Re: gs 3.53 and up
> > [...]
> > 3.53 *compiled with
> > optimizations turned off* works perfectly. This evidence leads me to
> > believe that there is nothing wrong with my postscript source (not
> > that I wrote it by hand. I came straight from dvips.). I have solved
> > the problem personally by downloading gs 4.00 and compiling it with no
> > gcc optimizations. However, I have to install it in /usr/local/
> > because I didn't have a functioning debianized version. This means I
> > either have to keep a broken gs 3.x or working gs 2.x installed in
> TeXtpert I'm not, but; [ This is only a guess ]
> Turning off the optimizations and getting _any_ kind of
> different end results, except for a speed or size difference
> should probably be considered a bug in gcc.
Or some other flaky stuff: My Cirix 486 used to play that trick
with me with gs-2.61. Also, I've seen programmes that compare
doubles: the result depends on the optimisation, as without optimisation,
all compares are done at "double" precision, but with optimisation,
sometimes the doubles are still in the FPU, and then the compare
happens at "extended" (80 bits) precision. Not a gcc bug.
> That having been said, you do know there are two different "gs"
> interpreters? gs is an Alladin product, donated to GNU
> (accord. to the Printing HOWTO). As such, the GNU version is
> just an older Alladin version, presumably the newer Alladin
> version has more features, and different bugs.
> So if you have source from one and a binary from the other, you
> may be hunting bugs that are no longer with us...
The latest GNU version is (AFAIK) 3.33. As his problems are
with gs 3.53 and up, he can only be using Alladin version.
Use Debian/GNU Linux!