Re: Fixes for GNU Sather 1.2.2

[Sather User]

Michael,

Thank you again for a detailed reply.

I won't draw it out further, beyond to lament that cc, cp, cd, as, ar,
nm and quite a few other programs have not yet been renamed in
compliance with the imperative that you so eloquently demonstrate.  It
was so obliging of GNU Sather to vacate that space!

(Who cares that Autoconf allows gcc to be renamed at compile time?)

And to suggest that further Sather development should avoid
unnecessary changes that are conspicuous, and straightforward to do,
but perhaps lack any merit other than being conspicuous.  The Sather
community is not yet so numerous that further fragmentation of it
(and, again, I agree that I am subscribed to the wrong list), or
changing features that seem to be characteristic of the language, can
be a good idea.  Build.  Don't destroy.

It is relevant to question the versioning.  In Library/Base/
flt_other.sa, both in Sather-1.2b and in GNU sather-1.2.3, I read:
"Arbitrary precision floating point.  Not yet implemented."  At many
other points there is plainly unfinished business.  Reasonably, Sather
1.2 won't come out of beta until some of them have been dealt with.

Just one more: Dispatching of routine and iter closures is not
implemented yet.  The compiler will let you assign one closure to
another only if they have exactly the same static types.

E.g. I can do

      opt.addopt("add", 'a', #(no_argument), bind(r_add(_,_)));

where:

      addopt(name:STR, val:CHAR, ha:HA_ENUM, fun:ROUT{CHAR, STR})
        is ... end;

but not where:

      addopt(name:STR, val:CHAR, ha:HA_ENUM, fun:$OB) is ... end;

I've taken that example from Ben Gomes's Contrib/gomes/get_opt.sa
after -convert (cs option, presumably sacomp option) from Sather-1.0
to Sather-1.1 code.

If GNU Sather-1.2.3 is really Sather-1.2b4, as I think it is,
Sather-1.2 will not be a patch-level release, and that particular
argument for restricting it to 32-bit integers on all architectures
disappears.

Oh, and by the way.  At my ALGOL 60 programming course the program
listings, printed by the Flexowriter from paper tapes (of course,
there were no CRT computer monitors) were all lower case, although the
keywords were underlined.  That was so nice and so civilized, compared
with that shitty FORTRAN (3 or 4?) which was all caps.  So when C came
along one of its plusses was lower case.

That doesn't mean I can't adjust to Sather class names.  It is what it
is.  Get over it.

Regards,
Mike

-- 
Michael Talbot-Wilson