[Webkit-unassigned] [Bug 37369] [GTK] Enable building whatever already exists of WebKit2

bugzilla-daemon at webkit.org bugzilla-daemon at webkit.org
Fri Nov 26 05:37:26 PST 2010


https://bugs.webkit.org/show_bug.cgi?id=37369





--- Comment #59 from Ravi Phaneendra Kasibhatla <ravi.kasibhatla at motorola.com>  2010-11-26 05:37:24 PST ---
(In reply to comment #57)
> (From update of attachment 74419 [details])
> View in context: https://bugs.webkit.org/attachment.cgi?id=74419&action=review
> 
> Very nice patch. Thanks for updating it. Just a couple more lingering issues. I think it's very close.
> 
> > WebCore/ChangeLog:8
> > +        No new tests. (OOPS!)
> 
> This line should be removed.
Done.
> 
> > WebCore/platform/network/soup/cache/webkit/soup-cache.h:27
> > +#define WEBKIT_API __attribute__((visibility("default")))
> > +#define WEBKIT_OBSOLETE_API WEBKIT_API __attribute__((deprecated))
> 
> This will actually break the GTK+ Windows build. You should probably include the MSVC support here too.
Done. Defined in same manner as defined in <webkit/webkitdefines.h>
> 
> > WebKit2/GNUmakefile.am:580
> > +# Ensure that the prefix header is built before
> 
> You are no longer using using a precompiled header, so I think this comment is out of date. :)
Done.
> 
> > WebKit2/WebKit2Prefix.h:33
> > +#if !defined (BUILDING_GTK__)
> >  #include <wtf/Platform.h>
> >  #include <wtf/DisallowCType.h>
> >  #ifdef __cplusplus
> >  #include <wtf/FastMalloc.h>
> >  #endif
> > +#endif /* !defined (BUILDING_GTK__) */
> 
> Why do you need to conditionally include these?
Changed to include <WebCore/config.h> at top of file for GTK port as per our discussion.
> 
> > WebKit2/WebProcess/WebCoreSupport/WebFrameLoaderClient.cpp:-33
> > -#include "PluginView.h"
> 
> I think the right way to solve this problem is to only include the big list of -I directories in WebCore if we are building WebKit 1. Otherwise, we'll probably keep running into this issue. Do you mind trying that?
Reverted this file. Solved the issue by building WebKit2 as a convenience target and then linking with actual webkit target as per our discussion yesterday.
> 
> > configure.ac:903
> > +HTTP_STACK="soup"
> > +WEBKITGTK_PC_NAME=${WEBKITGTK_PC_NAME}2
> > +AC_SUBST([HTTP_STACK])
> > +AC_SUBST([WEBKITGTK_PC_NAME])
> 
> If HTTP_STACK is hardcoded, why not just remove it? Generally we have standardized on soup. If you want to re-add support for cURL, maybe we can address it in another patch? I think it's much more feasible for WebKit2, if you really need it for something.
Done.
> 
> > configure.ac:989
> > +AC_CONFIG_FILES([
> 
> If posible. please indent the inside of this conditional.
> 
> > configure.ac:1008
> > +AC_CONFIG_FILES([
> > +WebKit2/gtk/${WEBKITGTK_PC_NAME}-${WEBKITGTK_API_VERSION}.pc:WebKit2/gtk/webkit2.pc.in
> > +]
> > +,[WEBKITGTK_API_VERSION=$WEBKITGTK_API_VERSION,WEBKITGTK_PC_NAME=$WEBKITGTK_PC_NAME]
> > +)
> 
> Same here.
Done.

-- 
Configure bugmail: https://bugs.webkit.org/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are the assignee for the bug.



More information about the webkit-unassigned mailing list