Re: MiNT goes UNiX, invitation for mailing list

[mh1@irz301.inf.tu-dresden.de (Michael Hohmuth)]

> > Btw., why not just using __MINT__ as a compiler switch, it is
> > predefined by the MiNTlibs and independent of the compiler.  For
> > compiler dependent code we can use atarist for GCC, ??? for PureC, and
> > so on.

> Because people have already used __MINT__ in an incompatible way, I
> believe. For instance, in tcsh, the startup file is named cshrc.csh
> (not .cshrc) on the atari.

Well, the `tcsh' port uses "#ifdef TOSFS" instead of "#ifdef __MINT__"
for this very purpose.  If you recompile without TOSFS defined, you
get what you want.

I don't know why you always cite the `tcsh' port as an example of how
__MINT__ is incompatibly used.  I believe that the `tcsh' port uses
__MINT__ in the correct way, meaning "if using MiNT Library, do this".
(And I should know, I did it.  ;-)

> So it would beb better to tie the choice between the two types:
> `maximum unix conformance' and `maximum tos compability' to a brand
> new switch, whether that would be a I_WANT_UNIX or I_WANT_TOS.

Perhaps we should extend the MiNT Library with an extra switch
_UNIX_SOURCE meaning "maximum unix conformance".  _TOS_SOURCE could
mean "maximum tos compability".  _MINT_SOURCE could mean what the
current default with mntlib is, but I could also imagine _UNIX_SOURCE
being the default.

Besides, this way we won't need another library.

Michael
-- 
Internet: hohmuth@freia.inf.tu-dresden.de