Message ID | AANLkTimQXgVE+yoBEpmUvn3gOP0eq4ge7YtKEYj5piO3@mail.gmail.com |
---|---|
State | New |
Headers | show |
On 10/18/2010 02:19 PM, Kai Tietz wrote: > As msvcrt's implementation of bessel-functions have a different > behavior in corner-cases then glibc's implementation, the test is for > now better to set to xfail. > > 2010-10-18 Kai Tietz > * gfortran.dg/bessel_7.f90: Set xfail for mingw targets. OK. (If I recall correctly, with MS's BESSEL_YN(n, 0.0) gives NAN rather than -INF for x=0.0 and all "n".) Tobias
2010/10/18 Tobias Burnus <burnus@net-b.de>: > On 10/18/2010 02:19 PM, Kai Tietz wrote: >> >> As msvcrt's implementation of bessel-functions have a different >> behavior in corner-cases then glibc's implementation, the test is for >> now better to set to xfail. >> >> 2010-10-18 Kai Tietz >> * gfortran.dg/bessel_7.f90: Set xfail for mingw targets. > > OK. > > (If I recall correctly, with MS's BESSEL_YN(n, 0.0) gives NAN rather than > -INF for x=0.0 and all "n".) Right. Applied at revision 165633. Cheers, Kai
Index: bessel_7.f90 =================================================================== --- bessel_7.f90 (revision 165609) +++ bessel_7.f90 (working copy) @@ -1,8 +1,11 @@ -! { dg-do run } +! { dg-do run { xfail *-*-mingw* } } ! ! PR fortran/36158 ! PR fortran/33197 ! +! For mingw targets this test is disabled as the MS implementation +! of Bessel has different results. +! ! Run-time tests for transformations BESSEL_YN ! implicit none