Hans Breuer <[EMAIL PROTECTED]> writes:

> while adapting make and export definition files for the
> win32 build of Gimp 1.3, there where some questions ariseing :
> - what are the supposed dependency between all the libgimp* 
>   libraries ?

you got it right in your graphics below.

> what I have at the moment is (gimp.exe works, but most
> plug-ins do not yet) :
> gimp.exe                      plug-ins
>   + gimpcolor.dll            + gimp.dll
>   + gimpwidgets.dll          + gimpui.dll
>      + gimpcolor.dll         |  + gimpwidgets
>      + gimpmath.dll          |    + gimpcolor.dll
>                              |    + gimpmath.dll
>                              + gimpcolor.dll
>                              + gimpmath.dll
>                              + gimpwidgets.dll
> all the above resolves fine except libgimwidgets dependcies,
> because there are the functions listed below, which need
> (and have) different implementations in Gimps and Plug-Ins
> context. Currently I'm linking gimpwidgets.dll against
> the functions exported by Gimp, but this probably is the
> reason why most plug-ins aren't running ...
> I could hack up a windoze specific version for those functions, 
> to dynamically detect, if they are running in Gimp's or Plug-In's 
> context, but does anyone now of a proper cross platform sollution ?
> critical functions:
>  gimp_palette_get_background
>  gimp_palette_get_foreground
>  gimp_standard_help_func
>  gimp_unit_get_abbreviation
>  gimp_unit_get_digits
>  gimp_unit_get_factor
>  gimp_unit_get_number_of_built_in_units
>  gimp_unit_get_number_of_units
>  gimp_unit_get_plural
>  gimp_unit_get_singular
>  gimp_unit_get_symbol

We probably need to export the symbols you listed so you can link
gimp against libgimpwidgets. libgimp exports the same symbols but
has an implementation that works through the PDB.

> Second question:
> - to make Gimp compile again I needed to include some headers 
> in other headers again (see attached patch). Yes, I've read the 
> docs, which say not to do so, but is it really wanted to get 
> a header order dependency ?

Please send the patch included in your email, not as zip archive.
Makes it easier to read and allows me to comment on the diff in
a reply. I have looked into it and parts of it look good, others
need a different solution imho.

Salut, Sven

Reply via email to