Re: [Bug-wget] GNU wget 1.17.1 released

[Eli Zaretskii]

> From: Tim Rühsen <address@hidden>
> Date: Sun, 13 Dec 2015 15:17:02 +0100
> Cc: address@hidden
> 
> Andries, thanks for insisting.
> 
> As Andries says, I came up with a polished version of his patch (17th 
> August), 
> but got no review resp. 'ok for pushing'.

AFAICS, the patch you posted does not cover the discussion in its
entirety, or at least doesn't follow the agreement reached near its
end.  I proposed a method to deal with the problem reported by
Andries, in a way that will work on Windows as well:

  http://lists.gnu.org/archive/html/bug-wget/2015-08/msg00154.html

Let me summarize it:

 . If the user asked for unmodified file names, do nothing with them

 . Otherwise, convert file names from their remote charset to the
   local charset using 'iconv'

 . 'iconv' needs the from-charset and the to-charset, which should be
   computed as follows:

   . if the user specified a from-charset, use that; otherwise assume
     UTF-8
   . if the user specified to-charset, use that; otherwise call
     nl_langinfo(CODESET) to find out the current locale's encoding,
     and use that

 . If 'iconv' fails, convert to ASCII using %NN hex-encoding

I believe the above is portable, with the sole exception of
nl_langinfo, which doesn't exist on Windows.  But there's a Gnulib
replacement; alternatively, a simple replacement can be put on
mswindows.c.

Comments?