#5569 acsch( 1.0e-6*%i );

[Dan Gildea]

It appears different lisps have different branch cuts for asinh, causing a test suite failure in rtest_trig.mac problem 202 (line 823) for cmucl, ecl, and gcl.

in sbcl:

(%i1) acsch( 1.0e-6*%i);

  0: (ASINH #C(0.0d0 -1000000.0d0))
    1: (ASINH -999999.9999995d0)
    1: COMMON-LISP:ASINH returned -14.508657738523969
  0: COMMON-LISP:ASINH returned #C(-14.508657738523969 -1.5707963267948966)
(%o1)            - 1.5707963267948966 %i - 14.508657738523969

in cmucl, ecl, gcl:

(%i1) acsch( 1.0e-6*%i );

  0: (COMMON-LISP:ASINH #C(0.0 -1000000.0))
  0: COMMON-LISP:ASINH returned #C(14.508657738523969 -1.5707963267948966)
(%o1)             14.508657738523969 - 1.5707963267948966 %i

[David Scherfgen]

Thanks for reporting. Adding these tests paid off!
Fix is on the way.

[David Scherfgen]

Fixed by commit [d18da1].
GCL 2.7.x needs another fix, because it doesn't signal on float division by zero, logarithm of zero, etc., but instead generates infinities. This will be a different commit.