Bug 17001 – Build error with Gtk port on Mac OS X (r29781)

• Status: RESOLVED
• Resolution: FIXED
• Component: WebKitGTK

[Randall Wood]

I am attempting to build webkit-gtk on Mac OS X 10.5 (PPC) and am having the following error:

 /usr/bin/g++-4.0 -DHAVE_CONFIG_H -I. -I./JavaScriptCore -I./JavaScriptCore/ForwardingHeaders -I./JavaScriptCore/wtf -I./JavaScriptCore/kjs -I./DerivedSources -DBUILDING_GTK__=1 -DWTF_CHANGES -DXP_UNIX -DNDEBUG -DENABLE_ICONDATABASE=0 -I./JavaScriptCore/bindings -I./WebCore -I./WebCore/ForwardingHeaders -I./WebCore/platform -I./WebCore/platform/network -I./WebCore/platform/graphics -I./WebCore/platform/text -I./WebCore/loader -I./WebCore/loader/icon -I./WebCore/css -I./WebCore/dom -I./WebCore/page -I./WebCore/bridge -I./WebCore/editing -I./WebCore/history -I./WebCore/xml -I./WebCore/html -I./WebCore/bindings/js -I./WebCore/platform/image-decoders -I./WebCore/platform/image-decoders/bmp -I./WebCore/platform/image-decoders/gif -I./WebCore/platform/image-decoders/ico -I./WebCore/platform/image-decoders/jpeg -I./WebCore/platform/image-decoders/png -I./WebCore/platform/image-decoders/xbm -I./WebCore/plugins -I./WebCore/rendering -I./WebCore/bindings/js -DWTF_USE_CURL=1 -I./WebCore/platform/network/curl -DENABLE_DATABASE=1 -I./WebCore/platform/sql -I./WebCore/storage -DENABLE_XPATH=1 -DENABLE_XSLT=1 -I./WebKit/gtk/WebView -DBUILDING_CAIRO__=1 -DBUILD_WEBKIT -I./WebCore/platform/gtk -I./WebCore/platform/graphics/gtk -I./WebCore/platform/graphics/cairo -I./WebCore/loader/gtk -I./WebCore/page/gtk -I./WebKit/gtk/WebCoreSupport -I./WebKit/gtk/WebView -I/opt/local/include -I/opt/local/include -fno-strict-aliasing -Wall -W -Wcast-align -Wchar-subscripts -Wreturn-type -Wformat -Wformat-security -Wno-format-y2k -Wundef -Wmissing-format-attribute -Wpointer-arith -Wwrite-strings -Wno-unused-parameter -fvisibility-inlines-hidden -fno-rtti -fno-exceptions -fvisibility=hidden -D_REENTRANT -DXTHREADS -I/opt/local/include/glib-2.0 -I/opt/local/lib/glib-2.0/include -I/opt/local/include -I/opt/local/include/gtk-2.0 -I/opt/local/lib/gtk-2.0/include -I/opt/local/include/atk-1.0 -I/opt/local/include/cairo -I/opt/local/include/pango-1.0 -I/opt/local/include/freetype2 -I/opt/local/include/libpng12 -I/usr/X11/include -I/opt/local/include/libxml2 -I/opt/local/include -I/opt/local/include -I/opt/local/include -I/opt/local/include/libxml2 -O2 -MT WebCore/page/libWebKitGtk_la-Frame.lo -MD -MP -MF WebCore/page/.deps/libWebKitGtk_la-Frame.Tpo -c WebCore/page/Frame.cpp  -fno-common -DPIC -o WebCore/page/.libs/libWebKitGtk_la-Frame.o
/opt/local/include/X11/X.h:108: error: conflicting declaration 'typedef XID Cursor'
/System/Library/Frameworks/ApplicationServices.framework/Frameworks/QD.framework/Headers/QuickdrawTypes.h:269: error: 'Cursor' has a previous declaration as 'typedef struct Cursor Cursor'
make[1]: *** [WebCore/page/libWebKitGtk_la-Frame.lo] Error 1
make: *** [all] Error 2

Complete logs are at http://homepage.mac.com/rhwood/webkit/r29781/install.log

It seems to me that perhaps something is trying to use QuickDraw because I am on a Mac, despite my wanting to build the gtk version?

[Mark Rowe (bdash)]

I'm curious what is pulling in the QuickDraw headers.  It may or may not be a bug that they are being pulled in when building the Gtk port on Mac OS X.  One way to find out where they are being included from is to intentionally introduce an error into the header.  For example, you could edit:

/System/Library/Frameworks/ApplicationServices.framework/Frameworks/QD.framework/Headers/QuickdrawTypes.h

and add a "#error Why am I here?" line somewhere near the beginning.  Then when you build, GCC will emit an error which will mention which file is including the one triggering the error.

[Randall Wood]

Adding the error message results in:

...[blah blah blah]...
In file included from /System/Library/Frameworks/ApplicationServices.framework/Frameworks/QD.framework/Headers/QD.h:42,
                 from /System/Library/Frameworks/ApplicationServices.framework/Headers/ApplicationServices.h:30,
                 from /System/Library/Frameworks/Carbon.framework/Headers/Carbon.h:24,
                 from ./JavaScriptCore/bindings/npapi.h:84,
                 from ./JavaScriptCore/bindings/npruntime.h:76,
                 from ./JavaScriptCore/bindings/npruntime_internal.h:28,
                 from ./JavaScriptCore/bindings/NP_jsobject.h:31,
                 from WebCore/page/Frame.cpp:70:
/System/Library/Frameworks/ApplicationServices.framework/Frameworks/QD.framework/Headers/QuickdrawTypes.h:1:2: error: #error Why am I here?
/opt/local/include/X11/X.h:108: error: conflicting declaration 'typedef XID Cursor'
/System/Library/Frameworks/ApplicationServices.framework/Frameworks/QD.framework/Headers/QuickdrawTypes.h:270: error: 'Cursor' has a previous declaration as 'typedef struct Cursor Cursor'

[Mark Rowe (bdash)]

Thanks!  So the problem would seem to be that both XP_MAC and XP_UNIX are defined on Mac OS X.  This pulls in both Quickdraw.h and Xlib.h, which have conflicting definitions for Cursor.

I'm guessing that for the Gtk port on Mac OS X, we probably don't want XP_MAC defined at all.  Whether or not XP_UNIX is defined should depend on whether you're using the X11 or "native" version of Gtk.

[Mark Rowe (bdash)]

It might be as simple as changing:

#if defined(__APPLE_CC__) && !defined(__MACOS_CLASSIC__)

to:

#if defined(__APPLE_CC__) && !defined(__MACOS_CLASSIC__) && !defined(XP_UNIX)

I'm a little wary of changing npapi.h though as it's technically an API header, but a simple change like this is unlikely to cause problems.

[Rodney Dawes]

(In reply to comment #3)
> I'm guessing that for the Gtk port on Mac OS X, we probably don't want XP_MAC
> defined at all.  Whether or not XP_UNIX is defined should depend on whether
> you're using the X11 or "native" version of Gtk.

We don't want XP_MAC or XP_UNIX defined I don't think. In fact, XP_UNIX only gets defined with the qmake build, when the GTK+ port is being built against X11. Is this problem occurring with the autotools build? Are you sure that the GTK+ you're building against, isn't using X11?

[Mark Rowe (bdash)]

I've just verified that the one-line change I suggested resolves this issue when building the Gtk port on Mac OS X.

[Mark Rowe (bdash)]

Created attachment 18690 [details]
Patch

[Darin Adler]

Comment on attachment 18690 [details]
Patch

Looks good. r=me

[Mark Rowe (bdash)]

Fix landed in r29806.