Re: [gtk-osx-users] Pango error implicit declaration of function is invalid in C99.

[John Ralls <jralls ceridwen us>]

> On Sep 2, 2020, at 8:57 AM, Pascal <p p14 orange fr> wrote:
>
> > Le 2 sept. 2020 à 16:30, John Ralls <jralls ceridwen us> a écrit :
> >
> > > On Sep 2, 2020, at 3:24 AM, Pascal <p p14 orange fr> wrote:
> > >
> > > Hello John,
> > >
> > > > Le 25 août 2020 à 17:21, John Ralls <jralls ceridwen us> a écrit :
> > > >
> > > > > On Aug 25, 2020, at 3:47 AM, Pascal <p p14 orange fr> wrote:
> > > > >
> > > > > > Le 22 août 2020 à 12:11, Pascal <p p14 orange fr> a écrit :
> > > > > >
> > > > > > > Le 22 août 2020 à 10:39, Pascal <p p14 orange fr> a écrit :
> > > > > > >
> > > > > > > Hello,
> > > > > > > I've a compiling error with Pango:
> > > > > > >
> > > > > > > % jhbuild build meta-gtk-osx-gtk3
> > > > > > > ...
> > > > > > >
> > > > > > > *** Error during phase build of pango: ########## Error running ninja   *** [19/29]
> > > > > > >
> > > > > > > [1] Rerun phase build
> > > > > > > [2] Ignore error and continue to install
> > > > > > > [3] Give up on module
> > > > > > > [4] Start shell
> > > > > > > [5] Reload configuration
> > > > > > > [6] Go to phase "wipe directory and start over"
> > > > > > > [7] Go to phase "configure"
> > > > > > > [8] Go to phase "clean"
> > > > > > > choice: 
> > > > > >
> > > > > > Digging in the error, the define CAIRO_HAS_FC_FONT is not set.
> > > > > > It seems that fontconfig should have been built before cairo.
> > > > >
> > > > > Fontconfig is in the "after" section of Cairo, so it should be ok:
> > > > >
> > > > > <autotools id="cairo"
> > > > >         autogenargs="--enable-pdf --enable-quartz --disable-xlib --without-x">
> > > > > <branch repo="freedesktop" module="cairo" revision="1.16"/>
> > > > > <dependencies>
> > > > >  <dep package="pixman"/>
> > > > >  <dep package="meta-gtk-osx-bootstrap"/>
> > > > >  <dep package="freetype-no-harfbuzz"/>
> > > > > </dependencies>
> > > > > <after>
> > > > >  <dep package="fontconfig"/>
> > > > > </after>
> > > > > </autotools>
> > > > >
> > > > > Is it thus correct with the following fix?
> > > > >
> > > > > diff --git a/modulesets/gtk-osx.modules b/modulesets/gtk-osx.modules
> > > > > index 297b4fe..ed49555 100644
> > > > > --- a/modulesets/gtk-osx.modules
> > > > > +++ b/modulesets/gtk-osx.modules
> > > > > @@ -107,10 +107,8 @@
> > > > >   <dep package="pixman"/>
> > > > >   <dep package="meta-gtk-osx-bootstrap"/>
> > > > >   <dep package="freetype-no-harfbuzz"/>
> > > > > -    </dependencies>
> > > > > -    <after>
> > > > >   <dep package="fontconfig"/>
> > > > > -    </after>
> > > > > +    </dependencies>
> > > > > </autotools>
> > > > >
> > > > > Should I then push it to github?
> > > >
> > > > You mean make a PR? You can't push directly to the repo and in any case the canonical repo is 
> > > > https://gitlab.gnome.org/GNOME/gtk-osx.
> > > >
> > > > However, FC is supposed to be optional in Pango, see the last line of meson_options.txt
> > >
> > > Yes I got the same, indeed pango doesn't directly depend on fontconfig but expect that CAIRO_HAS_FC_FONT 
> > > is defined in cairo include files. It is defined when fontconfig is seen when building cairo thus 
> > > fontconfig should be built before cairo.
> > > So I was wondering if the present cairo attributs in gtk-osx.modules should be changed.
> > >
> > > > I'm able to build meta-gtk-osx-gtk3 without FontConfig so I wonder how you got into a situation where 
> > > > Pango thinks it should use it but Cairo does not.
> > >
> > > Cairo configuration:
> > > cairo (version 1.16.0 [release]) will be compiled with:
> > > The following surface backends:
> > > Image:         yes (always builtin)
> > > <...>
> > > The following font backends:
> > > User:          yes (always builtin)
> > > FreeType:      yes
> > > Fontconfig:    no (requires fontconfig >= 2.2.95)
> > > Win32:         no (requires a Win32 platform)
> > > Quartz:        yes
> > >
> > > Pango configuration:
> > > *** Configuring pango *** [19/29]
> > > <...>
> > > Project name: pango
> > > Project version: 1.44.7
> > > <...>
> > > Dependency fontconfig found: NO unknown version, but need: ['>= 2.11.91']
> > > Run-time dependency fontconfig found: YES
> > > Message: fontconfig has FcWeightFromOpenTypeDouble: NO
> > >
> > > Would you get my full logs to your private e-mail?
> >
> > Not particularly. 
> >
> > I'd rather not require building fontconfig if it's not necessary and while I don't have a recent stable 
> > build without fc, I do have a recent unstable one without it so I suspect that they've fixed the 
> > dependency. There's a new release, 1.46.1. Can you test that today?
>
> I've got the same errors:
> pango-1.46.1/pango/pangocairo-fcfont.c:71:10: error: implicit declaration of function 
> 'cairo_ft_font_face_create_for_pattern' is invalid in C99 [-Werror,-Wimplicit-function-declaration]
>  return cairo_ft_font_face_create_for_pattern (fcfont->font_pattern);
>         ^
> pango-1.46.1/pango/pangocairo-fcfont.c:71:10: note: did you mean 'cairo_ft_font_face_create_for_ft_face'?
> /opt/xnadalib-2020/include/cairo/cairo-ft.h:56:1: note: 'cairo_ft_font_face_create_for_ft_face' declared 
> here
> cairo_ft_font_face_create_for_ft_face (FT_Face         face,
> ^
> 1 error generated.
> pango-1.46.1/pango/pangocairo-fcfontmap.c:109:5: error: implicit declaration of function 
> 'cairo_ft_font_options_substitute' is invalid in C99 [-Werror,-Wimplicit-function-declaration]
>    cairo_ft_font_options_substitute (pango_fc_fontset_key_get_context_key (fontkey),
>    ^
> pango-1.46.1/pango/pangocairo-fcfontmap.c:109:5: note: did you mean 'cairo_font_options_status'?
> /opt/xnadalib-2020/include/cairo/cairo.h:1397:1: note: 'cairo_font_options_status' declared here
> cairo_font_options_status (cairo_font_options_t *options);
> ^
> 1 error generated.
>
> HTH, Pascal.

Indeed it does. What's happening is that Pango's meson.build insists that it must have either both or neither 
of freetype and fontconfig. So by building freetype-no-harfbuzz to get a more complete harfbuzz I'm tripping 
up pango. Unfortunately gitlab.gnome.org seems to be broken ATM so I can't open an issue about it, but I'll 
try rearranging things a bit to work around the problem.

Regards,
John Ralls