[WPE] GLContextEGLWPE.cpp:44:96: error: invalid cast from type ‘GLNativeWindowType {aka long long unsigned int}’ to type ‘EGLNativeWindowType {aka unsigned int}

[Michael Catanzaro]

When building for 32-bit ARM:

./platform/graphics/egl/GLContextEGLWPE.cpp: In static member function ‘static void* WebCore::GLContextEGL::createWindowSurfaceWPE(EGLDisplay, EGLConfig, GLNativeWindowType)’:
./platform/graphics/egl/GLContextEGLWPE.cpp:44:96: error: invalid cast from type ‘GLNativeWindowType {aka long long unsigned int}’ to type ‘EGLNativeWindowType {aka unsigned int}’
     return eglCreateWindowSurface(display, config, reinterpret_cast<EGLNativeWindowType>(window), nullptr);

So, that cast is clearly illegal. GLNativeWindowType is larger than EGLNativeWindowType, so it's not safe to cast in that direction.

I notice that GLNativeWindowType is defined oddly in GLContext.h:

#if USE(EGL) && !PLATFORM(GTK) && !PLATFORM(WPE)
#include "eglplatform.h"
typedef EGLNativeWindowType GLNativeWindowType;
#else
typedef uint64_t GLNativeWindowType;
#endif

I'm genuinely uncertain where the declaration of EGLNativeWindowType is coming from, but I see this in ANGLE's eglplatform.h:

#elif defined(USE_OZONE) || defined(USE_WPE)

typedef intptr_t EGLNativeDisplayType;
typedef intptr_t EGLNativeWindowType;
typedef intptr_t EGLNativePixmapType;

And of course, intptr_t is a 32-bit type. I have no clue how this is supposed to work.

[Carlos Alberto Lopez Perez]

(In reply to Michael Catanzaro from comment #0)
> I'm genuinely uncertain where the declaration of EGLNativeWindowType is
> coming from, but I see this in ANGLE's eglplatform.h:
> 

EGLNativeWindowType is declared in /usr/include/EGL/eglplatform.h and has different definitions depending on the EGL platform you have defined (GBM, WaylandEGL, X11, etc)


I think ANGLE headers should not be used for building this file.

[Michael Catanzaro]

I don't know if ANGLE headers are actually used or not. I don't know where the definition of EGLNativeWindowType is coming from. But I'd be surprised that we have it defined to be intptr_t there if we're not using it.

[Carlos Alberto Lopez Perez]

I have been bitten by this, so I have checked it.

I think this started to happen with r217208 <https://trac.webkit.org/r217208>

Patch incoming ..

[Carlos Alberto Lopez Perez]

Created attachment 327039 [details]
Patch

[Build Bot]

Attachment 327039 [details] did not pass style-queue:


ERROR: Source/WebCore/ChangeLog:3:  Please consider whether the use of security-sensitive phrasing could help someone exploit WebKit: invalid cast  [changelog/unwantedsecurityterms] [3]
Total errors found: 1 in 2 files


If any of these errors are false positives, please file a bug against check-webkit-style.

[Michael Catanzaro]

Thanks. I'm not qualified to review this; please ask Miguel and Zan.

[Carlos Alberto Lopez Perez]

(In reply to Carlos Alberto Lopez Perez from comment #3)
> 
> I think this started to happen with r217208 <https://trac.webkit.org/r217208>
> 

When I wrote that I didn't noticed the age of the commit (6 months ago).. 
And IIRC I did ARMv7 builds of WPE trunk 1-2 months ago.

So I guess something else than r217208 caused this

In any case I still think the attached patch is ok, so any review of it is appreciated.

[WebKit Commit Bot]

Comment on attachment 327039 [details]
Patch

Clearing flags on attachment: 327039

Committed r225051: <https://trac.webkit.org/changeset/225051>

[WebKit Commit Bot]

All reviewed patches have been landed.  Closing bug.