Math test failure (still) on m68k

[Kars de Jong]

Hi,

Well, I hoped 'make check' would not fail, but it did with test-float and
test-ifloat. Looks like the deltas still aren't correct for Linux/m68k.
Did work with the double versions.

testing float (without inline functions)
Fail: j0 (2.0) = 0.22389...
Result:
 is:          2.23890811204910278320e-01   0x1.ca874400000000000000p-3
 should be:   2.23890781402587890625e-01   0x1.ca874000000000000000p-3
 difference:  2.98023223876953125000e-08   0x1.00000000000000000000p-25
Fail: jn (0, 2.0) = 0.22389...
Result:
 is:          2.23890811204910278320e-01   0x1.ca874400000000000000p-3
 should be:   2.23890781402587890625e-01   0x1.ca874000000000000000p-3
 difference:  2.98023223876953125000e-08   0x1.00000000000000000000p-25
Fail: jn (3, 2.0) = 0.12894...
Result:
 is:          1.28943279385566711426e-01   0x1.08136a00000000000000p-3
 should be:   1.28943249583244323730e-01   0x1.08136600000000000000p-3
 difference:  2.98023223876953125000e-08   0x1.00000000000000000000p-25
Fail: jn (10, 2.0) = 2.515...*10^-7
Result:
 is:          2.51538750717372749932e-07   0x1.0e167200000000000000p-22
 should be:   2.51538637030535028316e-07   0x1.0e166a00000000000000p-22
 difference:  1.13686837721616029739e-13   0x1.00000000000000000000p-43

Test suite completed:
  2278 test cases plus 2266 tests for exception flags executed.
  4 errors occured.

Kars.
-- 
------------------------------------------------------------------------------
Kars de Jong             Signaalkamp rules the waves!       Turrican@Discworld
--------======]**-----|      jongk@cs.utwente.nl      |-----**[======---------