> On Mon, 22 May 2000, Dmitry Timoshkov wrote:
>
> > James Sutherland <[EMAIL PROTECTED]> wrote:
> >
> > >> I would suggest to use FormatMessage() API instead of
> inventing new
> > >> interface. You even could add new error codes if you
> wish to extend
> > >> an existing functionality.
> > >
> > >Which API is that? It looks like a Win32 one from here,
> which tends to
> > >suggest I can't use it in Unix programs...
> >
> > Certainly, we are talking about the Wine debugging messages, no?
> > Wine internally should use its own (i.e. Win32) API or, in
> rare exceptions,
> > services provided by the underlying OS.
>
> The library I am talking about is for Unix programs. In Wine's case, I
> would consider internal debugging/failure messages to be a "rare
> exception" where Wine should use underlying services.
> Otherwise, how do
> you debug the debugger if it depends on itself?! (Imagine
> trying to print
> the error message for "Wine file and memory subsystem hosed" or
> equivalent...)
The point that you seem to have missed is that it is perfectly possible
to use something that has the same syntax and semantics as FormatMessage
without actually using the real FormatMesage that I admit has the
problems that you point out.
In short, of course the debug message version of FormatMessage must only
use Unix API:s that goes without saying. The real question is whether
using a FormatMessage like syntax is good or not, the rest is just
implementation details.
