Re: charset issue on Windows help needed

[Geert Janssens <geert gtk kobaltwit be>]

On Wednesday 10 September 2014 21:25:30 Fernando Rodriguez wrote:
> On Wednesday 10 September 2014 5:07:03 PM Fernando Rodriguez wrote:
> > On Wednesday 10 September 2014 7:37:28 PM Geert Janssens wrote:
> > > Hi,
> > >
> > > I'm stuck on the following issue. Program is GnuCash on Windows.
> > >
> > > Consider the following code snippet:
> > >
> > > struct stat statbuf;
> > > gchar* dirname = g_strdup(g_getenv("GNC_DOTGNUCASH_DIR");
> > > gint rc = g_stat (dirname, &statbuf);
> > >
> > > switch (errno)
> > > {
> > >
> > >   case ENOENT:
> > >     // Directory doesn't exist
> > >     // Here is code to create it which I cut for brevity
> > >     break;
> > >   
> > >   case EACCES:
> > >     // Directory can't be accessed
> > >     exit(1);
> > >   
> > >   case ENOTDIR:
> > >     // Not a directory
> > >     exit(1);
> > >   
> > >   default:
> > >     // Unknown error
> > >     exit(1);
> > >
> > > }
> > >
> > > // Continue code with valid, existing directory
> > > ...
> > >
> > > So this snippet reads the value of environment variable
> > > GNC_DOTGNUCASH_DIR and tests whether this is a valid directory.
> > >
> > > This works fine when GNC_DOTGNUCASH_DIR uses a limited character
> > > set like ascii. For example when set to "c:\gcdev\geert" this
> > > works well and the code continues.
> > >
> > > However if set to for example:
> > > c:\gcdev\Łukasz
> > > Things go wrong (note the unusual Ł).
> > >
> > > In this case the code branches into case ENOENT and creates a
> > > directory named c:\gcdev\Lukasz (note the plain L now)
> > > Before it continues.
> > >
> > > Setting a breakpoint at rc=g_stat... and examining the value of
> > > dirname at that point also shows it to have a value of
> > > c:\gcdev\Lukasz (with plain L).
> > >
> > > So it seems I'm losing diacritical information here and I can't
> > > pass the right directory to my code to use.
> > >
> > > What should I do to get the real value from the environment to be
> > > able to access the true directory ?
> > >
> > > Thanks,
> > >
> > > Geert
> > >
> > > P.S. my locale settings are all "Dutch_Belgium.1252" except for
> > > LC_ALL which is empty.
> > > _______________________________________________
> > > gtk-app-devel-list mailing list
> > > gtk-app-devel-list gnome org
> > > https://mail.gnome.org/mailman/listinfo/gtk-app-devel-list
> >
> > Well gchar is a typedef for char so it only supports ascii. I think
> > you'll probably have to use Win32 API calls on Windows to access
> > multibyte file names.
> >
> > Also according to the documentation you should use GStatBuf instead
> > of struct stat on Windows to get consistent behaviour on different
> > compilers.
> >
> >
> >
> > _______________________________________________
> > gtk-app-devel-list mailing list
> > gtk-app-devel-list gnome org
> > https://mail.gnome.org/mailman/listinfo/gtk-app-devel-list
>
> Actually try using g_utf8_strncpy() instead of g_strdup(). I'm not
> sure if that'll get you all the way as the docs say g_stat calls the
> standard library functions and AFAIK those dont support
> unicode/multibyte filenames.
Thanks for the suggestion.

However it won't make a difference because the string I get from g_getenv is already wrong. I 
can see this in gdb if I set a breakpoint on the g_stat line.

If I issue 
print (gchar*) g_getenv ("GNC_DOTGNUCASH_DIR")
it prints
"c:\\gcdev\\Lukasz"

So you may well be right I need something other than g_stat. I can only test that however 
when I manage to get the right string from the environment. :(

Geert