Re: [Vala] [patch] G.Scanner

[Xavier Bestel <xavier bestel free fr>]

Hi,

Le mardi 27 octobre 2009 à 18:10 +0100, Julian Andres Klode a écrit :
> Am Dienstag, den 27.10.2009, 17:55 +0100 schrieb Xavier Bestel:
> > Hi,
> >
> > could someone apply that please ?
> >
> > Thanks,
> >     Xav
> >
> > --- /usr/share/vala/vapi/glib-2.0.vapi.orig 2009-10-20 19:22:51.000000000 +0200
> > +++ /usr/share/vala/vapi/glib-2.0.vapi      2009-10-21 17:10:01.000000000 +0200
> > @@ -2125,13 +2125,17 @@
> >             public TokenValue next_value;
> >             public uint next_line;
> >             public uint next_position;
> > -           public Scanner (ScannerConfig? config_templ);
> > +           public Scanner (ScannerConfig? config_templ = null);
> This seems good to me.
>
> >             public void input_file (int input_fd);
> >             public void sync_file_offset ();
> >             public void input_text (string text, uint text_len);
> >             public TokenType peek_next_token ();
> >             public TokenType get_next_token ();
> >             public bool eof ();
> > +           public int cur_line ();
> > +           public int cur_position ();
> > +           public TokenType cur_token ();
> > +           public TokenValue cur_value ();
> Don't they do the same as the member line, position, token and value do,
> and are thus useless?

They are, but they correctly access the values through methods instead
of direct member access. Direct access is undocumented, so I guess it's
not really part of the API and can change anytime.

> >             public uint set_scope (uint scope_id);
> >             public void scope_add_symbol (uint scope_id, string symbol, void* value);
> >             public void scope_foreach_symbol (uint scope_id, HFunc func);
> > @@ -2213,6 +2217,7 @@
> >     }
> >  
> >     [SimpleType]
> > +   [CCode (destroy_function = "")]
> >     public struct TokenValue {
> >             [CCode (cname="v_symbol")]
> >             public void* symbol;
> I never experienced any kind of problem with TokenValue, and thus I
> can't see why such a patch would be needed. At least everything in
> http://git.debian.org/?p=users/jak/apt2.git;a=blob;f=libapt/configuration.vala;h=524b204049cabfa50f94c1a2aa31838483dcaac4;hb=HEAD
> works perfectly. It may be related to your cur_value() function not
> being weak although it probably should be (as it just returns the
> 'value' member), but I'm not sure here.

I'm not sure either. I tried the weak thing, it doesn't help.
The trouble is that Vala doesn't support union natively, and there's
absolutely no documentation on how the use the metadata to resolve such
a case (and no one answers when I ask).
In the meantime, it's the only solution I found which works.

> Of this patch, the only thing that looks useful to me is the default
> value in the constructor. The remaining changes just seem kind of
> useless to me.

But it doesn't look clean at all to me.

        Xav