Re: Filename encoding in FileChooser

[muppet <scott asofyet org>]

On Aug 11, 2004, at 5:38 PM, Ross McFarland wrote:

> On Wed, 2004-08-11 at 15:15, Daniel Flemming wrote:
> > Hmmm, that's interesting reading, and relevant, but it doesn't really 
> > answer my question. :) I have a FileChooserDialog set to 
> > GTK_FILE_CHOOSER_ACTION_SAVE, and I read its filename with a 
> > get_filename method call after running it with a run method call. 
> > What I want to know is, what encoding is put into the resulting 
> > scalar (from the get_filename call)? I thought it was probably 
> > Unicode because that's what gtk uses in general, and also the 
> > resulting filenames kinda looked like Unicode. But I wanted to make 
> > sure.

have you actually run into any problems, or are you just looking for 
what's the right thing up front?


> i don't know, but from what i read and seem to remember from past 
> dealings with the stuff is that the filename encoding isn't anything 
> in particular (depends on what the system's encoding is) and these 
> functions help you make sense of that. once again this is all 'i 
> thinks' and no 'i knows.'

that was certainly the case for the GtkFileSelection --- the filenames 
are in local encoding.  i do not know about GtkFileChooser.  let's have 
a look:

the gtk_file_path_* functions which actually manipulate paths in the 
file chooser are currently implemented just as macros, and they are 
compared with strcmp().  also, in gtk_file_system_unix_create_folder() 
(an internal function), the GtkFilePath is passed directly to mkdir(), 
and if mkdir() fails, it is converted to utf8 for display in an error 
message (actually a GError).  so, the GtkFilePath is most likely in a 
local encoding (the same as what's on disk).

gtk_file_chooser_get_file_name() returns the result of 
gtk_file_system_path_to_filename(), which invokes a vtable method; 
choosing the GtkFileSystemUnix implementation, that function is 
gtk_file_system_unix_path_to_filename(), which does this:

  static gchar *
  gtk_file_system_unix_path_to_filename (GtkFileSystem     *file_system,
                                       const GtkFilePath *path)
  {
    return g_strdup (gtk_file_path_get_string (path));
  }

based on this reading, i would expect that the result of 
gtk_file_chooser_get_filename() is in local encoding (the same as on 
your disk, not utf8).


however, at the binding level, since gtk_file_chooser_get_filename() 
returns a gchar, we run that through the T_GCHAR typemap, which results 
in this code:

  XS(XS_Gtk2__FileChooser_get_filename)
  {
      dXSARGS;
      if (items != 1)
        Perl_croak(aTHX_ "Usage: Gtk2::FileChooser::get_filename(chooser)");
      {
        gchar_own *     RETVAL;
        GtkFileChooser *        chooser = SvGtkFileChooser (ST(0));

        RETVAL = gtk_file_chooser_get_filename(chooser);
        ST(0) = sv_newmortal();
        /* used when we can directly own the returned string. */
        /* we have to copy in the case when perl's malloc != gtk's malloc,
           * so best copy all the time. */
        sv_setpv ((SV*)ST(0), RETVAL);
        SvUTF8_on (ST(0));
          g_free (RETVAL);

      }
      XSRETURN(1);
  }


we're setting the sv's utf8 flag on, which means perl will treat it as 
utf8, but since it probably is *not* utf8, this may be a bug in the 
bindings.

can anyone confirm this?


--
The door is locked.  I tried to open it, but the lock is harder to pick 
than a broken nose.
  -- Sensei, on 'I, Ninja'