glibc 2.1.94 testing on alpha

[Ulrich Drepper]

Christian Iseli <chris@ludwig-alpha.unil.ch> writes:

> I updated my binutils, gcc, and glibc (CVS from today) and reran
> make check.

Your compiler seems t obe completely broken, at least when it comes to
math.  The test-float.out file contains

Failure: Test: frexp (12.8, &x) == 0.8
Result:
 is:          1.28000001907348632812e+01   0x1.99999a0000000^@^@^@^@^@^@^@p+3
 should be:   8.00000011920928955078e-01   0x1.99999a0000000^@^@^@^@^@^@^@p-1
 difference:  1.20000000000000000000e+01   0x1.8000000000000^@^@^@^@^@^@^@p+3
 ulp       :  201326592.0000
 max.ulp   :  0.0000


This is completely beyond everything one can reasonably expect (the
garbage at the end of the numbers left aside).  The printf output
problem can very well be associated with this.  I ran Alpha tests here
and test-float contains not a single problem and where output is
printed it looks reasonable.

This is with a mainline gcc.  I don't know which one you should use.

-- 
---------------.                          ,-.   1325 Chesapeake Terrace
Ulrich Drepper  \    ,-------------------'   \  Sunnyvale, CA 94089 USA
Red Hat          `--' drepper at redhat.com   `------------------------