(-1))
    , m_eventModel(NPEventModel(-1))
#endif

When building with GTK and target=quartz, both GTK and XP_MACOSX are true. This was discussed at length in https://bugs.webkit.org/show_bug.cgi?id=17001. The initial error was committed in r43633 and r46028 added the member variables which induced the failure.

[John Ralls]

Created attachment 38583 [details]
This patch allows PluginView.cpp to compile.

[Jan Alonzo]

Created attachment 38768 [details]
Patch v1

[Jan Alonzo]

Hi John. Can you please try if the attached patch works for you? Thanks.

[Jan Alonzo]

(In reply to comment #2)
> Created an attachment (id=38768) [details]
> Patch v1

Landed as http://trac.webkit.org/changeset/47882.

[John Ralls]

No, in fact it makes the situation much worse.

Apparently you missed out on quartz being the native, non-X11, windowing system for OSX. It is a supported windowing system in Gtk. 

Regards,
John Ralls

[Jan Alonzo]

(In reply to comment #5)
> No, in fact it makes the situation much worse.
> 
> Apparently you missed out on quartz being the native, non-X11, windowing system
> for OSX. It is a supported windowing system in Gtk. 
> 
> Regards,
> John Ralls

Are you able to provide a compile log this time? On Mac, WebKitGtk only support X11 plugins.

[John Ralls]

Created attachment 38823 [details]
Error output from making webkit r47899 + config.log

Sorry this took so long, I had to build a complete unstable Gtk tree to get the current WebKit SVN to build. I figured you didn't want to see the (probably bogus) results from my applying your patch to 1.1.12.  You'll see from the error output that there are a bunch of X11-related macros which aren't defined, because I'm not building against X11.

If WebKit Gtk doesn't support the quartz backend, then why does it have a configure option --target=quartz?

Regards,
John Ralls

[Jan Alonzo]

Comment on attachment 38768 [details]
Patch v1

Clearing review flag so it won't appear in the commit queue.

[John Ralls]

Created attachment 41762 [details]
New patch against r49997 which restores ability of WebKit-GTK to build with target=quartz 

I've finally gotten back to working on this. This patch, against r50006, permits WebKit to build successfully with target=quartz and without X11. It passes all the unittests and GtkLauncher works. I'm not sure how well (or even if) plugins will work with this build: I haven't found any tests for them.

[Eric Le Lay]

(In reply to comment #9)
> Created an attachment (id=41762) [details]
> New patch against r49997 which restores ability of WebKit-GTK to build with target=quartz 
> 
> I've finally gotten back to working on this. This patch, against r50006, permits WebKit to build successfully with target=quartz and without X11. It passes all the unittests and GtkLauncher works. I'm not sure how well (or even if) plugins will work with this build: I haven't found any tests for them.

Hi John, don't you have issues compiling gtkxtbin.c (webkit 1.2.0 & 1.2.1) ? I ended up totally disabling it (see https://trac.macports.org/ticket/24622).

[John Ralls]

I haven't yet tried building 1.2.x; the last version I've built is 1.1.15.4; the patches to get that built are at http://github.com/jralls/gtk-osx-build/tree/master/patches/.

Since 1.1.15.4 works fine for Gnucash's needs, I'm not likely to look at 1.2 until late this summer.

[Martin Robinson]

Quartz support will need to be worked into WebKit2GTK.