At 23:49 23.02.01 +0800, James Henstridge wrote:
>On Fri, 23 Feb 2001, Hans Breuer wrote:
>> Thanks for breaking the win32 build. :-(
>> ...
>
>Kjartan was doing a clean up, as some files were not seeing the ENABLE_NLS
>define, so even though translatable strings were being marked, they
>weren't being translated.
>
>The extent of the change is just including config.h at the top of each
>file (before any other headers), and removing the includes from the
>headers themselves.  So the config.h defines should be visible in
>the headers.  Having the config.h include in the source files rather than
>the headers is the recommended practice.
>
>How exactly did this break the win32 build? (I am curious).
>
the M$ <math.h> and friends are defining less prototypes than their
gcc counterparts. So I simply added some missing bits to config.h on
win32 (ok - it was a hack). geometry.h uses finite() for example.

My compiler switches are stopping the build if there is any missing
prototype - et voila ...

>> 
>> What's the right way to go, if config.h definitions are required e.g. 
>> for inline functions in header files ?
>
Ok, so there is only a special case for config.h (not for almost any header
file inclusion in header files, like The Gimp's current policy).
I only wanted to make sure to not fix the build with an in-appropriate
sollution ...

Thanks,
        Hans

-------- Hans "at" Breuer "dot" Org -----------
Tell me what you need, and I'll tell you how to 
get along without it.                -- Dilbert

Reply via email to