[Cairo] Re: Patch for cairo --without-x

[James Henstridge]

On 29/11/2003 7:04 AM, Sasha Vasko wrote:

>
>
> James Henstridge wrote:
>
>> If we decide to require automake >= 1.7 to build Cairo from CVS 
>> (which I highly recommend), then it should be pretty easy to 
>> conditionally install the cairo-xlib.h with syntax like this:
>>
> <snip>
>
>> This is easy to read, and should work okay.  I would think absense of 
>> a header is a better indication that the feature isn't available, 
>> compared to having a present but effectively empty header installed 
>> on the system
>
>
> Conditional installation of headers is evil, as it requires apps to 
> check not only for cairo, but also for its individual headers, while 
> configuring self. Also it makes impossible explicit include of it in 
> main header cairo.h.

If an app requires xlib support, is it really that much of a problem to 
check for a cairo-xlib.h header?  If using autoconf, it is only a single 
line in the configure.in file to perform such a check.

If you want to automatically include cairo-xlib.h when appropriate, it 
would be just as easy to do:
  #ifdef CAIRO_SUPPORTS_X
  #  include <cairo-xlib.h>
  #endif

If xlib support isn't available, it won't try to include the header.

I would recommend putting definitions like CAIRO_SUPPORTS_X into a 
separate header file (cairo-config.h, maybe?), which could be generated 
in a manner similar to glib's glib-config.h (this is a bit cleaner than 
getting autoconf to expand substitutions in all headers).

>> (would we install a cairo-win32.h header on unix systems in the future?)
>
>
> Well yeah, from this perspective better solution would be to rename 
> header to something like  cairo-interfaces.h, and have all the 
> optional stuff in this single header.

I would expect that the xlib and win32 specific functions would be quite 
different.  One is going to be dealing with X drawables, etc while the 
other will be dealing with the Windows equivalents.

>> For conditional compilation of the actual Cairo code, it would be 
>> better to do it the normal way.  That is, just use an AC_DEFINE() 
>> statement in the configure script to define CAIRO_HAS_X_BACKEND.  If 
>> config.h is 
>
> That was the way I did it at first, but as Carl noted that forces 
> compilation of effectively empty file, which makes conditional 
> compilation a preffered method.

I was actually thinking about using a CAIRO_HAS_X_BACKEND define for 
xlib specific stuff outside of cairo_xlib_surface.c (the makefile 
conditional would handle skipping cairo_xlib_surface.c).  If there isn't 
any such code, then you can ignore the above statement.


James.

-- 
Email: james at daa.com.au
WWW:   http://www.daa.com.au/~james/