[Bug math/16424] New: Various test failures in math-ldouble on ppc32.

[Bug math/16424] New: Various test failures in math-ldouble on ppc32.

[brooks at gcc dot gnu.org]

https://sourceware.org/bugzilla/show_bug.cgi?id=16424

            Bug ID: 16424
           Summary: Various test failures in math-ldouble on ppc32.
           Product: glibc
           Version: 2.19
            Status: NEW
          Severity: normal
          Priority: P2
         Component: math
          Assignee: unassigned at sourceware dot org
          Reporter: brooks at gcc dot gnu.org

Created attachment 7343
  --> https://sourceware.org/bugzilla/attachment.cgi?id=7343&action=edit
test-ldouble.out

This is failing with GLIBC trunk as of 10am PST on 2014-01-09.

Configure with "../glibc/configure --prefix=/usr CC='gcc -m32' CXX='g++ -m32'
--target=powerpc32-unknown-linux-gnu --build=powerpc32-unknown-linux-gnu
--host=powerpc32-unknown-linux-gnu" on an IBM Power7; run "make" and "make
check".

Aside from bug 16422 (llrint failures) and bug 16423 (logb_downward failure)
that are common with float and double, I'm seeing quite a number of
very-large-ulps failures in the long-double tests on powerpc32.  I'll attach
the full math-ldouble.out file, but here are some notable ones:

----

Failure: cbrt (-0x4p-1024): Exception "Underflow" set
Failure: Test: cbrt (-0x4p-1024)
Result:
 is:         -2.81264428523626182350e-103  -0x1.428a2f98d728ab896ccbp-341
 should be:  -2.81264428523626190396e-103  -0x1.428a2f98d728ae223ddbp-341
 difference:  8.04551122898042983566e-120   0x1.4c688811936040000000p-396
 ulp       :  1461947279494529.0000
 max.ulp   :  0.0000

...

Failure: cbrt (-0x8p-972): Exception "Underflow" set
Failure: cbrt (-0xf.ffffffffffff8p+1020): Exception "Invalid operation" set
Failure: cbrt (-0xf.ffffffffffff8p+1020): Exception "Overflow" set
Failure: Test: cbrt (-0xf.ffffffffffff8p+1020)
Result:
 is:         qNaN
 should be:  -5.64380309412236197351e+102  -0x1.428a2f98d728aac622b1p+341

...

Failure: sqrt_towardzero (0x8.a1p+8): Exception "Inexact" set

...

Failure: tgamma (0x1p-1024): Exception "Underflow" set

...

Failure: drem (-0x1.80000000000002p1, 2.0): Exception "Inexact" set
Failure: remainder (-0x1.80000000000002p1, 2.0): Exception "Inexact" set
Failure: nextafter (1.1, qNaN): Exception "Inexact" set
Failure: nexttoward (1.1, qNaN): Exception "Inexact" set

...

Failure: Real part of: cacos (1.0 - 0x1.fp-1025 i): Exception "Underflow" set
Failure: Real part of: cacos (-1.0 - 0x1.fp-1025 i): Exception "Underflow" set
Failure: Test: Real part of: cacos (0x1.0000000000001p0 + 0x1.fp-1025 i)
Result:
 is:          2.55717850395349441328e-301   0x1.5eb99734b41ee0000000p-999
 should be:   2.55717850395349434261e-301   0x1.5eb99734b41edd458700p-999
 difference:  7.06747072537819626541e-318   0x0.000000015d3c80000000p-1022
 ulp       :  1430472.0000
 max.ulp   :  0.0000

...

Failure: Real part of: cacosh (1.0 - 0x1.fp-1025 i): Exception "Underflow" set
Failure: Real part of: cacosh (-1.0 - 0x1.fp-1025 i): Exception "Underflow" set
Failure: Test: Imaginary part of: cacosh (0x1.0000000000001p0 + 0x1.fp-1025 i)
Result:
 is:          2.55717850395349441328e-301   0x1.5eb99734b41ee0000000p-999
 should be:   2.55717850395349434261e-301   0x1.5eb99734b41edd458700p-999
 difference:  7.06747072537819626541e-318   0x0.000000015d3c80000000p-1022
 ulp       :  1430472.0000
 max.ulp   :  0.0000

...

[similar things for casin, casinh, catan, catanh omitted]

...

Failure: Real part of: csqrt (-0x8p-152 - 0x4.0000000000004p-1024 i): Exception
"Underflow" set
Failure: Test: Real part of: csqrt (-0x8p-152 - 0x4.0000000000004p-1024 i)
Result:
 is:          2.97200210516769104218e-286   0x1.6a09e667f3bce0000000p-949
 should be:   2.97200210516769103231e-286   0x1.6a09e667f3bcdfa95162p-949
 difference:  9.87515318423588497104e-304   0x1.5aba79b5752400000000p-1007
 ulp       :  190615964531346.0000
 max.ulp   :  0.0000

...

est suite completed:
  21433 test cases plus 20373 tests for exception flags and
    20373 tests for errno executed.
  222 errors occurred.

-- 
You are receiving this mail because:
You are on the CC list for the bug.

[Bug math/16424] Various test failures in math-ldouble on ppc32.

[brooks at gcc dot gnu.org]

https://sourceware.org/bugzilla/show_bug.cgi?id=16424

--- Comment #1 from Brooks Moses <brooks at gcc dot gnu.org> ---
Created attachment 7344
  --> https://sourceware.org/bugzilla/attachment.cgi?id=7344&action=edit
test-ildoubl.out

There are also many of the same failures in test-ildoubl.out, so attaching that
as well.

-- 
You are receiving this mail because:
You are on the CC list for the bug.

[Bug math/16424] Various test failures in math-ldouble on ppc32.

[brooks at gcc dot gnu.org]

https://sourceware.org/bugzilla/show_bug.cgi?id=16424

Brooks Moses <brooks at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
   Target Milestone|---                         |2.19

-- 
You are receiving this mail because:
You are on the CC list for the bug.

[Bug math/16424] Various test failures in math-ldouble on ppc32.

[joseph at codesourcery dot com]

http://sourceware.org/bugzilla/show_bug.cgi?id=16424

--- Comment #2 from joseph at codesourcery dot com <joseph at codesourcery dot com> ---
Please don't file "various test failures" bugs; file individual bugs for 
specific concrete problems identified as being different from other 
currently open bugs (and, preferably, identified as bugs in glibc rather 
than GCC - I strongly suspect most spurious underflows for IBM long double 
are actually libgcc bugs), but not omnibus bugs; it's a pain if the state 
of a bug is "partly fixed" as Bugzilla can't track that.  If in doubt, 
take the risk of two bugs being the same problem rather than one bug 
covering two different problems.

Various of the failures you list are not present if you use current GCC 
with my recent libgcc fixes (and Adhemerval's fix for spurious inexact 
exceptions on addition of NaNs).  Others are not present with my GCC patch 
<http://gcc.gnu.org/ml/gcc-patches/2014-01/msg00157.html>; it doesn't make 
sense to study IBM long double failures without all those fixes in your 
libgcc.  Adhemerval said he was looking at the remaining failures, so 
please coordinate with him.

-- 
You are receiving this mail because:
You are on the CC list for the bug.

[Bug math/16424] Various test failures in math-ldouble on ppc32.

[azanella at linux dot vnet.ibm.com]

https://sourceware.org/bugzilla/show_bug.cgi?id=16424

Adhemerval Zanella Netto <azanella at linux dot vnet.ibm.com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |azanella at linux dot vnet.ibm.com

-- 
You are receiving this mail because:
You are on the CC list for the bug.

[Bug math/16424] Various test failures in math-ldouble on ppc32.

[jsm28 at gcc dot gnu.org]

https://sourceware.org/bugzilla/show_bug.cgi?id=16424

Joseph Myers <jsm28 at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|NEW                         |RESOLVED
         Resolution|---                         |INVALID

--- Comment #3 from Joseph Myers <jsm28 at gcc dot gnu.org> ---
Closing as an omnibus bug rather than anything specific.

-- 
You are receiving this mail because:
You are on the CC list for the bug.

[Bug math/16424] Various test failures in math-ldouble on ppc32.

[fweimer at redhat dot com]

https://sourceware.org/bugzilla/show_bug.cgi?id=16424

Florian Weimer <fweimer at redhat dot com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
              Flags|                            |security-

-- 
You are receiving this mail because:
You are on the CC list for the bug.