Re: config parser patch for more graceful parsing

From: Lorenzo Bettini
> > I hope this patch could be in the next package.
> > 
> > --- gengetopt-2.18.orig/src/skels/c_source.h_skel	Fri Jul 14 20:07:31 2006
> > +++ gengetopt-2.18/src/skels/c_source.h_skel	Tue Nov 21 19:36:40 2006
> > @@ -653,8 +653,11 @@
> >              }
> >          }
> >        else
> > -        { /* read up the remaining part up to a delimiter */
> > -          next_token = strcspn (farg, " \t\r\n#\'\"");
> > +        { /* read up the remaining part up to end of line or comment sign
> */
> > +          char *ch = (str_index + (next_token = strcspn (farg,
> "\r\n#")));
> > +          /* remove trailing whitespaces */
> > +          for ( --ch; (*ch >= '\t' && *ch <= '\r') || *ch == ' '; --ch )
> > +            --next_token;
> >            str_index += next_token;
> >          }
> 
> I'm really sorry about that, it must have escaped me.

Oh never mind that. You can realize now that I could wait for it for a long time ;)

> I noticed that the strcspn does not include ' nor " is that correct?

Yes, but be careful! First it goes to the end of the line and then steps back on (thus removes) trailing whitespaces. If you simply use strcspn() w/o ' and " parser can't cope with the strings mentioned above:

> key  = MC JMS  

This way key_arg will be "MC JMS"

The another benefit of this patch is values containing more than one "word" does not have to be quoted at all.