[RFC] cleanup cairo detection

[Riccardo Magliocchetti]

Hi Michael,

Il 14/05/2012 12:25, Michael Stahl ha scritto:
> On 11/05/12 18:43, Riccardo Magliocchetti wrote:

>> +dnl Check and warn is cairo has been enabled by mistake
>> +CAIRO_BY_MISTAKE=no
>> +if test $_os = Darwin -o $_os = WINNT; then
>> +    CAIRO_BY_MISTAKE=yes
>> +    enable_cairo_canvas=no
>> +    if test $_os = WINNT; then
>> +        dnl We only need cairo for Windows if we
>> +        dnl build librsvg or directx disabled
>> +        if test "$ENABLE_LIBRSVG" != NO -o -z "$ENABLE_DIRECTX"; then
>> +            enable_cairo_canvas=yes
>> +            CAIRO_BY_MISTAKE=no
>
> it seems to me that this ^^^ ...
>
>> +        fi
>> +    fi
>> +elif test -z "$enable_cairo_canvas"; then
>> +    enable_cairo_canvas=yes
>> +fi
>>
>> -SYSTEM_CAIRO=""
>> +if test "$ENABLE_LIBRSVG" != NO; then
>> +    enable_cairo_canvas=yes
>> +fi
>> +
>> +if test "$CAIRO_BY_MISTAKE" = "yes" -a "$enable_cairo_canvas" != "yes"; then
>
> ... and this ^^^ will cause the cairo canvas to be sometimes enabled on
> Windows, where it was intentionally always disabled before.

That logic to enable cairo if librsvg is enabled on windows without 
directx is already in git master buried inside some conditionals, i've 
just made it more clear.

>> +    AC_MSG_ERROR([The cairo canvas should not be used for this platform])
>> +    enable_cairo_canvas=no
>> +fi
>
> hmm... i don't know why we don't have an ENABLE_CAIRO variable
> independent from ENABLE_CAIRO_CANVAS (i think this existed some time in
> the past), but perhaps adding CAIRO to BUILD_TYPE serves the same purpose...

Don't know either, adding an enable cairo swith coud be tricky since 
some other stuff like librsvg depends on it.

The only behaviour change i need here is not adding cairo to BUILD_TYPE 
if enable_cairo_canvas is not set to yes.

thanks,
riccardo