Re: [PATCH] windows-nat: Decode system error numbers

[Corinna Vinschen <vinschen@redhat.com>]

On Nov  9 16:58, Pedro Alves wrote:
> On Wednesday 09 November 2011 16:42:50, Eli Zaretskii wrote:
> > > Date: Wed, 9 Nov 2011 12:02:40 +0000
> > > From: "Maciej W. Rozycki" <macro@codesourcery.com>
> > > 
> > > +  const char *msg = "Unspecified error.";
> > > +  unsigned long err;
> > > +  char buf[1025];
> > > +  size_t size;
> > > +
> > > +  if (ok)
> > > +    return;
> > > +
> > > +  err = GetLastError();
> > > +  size = FormatMessage (FORMAT_MESSAGE_FROM_SYSTEM
> > > +			| FORMAT_MESSAGE_IGNORE_INSERTS,
> > > +			NULL,
> > > +			err,
> > > +			MAKELANGID (LANG_NEUTRAL, SUBLANG_DEFAULT),
> > > +			buf, (sizeof (buf) - 1) / sizeof (TCHAR), NULL);
> > 
> > Will this DTRT with Cygwin, which AFAIK wants the wide versions of the
> > APIs?  Is, for example, "char buf[1025];" appropriate in that case?
> 
> Yeah, if UNICODE is defined, FormatMessage maps to FormatMessageW and 
> buf will be filled with a wide string, though I'm not sure
> if __USEWIDE implies UNICODE.

Then again, buf is defined as char, not wchar_t or WCHAR.  It would be
better to make buf a wchar_t, always call FormatMessageW, and then call
printf_filtered with %ls as string format.  That should work on all
supported platforms.


Corinna

-- 
Corinna Vinschen
Cygwin Project Co-Leader
Red Hat