Stefan Teleman wrote:
> Paul Floyd wrote:

>> I just wrote what I read.
> 
> Not really. You wrote *some* of what you read:
> 
> http://doc.trolltech.com/4.4/supported-platforms.html
> 
> Supported Features:
> 
> Compiler    Features
>         Concurrent    XmlPatterns    WebKit    CLucene    Phonon
> 
> [ ... ]
> SunCC <= 5.6    X                    X    X
> SunCC 5.7 & 5.8    X        X            X    X
> SunCC 5.9 and    X        X        X    X    X
> up   

I'd have no beef if the third X from the end of that last line weren't 
there.

>> Qt 4.4.0 is the first version of Qt that I've used that won't build 
>> out of the box on Solaris with the Sun compiler.
>  
> WebKit was introduced *first* in QT 4.4.0, and it is WebKit which is 
> where the vast majority of compile failures were reported. The compile 
> failures in qmake are caused by your $(CXXFLAGS) not passing 
> -D_XOPEN_SOURCE=500. This define is mandatory, and not only for QT.

Here's a rough list of the files I changed in building Qt:
  find . -name "*~"
./src/corelib/thread/thread.pri~
./src/3rdparty/webkit/JavaScriptCore/kjs/math_object.cpp~
./src/3rdparty/webkit/JavaScriptCore/kjs/date_object.cpp~
./src/3rdparty/webkit/JavaScriptCore/kjs/operations.cpp~
./src/3rdparty/webkit/JavaScriptCore/kjs/collector.cpp~
./src/3rdparty/webkit/JavaScriptCore/API/JSObjectRef.h~
./src/3rdparty/webkit/JavaScriptCore/API/JSValueRef.h~
./src/3rdparty/webkit/JavaScriptCore/API/JSBase.h~
./src/3rdparty/webkit/JavaScriptCore/API/JSContextRef.h~
./src/3rdparty/webkit/JavaScriptCore/API/JSStringRef.h~
./src/3rdparty/webkit/WebCore/dom/QualifiedName.cpp~
./src/3rdparty/webkit/WebCore/ksvg2/svg/SVGPathElement.cpp~
./src/3rdparty/webkit/WebCore/ksvg2/svg/SVGPolyElement.cpp~
./src/3rdparty/webkit/WebCore/css/CSSSelector.cpp~
./src/3rdparty/webkit/WebCore/css/CSSStyleSelector.cpp~
./src/3rdparty/webkit/WebCore/xml/XPathValue.cpp~
./src/3rdparty/webkit/WebCore/xml/XPathFunctions.cpp~
./src/3rdparty/webkit/WebCore/editing/markup.cpp~
./src/3rdparty/webkit/WebCore/html/HTMLInputElement.cpp~
./src/3rdparty/webkit/WebCore/html/HTMLParser.cpp~
./mkspecs/solaris-cc/qmake.conf~

(I probably edited a few files with vi, so no ~ traces). They're all in 
webkit apart from the thread profile include file.

I expect that _XOPEN_SOURCE will fix one problem with munmap.

> Modulo the overloading ambiguity compile failure in the example you 
> mentioned earlier. I have not had the time to look into it thoroughly, 
> and i have not yet determined that it is indeed a QT 4.4.x bug or a 
> Studio 12 bug.

There's one other resolution problem, which could be a Studio problem, 
and a few that are operator+ ambiguities between String and QString 
(easily fixed by explicitly creating String instances).

Most of the remaining errors are GCC (and perhaps other compilers) 
nonstandard code: use of "__inline" and using C99 features in C++ 
[including "<stdbool.h>", use of mathematical macros like isfinite]

The use of stdbool.h is probably the only one that might be dangerous 
(since sizeof(C++ bool) is different to sizeof(C99 _Bool)).

> Reporting compiler failures for WebKit along with suggested fixes is the 
> right thing to do. Just because you ran into compile failures in 4.4.0 
> and you reported them, it doesn't mean that they won't be fixed in 
> 4.4.1, or that "Solaris is no longer supported", or that "TT developers 
> no longer have Solaris workstations". Maybe it takes time to synchronize 
> compiler fixes for WebKit between TT and upst

I don't really know what the relationship is between Trolltech/Qt and 
KDE/WebKit. If TT doesn't own (or hasn't forked) WebKit, then I can see 
that they are going to have problems supporting it.

> Do you still compile QT4 with SunCC 5.6 ?

No, just Studio 12 and Express.

A+
Paul
-- 
Paul Floyd                 http://paulf.free.fr


Reply via email to