Bug 59438

(In reply to comment #0)
> Currently the alias doesn't respect the tests.

I think it was fixed in 2.10.0 with http://cgit.freedesktop.org/fontconfig/commit/?id=ddefa5021f7785514f373aab6a8e6191a867278e

<alias>
        <test name="lang">
                <string>ja</string>
        </test>
        <test name="family" qual="all" compare="not_eq">
                <string>IPAGothic</string>
        </test>
        <family>sans-serif</family>
        <default><family>IPAGothic</family></default>
</alias>

$ FC_DEBUG=1028 fc-match
...
        Loading config file /home/tagoh/.config/fontconfig/conf.d/00-test-in-alias.conf
Add Subst match
        pattern any lang Equal "ja"
        pattern all family NotEqual "IPAGothic"
        pattern any family Equal(ignore blanks) "sans-serif"
edit
        Edit family AppendLast "IPAGothic";

That said, as you say fontconfig does deal with <alias> inside <match> too:

        <match>
                <alias>
                        <family>sans-serif</family>
                        <default><family>foo</family></default>
                </alias>
        </match>

$ FC_DEBUG=1028 fc-match
...
        Loading config file /home/tagoh/.config/fontconfig/conf.d/000-bz59438.conf
Add Subst match
        pattern any family Equal(ignore blanks) "sans-serif"
edit
        Edit family AppendLast "foo";

Add Subst match
edit

Proposed fix:
  http://cgit.freedesktop.org/~tagoh/fontconfig/commit/?h=bz59438

Hmm, that just looks like one of issues because the elements are parsed by FcEndElement() when an element enclosure is being closed. I guess there should be much better approach to get all happier. that said, the code itself doesn't know what element is allowed to be the parent but DTD. I suppose libxml is capable to check it with enabling the feature.

Or we can make it valid to have alias in match.  Otherwise, lets just close this.  Having a way to enable DTD verification (envvar and cmdline arg to fc-* tools) in libxml is useful.

Okay, here is another proposal:
  http://cgit.freedesktop.org/~tagoh/fontconfig/commit/?h=bz59438-2

the test case in comment#1 in question would introduces the warning in this case and no more make the empty match pattern:

$ FC_DEBUG=1028 ./fc-match/fc-match
...
Fontconfig warning: "/home/tagoh/.config/fontconfig/conf.d/000-bz59438.conf", line 7: No <test> nor <edit> elements in <match>
...
        Loading config file /home/tagoh/.config/fontconfig/conf.d/000-bz59438.conf
Add Subst match
        pattern any family Equal(ignore blanks) "sans-serif"
edit
        Edit family AppendLast "foo";
...

This doesn't prevent <alias> inside <match> exactly though.

(In reply to comment #3)
> Or we can make it valid to have alias in match.  Otherwise, lets just close
> this.  Having a way to enable DTD verification (envvar and cmdline arg to
> fc-* tools) in libxml is useful.

Sure. but I suppose that feature isn't available in expat. it may be a good opportunity to switch the default parser to libxml though, as we had Bug#34266 in bz, we need to polish a bit for libxml support perhaps. there seems some distros enabling it by default, it's not used in Fedora. so I'm not sure how better it works on. maybe I should try...

just merged a patch in comment#4 regarding Bug#71085. let me close this at this moment so that it works enough for this purpose. and may be better filing an RFE for DTD validation perhaps.