Re: lynx-dev [PATCH] to dev15 - fix for win32

[Klaus Weide]

On Mon, 22 Nov 1999, Leonid Pauzner wrote:

> > * Bug in LYGetHostByName code fixed (Windows only problem). - SH
> 
> This is not windows-only changes - there is the #else case.
> 
> [I have no sources handy but seems strange how else that was worked
> on Unixes previously: valid_hostname(host) was called for unitialized
> argument?]

No, the `host = str;' that comes later is an unnecessary duplication; the
function begins with:

PUBLIC struct hostent * LYGetHostByName ARGS1(
        CONST char *,   str)
{
#ifndef _WINDOWS_NSL
    CONST char *host = str;
#endif /* _WINDOWS_NSL */


When LYGetHostByName is called from HTParseInet (the usual case, except
for "guessing"), then the _WINDOWS_NSL-specific `strncpy(host, str,
(size_t)512);' in LYGetHostByName is *also* a duplication of something
that already happens in HTParseInet.  (In fact it seems the strncpy in
LYGetHostByName would copy `host' onto itself - `str' already IS a
pointer to `host' -, that doesn't look like healthy practice.)

> > --- lynx283.d15/WWW/Library/Implementation/HTTCP.c    Thu Oct 21 23:56:48 
> > 1999
> > +++ lynx283.w32/WWW/Library/Implementation/HTTCP.c    Mon Nov 22 10:21:04 
> > 1999
> > @@ -693,6 +693,12 @@
> >       return NULL;
> >      }
> 
> > +#ifdef _WINDOWS_NSL  /* 1999/11/21 (Sun) 21:35:19 */
> > +    strncpy(host, str, (size_t)512);
> > +#else
> > +    host = str;
> > +#endif /* _WINDOWS_NSL */
> > +
> >      if (!valid_hostname(host)) {
> >       lynx_nsl_status = HT_NOT_ACCEPTABLE;
> >  #ifdef NO_RECOVERY
> > @@ -704,12 +710,6 @@
> >  #endif
> >       return NULL;
> >      }
> > -
> > -#ifdef _WINDOWS_NSL
> > -    strncpy(host, str, (size_t)512);
> > -#else
> > -    host = str;
> > -#endif /*  _WINDOWS_NSL */
> 

   Klaus