[Bug math/22826] New: 2.26: float128-related test failures ?

["dilfridge at gentoo dot org" <sourceware-bugzilla at sourceware dot org>]

https://sourceware.org/bugzilla/show_bug.cgi?id=22826

            Bug ID: 22826
           Summary: 2.26: float128-related test failures ?
           Product: glibc
           Version: unspecified
            Status: UNCONFIRMED
          Severity: normal
          Priority: P2
         Component: math
          Assignee: unassigned at sourceware dot org
          Reporter: dilfridge at gentoo dot org
  Target Milestone: ---

Using 2.26 branch I'm getting on Gentoo strange float128-related test failures. 

* binutils 2.29.1
* gcc-6.4
* host system has glibc-2.25
* errors only occur when building the x86 library on a x86_64 system, not on a
pure x86 system.

FAIL: math/test-float128-finite-tgamma
FAIL: math/test-float128-finite-trunc
FAIL: math/test-float128-tgamma
FAIL: math/test-float128-trunc
FAIL: math/test-ifloat128-tgamma
FAIL: math/test-ifloat128-trunc

The test outputs don't look like a rounding error, more like "completely
wrong".

Example: 
testing _Float128 (finite-math-only)
Failure: Test: tgamma (-0x1.3ffffep+4)
Result:
 is:         -1.69685262160625547111063374870045606e-15 
-0x1.e9159fc95cbf97ad7d6c16732224p-50
 should be:   2.15500262145255367562240404835791845e-13  
0x1.e543717808f65ad8ff39bc4af093p-43
 difference:  2.17197114766861623033351038584492295e-13  
0x1.e9159cb79bafda085a349477d6d7p-43
 ulp       :  9919810589228634203671706476205783.0000
 max.ulp   :  4.0000
Failure: Test: tgamma (-0x1.3ffffffffffffffep+4)
Result:
 is:         -1.68358626357708163562577078926814561e-15 
-0x1.e542bd5e58f06448fa82f80fa13cp-50
 should be:   2.36943678934055020753475621849318276e-01  
0x1.e542ba4020225131d4671fcf32c6p-3
 difference:  2.36943678934056704339739198930953903e-01  
0x1.e542ba4020261bb74f23d1affb58p-3
 ulp       :  9842255358231899528293245397236568.0000
 max.ulp   :  4.0000
Failure: Test: tgamma (-0x1.3fffffffffffffffffffffffff8p+4)
Result:
 is:         -1.68358626357706966420660912636119321e-15 
-0x1.e542bd5e58ec990c548409215999p-50
 should be:   1.04208932046406702025568537090748965e+12  
0x1.e542ba402022507a9cad2bf91e89p+39
 difference:  1.04208932046406702025568537259107594e+12  
0x1.e542ba402022507a9cad2cebbfe8p+39
 ulp       :  9842255358231829543284713326362600.0000
 max.ulp   :  4.0000
Failure: Test: tgamma (-0x1.3fffffffffffffffffffffffffffp+4)

Any advice would be appreciated.

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