h.g. muller wrote:
To return to thi issue: I did not succeed in testing the Polyglot from
the Gold Pack with Win95. When I tried to write it on a floppy from my
Win2K machine (where I brought it with a memory stick), the floppy
invriable becomes unreadable both on my Win95 laptop and the Win2k
machine, and is even unformattable on the Win2k machine after that.
Wen I reboot the Win2k machine under Win98, the Polyglot seems to
work, though. But that probably doen't prove anything, as I suppose
the problem you mension is specific to Win95 only.
It is. I have a 95 machine that I keep around for one very old
hardware-specific game. But I do test chess software on it from time to
time, mostly because I can do so easily-- it's on my home network, so I
don't need to use sneaker-net with a floppy. Engines are what bother me
the most when it comes to 95 support. They don't make any system calls
that 95 doesn't have, but they won't work because they were compiled
with Visual Studio 2005 or later. (Its CRT initializes by making calls
95 doesn't have)
It is technically possible to create 95 compatible programs using Visual
Studio 2005, but you have to go through extra steps to do so, and in
practice, nobody does. The latest Hiarcs to work in 95 is 10, Rybka
1.2f doesn't work in 95, and yet the latest naum and bright work in 95.
I have several Togas that work in 95, but most speed-compiles tend not
to because of the compiler used (Cyclone 3.4 being one exception that
DOES work in 95). I have my own speed compiles of various Toga versions
that work in 95, and generally are faster too. Heh. Anyway, it's a
crapshoot when it comes to engines.
I would be very interested to have your patches to make Polyglot work
under Win95. Although in case a choice has to be made, I would
definitely prefer including a Polyglot that can run without
cygwin1.dll over one that would work on Win95, it would be good to
have both.
Anything relating to the affinity option must be stripped. You could
leave in the option, and just not make the system call, and that would
work, but then you'd have an uci option that doesn't do anything. I
basically just conditionally compiled that code depending on a
definition: WIN95. I can send you the polyglot source I use, but it's
rather customized, but based on 1.4w12 mostly. Except I've converted it
to C. It doesn't have any of the book changes by Michel van den Bergh,
nor any of his winboard extensions.
I use the same source for windows and linux, which, btw, is why I am in
favor of linking with cygwin1.dll-- that way you run much more code in
common. That makes it very easy to keep source that compiles in either
OS, and keep out bugs specific to one OS.
