Redis-1.0.gir does not work properly with vala #1

Closed
fabiand opened this Issue Oct 22, 2012 · 7 comments

Comments

Projects
None yet
3 participants
Contributor

fabiand commented Oct 22, 2012

Basically it would be nice if the gir was usable with vala, currently a couple of errors are thrown:

[fabiand@apu tests]$ ls
Makefile.include  test-redis-client.c  vredis.vala  vredis.vala~
[fabiand@apu tests]$ cat vredis.vala

using Redis;

void main()
{
//    var client = Redis.Client();
}
[fabiand@apu tests]$ valac --verbose --girdir=.. --pkg Redis-1.0 vredis.vala 
Loaded package `/usr/share/vala-0.16/vapi/glib-2.0.vapi'
Loaded package `/usr/share/vala-0.16/vapi/gobject-2.0.vapi'
Loaded package `../Redis-1.0.gir'
Loaded package `/usr/share/gir-1.0/GLib-2.0.gir'
Loaded package `/usr/share/gir-1.0/GObject-2.0.gir'
Loaded package `/usr/share/gir-1.0/Gio-2.0.gir'
Gio-2.0.gir:44356.7-44356.47: warning: Virtual method `G.Resolver.lookup_service_async' conflicts with method of the same name
      <virtual-method name="lookup_service_async">
      ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
Gio-2.0.gir:46096.7-46096.31: warning: Signal `G.Settings.change_event' conflicts with method of the same name
      <virtual-method name="change_event">
      ^^^^^^^^^^^^^^^^^^^^^^^^^

(valac:10800): GLib-CRITICAL **: g_str_has_suffix: assertion `str != NULL' failed

** (valac:10800): CRITICAL **: vala_gir_parser_node_lookup: assertion `name != NULL' failed
Gio-2.0.gir:59053.7-59057.24: error: `UnixSocketAddress' already contains a definition for `abstract'
Gio-2.0.gir:58925.7-58927.103: note: previous definition of `abstract' was here
Compilation failed: 1 error(s), 2 warning(s)
[fabiand@apu tests]$ rpm -q glib2
glib2-2.32.4-2.fc17.x86_64
[fabiand@apu tests]$ rpm -q vala
vala-0.16.1-1.fc17.x86_64
[fabiand@apu tests]$ 
Owner

chergert commented Oct 22, 2012

I, too, would love this to work. I haven't yet figured out how to fix that. I think it is a bug in Vala since I don't actually use either of those classes (they simply come from Gio-2.0).

Owner

chergert commented Oct 22, 2012

I just talked to someone from #vala, and they suggested using --pkg gio-2.0 to prevent valac from using the Gio-2.0.gir.

Let me know if that helps.

nemequ commented Oct 22, 2012

As Christian said, the solution is to pass --pkg gio-2.0 to valac.

The reason for this is that Redis-1.0.gir includes Gio-2.0.gir, which Vala can't parse properly without additional metadata and a custom vala file. There are lot of reasons why this is the case, many of which are listed in the Generating a VAPI from a GIR section of the Upstream Guide on the Vala wiki. While both the Vala and GObject Introspection teams are working to resolve these issues, it is unlikely that all of these issues will ever be resolved.

In order to prevent valac from parsing the GIO GIR, you simply have to get valac to parse the GIO VAPI (which is generated from the GIR with the help of the files I mentioned above, and shipped with valac) first. Once it does so, the gio-2.0 package will have been provided, and when valac looks at Gio-2.0.gir and sees a package element for gio-2.0 it will know to ignore the rest of the GIR).

So, if you ever see errors from valac referring to a GIR for which a VAPI exists, the first thing you should try is asking valac to parse the corresponding VAPI.

Contributor

fabiand commented Oct 23, 2012

Okay.
An easy thing to do is then, to generate - additionally to the gir - a vapi for redis-glib using vapigen.

When I tried

vapigen --lib redis-1.0 --pkg gio-2.0 Redis-1.0.gir

I noted that vapigen interpretes ClientError as a n enum, and not es error.

Contributor

fabiand commented Oct 23, 2012

It was mentioned that this shoudl work with a recent g-i version.

nemequ commented Oct 23, 2012

I'm not sure where you got that, but it's no different with git master of gobject-introspection. The issue is that the enum isn't registered with GType. You can use something like http://fpaste.org/yHVp/ to do that, or use glib-mkenums.

Owner

chergert commented Oct 23, 2012

I'm usually pretty bad about creating gtypes for errors. Fixed in master.

fabiand closed this Nov 23, 2012

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment