[ starting a new thread intentionally ]
MacArthur, Ian (SELEX GALILEO, UK) wrote in thread
"[fltk.development] Fltk-1.3 and Fl_Preferences":
[ Albrecht Schlosser wrote: ]
>> I took a look at your patch and stumbled over this:
>>
>> -#ifdef WIN32
>> +#ifdef _WIN32
>>
>> Why did you change this? My understanding is that FLTK defines
>> "WIN32" (without underscore) for all FLTK Windows builds, and
>> this is used everywhere in the code. I assume that this has been
>> done to be independent of maybe different macros of different
>> compilers.
>
> Yes - I probably should not have changed that. It should not matter
> but...
It does matter if _WIN32 is not defined, however. And maybe there
are other compilers that don't define it (but see below).
> Anyway, I changed it because:
>
> - we define WIN32 in our builds
That I don't understand.
> - the gcc win32 tools define WIN32 anyway, and also _WIN32 and
> quite a few other variants... On this mingw setup, all of these
> are set by default for every build:
> WIN32 _WIN32 __WIN32 __WIN32__
okay, I can see this with cygwin/gcc -mno-cygwin, too.
> - the MS tools do not define WIN32, but do define _WIN32 and a few
> others, I forget what.
That's why _we_ define WIN32, IMHO.
And what about Borland, Watcom, and ... (who knows)?
Cygwin (in "cygwin mode") defines none of them, and we rely on
WIN32 being defined.
> So, for maximum portability, we should not be depending on WIN32 being
> defined, but it seems that we can depend on _WIN32 being defined.
> Possibly.
Again: my assumption is that we (FLTK) define WIN32 to have to test
only one macro. But maybe that's for other historical reasons.
Otherwise we would end up with something like this (found in
png/pngconf.h:
#if defined(PNG_DLL) || defined(_DLL) || defined(__DLL__ ) || \
(( defined(_Windows) || defined(_WINDOWS) || \
defined(WIN32) || defined(_WIN32) || defined(__WIN32__) ))
> There have been a few posts from MS VS users who have got themselves
> into trouble recompiling files from the tarball and NOT setting WIN32
> in their builds and wondering why things go bad...
Isn't that their problem ;-) ? Using FLTK code in a non-FLTK
environment?
> A number of the files in the test folder use WIN32, for example, and
> probably ought to use _WIN32... That sort of thing...
>
>> For an example see fltk-config --cxxflags.
>>
>> Statistics: If I counted right (with grep and wc -l) then we have
>> 181 occurrences of "#if ... WIN32" and only 9 of "#if ... _WIN32"
>> (I didn't search in fluid and test dirs, however).
>>
>> One of them is in FL/Fl_Table.H and 8 are in src/fl_utf.c, which
>> are both new files from "external" sources. IMHO we should
>> change _these_ to have identical "#if defined ..." statements,
>> but not change existing ones to _WIN32.
>>
>> Does anybody know more about this?
>
> I'd guess that Greg or OksiD added those files, and probably developed
> them in a MS VS environment, hence the use of _WIN32.
Yes, that's my assumption, too, and that's why I wrote 'external'.
>>From my comments above, I imagine it is clear that I think we should
> actually change everything to _WIN32.
> But I don't feel strongly enough about it to actually make the
> changes...
That's why I asked if anybody knows more about it.
My main point is that we are consistently using WIN32 or _WIN32
everywhere. If we decide to use _WIN32 everywhere, then we must
also change the define in fltk-config and Makefiles (or makeinclude)
or elsewhere. But please don't *mix* both, otherwise we'll maybe get
into trouble.
Perhaps Matt and Mike could tell us more ?
Albrecht
_______________________________________________
fltk-dev mailing list
fltk-dev@easysw.com
http://lists.easysw.com/mailman/listinfo/fltk-dev
