Bug 33822 – REGRESSION(r53273): TransformationMatrix.h needs to use alternative typedef for XFORM when compiling with MinGW

• Component: Tools / Tests
• Status: RESOLVED FIXED

[Daniel Bates]

Changeset 53273 <http://trac.webkit.org/changeset/53273> broke the Qt Windows build when using the MinGW compiler.

In particular, we cannot typedef/forward-declare XFORM in TransformationMatrix.h because this causes a redefinition error with the struct of the same name defined in C:\Qt\2009.05\mingw\include\wingdi.h. Also, the MinGW wingdi.h does not define the macro definition _XFORM_ (like in the Microsoft SDK), so we cannot determine whether this struct is already defined using the preprocessor.

For completeness, the following is a snippet from the build output including the error message:

.\..\JavaScriptCore\pcre" [...] -o obj\debug\SVGElementFactory.o generated\SVG
ElementFactory.cpp
In file included from ..\..\..\WebCore\svg/SVGTransform.h:25,
                 from ..\..\..\WebCore\svg/SVGTransformList.h:26,
                 from ..\..\..\WebCore\svg/SVGTransformable.h:27,
                 from ..\..\..\WebCore\svg/SVGStyledTransformableElement.h:27,
                 from ..\..\..\WebCore\svg/SVGAElement.h:28,
                 from generated\SVGElementFactory.cpp:34:
..\..\..\WebCore\platform\graphics\transforms/TransformationMatrix.h:47: error:
conflicting declaration 'typedef struct tagXFORM XFORM'
c:\qt\2009.05\mingw\bin\../lib/gcc/mingw32/4.4.0/../../../../include/wingdi.h:16
34: error: 'XFORM' has a previous declaration as 'typedef struct _XFORM XFORM'
make[2]: *** [obj/debug/SVGElementFactory.o] Error 1

[Daniel Bates]

Created attachment 46877 [details]
Patch

No functionality was changed. So, no new tests.

[Mark Rowe (bdash)]

Why on earth would MinGW declare this type in a manner that conflicts with how the Windows API declares it?

[Adam Roben (:aroben)]

Comment on attachment 46877 [details]
Patch

We try to avoid including headers from other headers whenever possible, in an effort to reduce code size and compile times. Can we just insert the correct forward-declaration for MinGW as appropriate?

[Daniel Bates]

Created attachment 46929 [details]
Patch

Modified the patch based on a conversation with Adam Roben on IRC today (01/19).

[Adam Roben (:aroben)]

Comment on attachment 46929 [details]
Patch

r=me

[Daniel Bates]

Comment on attachment 46929 [details]
Patch

Clearing flags on attachment: 46929

Committed r53485: <http://trac.webkit.org/changeset/53485>

[Daniel Bates]

All reviewed patches have been landed.  Closing bug.