Re: [Devel] new FT error scheme

[Werner LEMBERG]

> > The current code like
> > 
> >   #define TT_Err_xxx  FT_Err_xxx
> > 
> > isn't really meaningful.
> > 
> > What do you think about the following:
> > 
> >   #define FT_Err_Ok   0x00
> >   #define FT_Err_Foo  0x01 + FT_ERR_OFFSET
> >   #define FT_Err_Bar  0x02 + FT_ERR_OFFSET
> >   ...
> > 
> > In freetype.h:
> >   #undef  FT_ERR_OFFSET
> >   #define FT_ERR_OFFSET 0x000
> > 
> > In module foo:
> > 
> >   #undef  FT_ERR_OFFSET
> >   #define FT_ERR_OFFSET 0x100
> > 
> > This would give better error codes (the high byte gives the module,
> > the low byte the error code), and we could uniformly use FT_Err_*
> > everywhere.
> >
>  
> Hummm.. I don't understand your scheme. It seems that the value of
> FT_Err_Foo would not be the same if you include "foo.h" or
> "freetype.h", which is hardly a good idea :-)
> 
> Is that really what you meant?

Basically yes.  Consider the error `glyph not found'.  This is, say,
0x02.  If this error occurs in module `xxx', you'll get e.g. 0x102.
If this occurs in module `yyy', the error value is e.g. 0x402.  If you
AND the error code with 0xFF, you get the `real' error code (as
presented in freetype.h).  If you AND with 0xFF00 you get the module
in which the error has happened.  This is a useful additional
information IMHO.


    Werner