Installer for Windows, was: Maxima 5.30.0 / wxMaxima 13.04.1 Windows Binaries

>>>>> "Nico" == Nico Bredenbals <nbred at live.uni-paderborn.de> writes:

    Nico> Hey,
    Nico> 2) seems to be solved by a new Version from Andrej, I will repackage the binary with
    Nico> this version.
    Nico> 1) I came across the two failed tests during "make check" of maxima, but I thought they
    Nico> occured due to bogus tests:
    Nico> 99.25 - (397/4) = 99.25 - 99.25 = 0 which, while not beeing simplified to its fullest,
    Nico> matches the expected result of 0. I'll see if I can find a newer GCL version with ANSI
    Nico> support for windows and check if this changes the test result. Can anyone confirm that
    Nico> this error does not occur on any other platform?
    Nico> BTW, running the testsuite with the current maxima 2.8 windows release gives 2 errors
    Nico> not occuring with my 3.0 compilation:
    Nico> Running tests in rtest16:
    Nico> ********************** Problem 196 ***************
    Nico> Input:
    Nico> ?????? 60
    Nico> ????? 2?? - 1
    Nico> float(-------) - 1
    Nico> ???????? 60
    Nico> ??????? 2
    Nico> Result:
    Nico> - 1.1102230246251565E-16
    Nico> This differed from the expected result:
    Nico> 0.0
    Nico> ********************** Problem 197 ***************
    Nico> Input:
    Nico> ?????? 1000
    Nico> ????? 2???? - 1
    Nico> float(---------) - 1
    Nico> ???????? 1000
    Nico> ??????? 2
    Nico> Result:
    Nico> - 1.1102230246251565E-16
    Nico> This differed from the expected result:
    Nico> 0.0
    Nico> 619/621 tests passed
    Nico> From my point of view this is pretty accurate regarding 32Bit FP, but I am no expert in
    Nico> this field and it's failed test non the less.

These are known issues with gcl.  While the error is small, it should
be zero.  I think the issue is that gcl is not correctly rounding.

Ray