Sorry, I meant gobject-introspection fix from bug #663048

+INTROSPECTION_GIRS = Rsvg-@RSVG_API_MAJOR_VERSION@.0.gir
[...]
+Rsvg-@RSVG_API_MAJOR_VERSION@_0.gir: librsvg-@RSVG_API_MAJOR_VERSION@.la

Shouldn't hardcode the 0, but should define RSVG_API_MINOR_VERSION=0 in configure.in and use that instead.

+Rsvg_@RSVG_API_MAJOR_VERSION@_0_gir_FILES = $(addprefix $(srcdir)/, $(introspection_files))

This is wrong, not all files in $introspection_files exist in srcdir.

Thanks for the quick review. All of the files are in srcdir according to my eyes - what am I missing?

librsvg-feature.h and librsvg-enum-types.[ch] are in builddir. 

I think using ..._gir_FILES = $(librsvg_@RSVG_API_MAJOR_VERSION@_la_SOURCES) directly (and adding the include headers to that variable, which is the correct thing to do anyway) should work.

I've done that on the 'introspection' branch; can you check that this works (I don't have introspection buildable here).

Why is libxml-2.0 in the INCLUDES? librsvg exposes no API with it; it's only used internally.

..._gir_CFLAGS had $(INCLUDES) in it which is not used in librsvg's Makefile.am at all; did you mean to put $(librsvg_@RSVG_API_MAJOR_VERSION@_la_CFLAGS) (and/or _CPPFLAGS) instead, or is it ok to just omit that (as I have done) ?

Without libxml2 stuff, this happens:

  GISCAN Rsvg-2.0.gir
In file included from /home/dsd/projects/librsvg/rsvg-css.h:31:0,
                 from <stdin>:5:
/home/dsd/projects/librsvg/rsvg-private.h:33:24: fatal error: libxml/SAX.h: No such file or directory
compilation terminated.
Error while processing the source.
make[2]: *** [Rsvg-2.0.gir] Error 1

Created attachment 200359 [details] [review]
fixup patch

Here is a patch for your introspection branch which makes it work for me. I also added some extra annotations needed by Sugar. I'll also attach a python test case which uses these functions.

Created attachment 200360 [details]
python test app

-Rsvg_@RSVG_API_VERSION_U@_gir_NAMESPACE = RSVG
+Rsvg_@RSVG_API_VERSION_U@_gir_NAMESPACE = Rsvg

Is that just a preference, or required? (I don't really care if it's capitalised or not, just curious.)

Could you also attach the generated .gir file, for inspection?

(In reply to comment #5)
> Without libxml2 stuff, this happens:
> 
>   GISCAN Rsvg-2.0.gir
> In file included from /home/dsd/projects/librsvg/rsvg-css.h:31:0,
>                  from <stdin>:5:
> /home/dsd/projects/librsvg/rsvg-private.h:33:24: fatal error: libxml/SAX.h: No
> such file or directory
> compilation terminated.
> Error while processing the source.
> make[2]: *** [Rsvg-2.0.gir] Error 1

Hmm. Maybe this is a case for _gir_PACKAGES instead of _gir_INCLUDES ?

The NAMESPACE change is required. Otherwise the error is:

  GISCAN Rsvg-2.0.gir
Namespace is empty; likely causes are:
* Not including .h files to be scanned
* Broken --identifier-prefix


I can put libxml2 in PACKAGES rather than INCLUDES and indeed it works - nice. So I guess the following is correct:

Rsvg_@RSVG_API_VERSION_U@_gir_PACKAGES = \
	libxml-2.0 \
	pango \
	$(NULL)
Rsvg_@RSVG_API_VERSION_U@_gir_INCLUDES = \
	GLib-2.0 \
	Gio-2.0 \
	GdkPixbuf-2.0 \
	cairo-1.0 \
	$(NULL)

That is, Requires specified by librsvg-2.0.pc are listed in INCLUDES, and the other stuff required for g-ir-scanner to work is listed in PACKAGES.

Created attachment 200364 [details]
output gir file

Thanks, looks ok to me.

I've squashed the commits and pushed to master; please verify I got everything right :-)

Thanks! Almost but not quite - g-ir-scanner fails:
  GISCAN Rsvg-2.0.gir
Package libxml2-2.0 was not found in the pkg-config search path.
Perhaps you should add the directory containing `libxml2-2.0.pc'
to the PKG_CONFIG_PATH environment variable
No package 'libxml2-2.0' found
Package Pango-1.0 was not found in the pkg-config search path.
Perhaps you should add the directory containing `Pango-1.0.pc'
to the PKG_CONFIG_PATH environment variable
No package 'Pango-1.0' found


Please check the PACKAGES and INCLUDES lines against the ones I suggested above.

Hmm, that's a bad trap. Should be fixed on master now.

Thanks for being so responsive here and for cleaning up my work - its now working fine! Looking forward to using this in Sugar very soon :)