[Tinymux] TinyMUX for 64-bit Windows

[Stephen Dennis]

On 1/17/06, Michael J Wenk <wenk at praxis.homedns.org> wrote:
> On Mon, Jan 16, 2006 at 09:58:07PM -0800, Stephen Dennis wrote:
> > Perhaps more people use TinyMUX on Windows than are willing to admit
> > it, but at this point, no one should really need TinyMUX compiled
> > natively for 64-bit edition of Windows.
> >
> > However, it does work, and it is available (as a TinyMUX 2.6 Alpha).
> > It doesn't give you anything that you don't already have. All the
> > built-in limits (described in LIMITS) are the same.
> >
> > Jake and I spent several days tag-teaming all of the warnings,
> > behavioral changes, differently-named functions, and seemingly
> > arbitrary behavior differences.  Did you know that VS 2005's
> > localtime() doesn't like dates beyond the year 3000 now, and it
> > helpfully asserts if it sees one.
> >
> > Anyway. Just to be clear. This is something for 2.6. TinyMUX 2.4 does
> > not support this. Even if you have VS 2005, you can't compile TinyMUX
> > 2.4 for Win64.
> >
> > The primary beneficiary of this is probably more precise source code
> > for AMD64 under Linux or FreeBSD. There is also the possibility that
> > someone has access to a DEC Tru64 box. At one time, I ported TinyMUX
> > to Tru64 Unix, and it worked, but there were still a large number of
> > ignorable warnings.  There shouldn't be any more warnings, now.
> >
> >
> > Brazil
>
> Back in the day(96 or 97) we ran TinyMUX (would have been a 1.X version of some
> form) on 64 bit DEC Alpha running either OSF/1 or Digital Unix.  There
> were a couple of warnings we worked through, and (more likely the
> machine itself rather than the compile) we could not use gcc to
> compile it, but it actually ran fairly well.  I didn't really notice any
> benefit to it being 64 bit.
>
> Its cool to see 2.X on 64 bit, tho again, I really don't see the point
> except for those souls that find themselves a site on a 64 bit OS.

My foray into Tru64 would have been in 2001 in Bellevue, WA. Hard disk
space was surprisingly limited, and it did not come with a C++
compiler. Anyway, to make a long story longer, on Tru64 with gcc's C++
compiler, the size of an integer was 4 bytes, but the size of a
pointer was 8 bytes. And, alignment was really weird. Any cast of
pointer to integer was hopelessly broken under those conditions.

Sizes and alignments of 64-bit Windows are rather similar to Tru64.
Size of time_t, size_t, and pointers are 4 bytes. The size of integers
remains 4.

As long as the size of an integer was 8 bytes on the DEC Alpha running
OSF/1 or DEC Alpha, I can see the TinyMUX 1.x codebase working with
warning. Otherwise, I'm not sure how you did it without a lot of
changes.


Brazil