Bug 21427 – Implement NSAPI support for WebKit/Mac inside WebCore/plugins

• Component: Plug-ins
• Status: RESOLVED FIXED

[Tor Arne Vestbø]

First steps towards NSAPI support for QtWebKit/Mac. 

PluginPackageMac and PluginViewMac should have abstractions for any Qt specific stuff, just like PluginPackage/ViewWin.

No changelog for now, I'll add that once the patches are reviewed and ready to be put in. 

Does the general approach look OK?

[Tor Arne Vestbø]

Created attachment 24143 [details]
Moved implementation of _webkit_isCaseInsensitiveEqualToString to WebCore

Basis for patch 0003

[Tor Arne Vestbø]

Created attachment 24144 [details]
Move _web_encodingForResource from WebKit into WebCore

Basis for patch 0003

[Tor Arne Vestbø]

Created attachment 24145 [details]
Implement PluginPackageMac and PluginViewMac

Sorry, was referencing this one as 0003, but Bugzilla seems to remove the patch number.

[Simon Hausmann]

On a more general note: What we're trying to achieve here is an implementation of NPAPI based on the existing PluginPackage/PluginView code that can be used across all ports that run on Mac OS X. It will take a while to become perfect, but we want to start in small steps. These patches are the basis to continue and implement more in smaller steps.

[Tor Arne Vestbø]

Created attachment 25016 [details]
Rebased against ToT

[Tor Arne Vestbø]

Created attachment 25017 [details]
Move _web_encodingForResource from WebKit into WebCore and change return type

Rebased against ToT

[Tor Arne Vestbø]

Created attachment 25019 [details]
Implement PluginPackageMac and PluginViewMac, and enable NPAPI support for QtWebKit/Mac

Rebased against ToT

[Simon Hausmann]

Comment on attachment 25016 [details]
Rebased against ToT

Looks good to me, but we should probably mention the reason behind this in the ChangeLog

[Simon Hausmann]

Comment on attachment 25017 [details]
Move _web_encodingForResource from WebKit into WebCore and change return type

Simple patch, looks good to me. We should also mention the reason here.

[Simon Hausmann]

Comment on attachment 25019 [details]
Implement PluginPackageMac and PluginViewMac, and enable NPAPI support for QtWebKit/Mac

> From 97e0aaaa77ee021c33364169ad55a8c001573073 Mon Sep 17 00:00:00 2001
> From: =?utf-8?q?Tor=20Arne=20Vestb=C3=B8?= <[e-mail]>
> Date: Wed, 17 Sep 2008 16:56:24 +0200
> Subject: [PATCH] =?utf-8?q?2008-11-10=20=20Tor=20Arne=20Vestb=C3=B8=20=20<[e-mail]>?=
> MIME-Version: 1.0
> Content-Type: text/plain; charset=utf-8
> Content-Transfer-Encoding: 8bit
> 
>         Reviewed by NOBODY (OOPS!).
> 
>         Implement PluginPackageMac and PluginViewMac, and enable NPAPI support for QtWebKit/Mac

It seems the ChangeLog is missing here.

> --- a/WebCore/plugins/PluginView.cpp
> +++ b/WebCore/plugins/PluginView.cpp
> @@ -138,8 +138,12 @@ void PluginView::frameRectsChanged() const
>  
>  void PluginView::handleEvent(Event* event)
>  {
> -    if (!m_plugin || m_isWindowed)
> +    if (!m_plugin)
>          return;
> +#if !defined(XP_MACOSX)
> +    if (m_isWindowed)
> +        return;
> +#endif

This #ifdef probably deserves a comment.

> @@ -419,6 +423,12 @@ NPError PluginView::setValue(NPPVariable variable, void* value)
>      case NPPVpluginTransparentBool:
>          m_isTransparent = value;
>          return NPERR_NO_ERROR;
> +#if defined(XP_MACOSX)
> +    case NPPVpluginDrawingModel:
> +	return NPERR_NO_ERROR;
> +    case NPPVpluginEventModel:
> +	return NPERR_NO_ERROR;
> +#endif

The indentation is off here.

>          PlatformPluginWidget platformPluginWidget() const { return m_window; }
> +        void setPlatformPluginWidget(PlatformPluginWidget widget) {
> +            if (widget != m_window)
> +                m_window = widget;
> +        }

The assignment is probably faster enough to not require a check for inequality :)


> +static int modifiersForEvent(UIEventWithKeyState *event)

Misplaced * :)


The rest looks good to me

[Simon Hausmann]

Comment on attachment 25019 [details]
Implement PluginPackageMac and PluginViewMac, and enable NPAPI support for QtWebKit/Mac

Ooops, accidentially r+'ed instead of r-. The comments need to be addressed :)

[Tor Arne Vestbø]

First two patches landed in r38332 and r38333

[Tor Arne Vestbø]

Created attachment 25095 [details]
Updated patch

Fixed comments

[Simon Hausmann]

Comment on attachment 25095 [details]
Updated patch

>  public:
>          PlatformPluginWidget platformPluginWidget() const { return m_window; }
> +        void setPlatformPluginWidget(PlatformPluginWidget widget) {
> +                m_window = widget;
> +        }

From the looks of it other code in WebCore puts simple inline setters in one line of code.

> +    if (!dispatchNPEvent(record)) {
> +        LOG(Events, "PluginView::setFocus(): Get-focus event not accepted");
> +    }
[...]
> +    if (!m_isStarted) {
> +        // TODO: Draw the "missing plugin" image
> +        return;
> +    }
[...]
> +    if (!dispatchNPEvent(event)) {
> +        LOG(Events, "PluginView::paint(): Paint event not accepted");
> +    }
[...]
> +    if (!dispatchNPEvent(record)) {
> +        LOG(Events, "PluginView::handleKeyboardEvent(): Keyboard event type %d not accepted",
> +                record.what);
> +    } else {
> +        event->setDefaultHandled();
> +    }
[...]
> +    if (!dispatchNPEvent(record)) {
> +        LOG(Events, "PluginView::nullEventTimerFired(): Null event not accepted");
> +    }

All of these one line control clauses should not use braces.

I'm really nitpicking here now :-), so r+ under the condition that you fix them before landing.

[Mark Rowe (bdash)]

Have these patches all been landed?  If so, why is this bug still open?

[Tor Arne Vestbø]

Ah, sorry Mark, yes they've been landed in r38332, r38333, and r38371. Closing.