glibc 2.2.3 make check errors on alphaev6-unknown-linux-gnu

["Ken Whaley" <ken at believe dot com>]

Are there known glibc 2.2.3 problems on alphaev6-unknown-linux-gnu
in math/ ?

I'm building glibc-2.2.3 with --enable-add-ons and
--kernel-version=2.2.14

I see consistent failures in math: test-fenv, test-float, and
test-ifloat across different binutil's and gcc's.   test-fenv runs
but has 35 failures (attached), and both test-float and test-ifloat
get floating point exceptions and dump core. Some of the math tests
get compile warnings that the equivalent build on i686 doesn't have
(see below) (all tests pass on i686 for me).

I get these failures building glibc-2.3.3 with binutils-2.10.90(alphapca56),
binutils-2.11(ev6), and binutils-2.11.90.0.15(ev6), using both
gcc-2.95.3+patches(atexit) (there are several more glibc-2.2.3 make check
failures on alphaev6 as well with gcc-2.95.3+patches, it doesn't look
like it can build a good glibc-2.2.3, as people here have warned],
and gcc-2.96-85 RH.

gcc
test-ifloat.c -c -O2 -Wall -Winline -Wstrict-prototypes -Wwrite-strings -g -
mieee  -U__LIBC_INTERNAL_MATH_INLINES -D__FAST_MATH__  -Wno-uninitialized  -
I../include -I. -I/usr/src/glibc-2.2.3/obj-alpha-bu21090-gcc29685-glc213/mat
h -I.. -I../libio  -I/usr/src/glibc-2.2.3/obj-alpha-bu21090-gcc29685-glc213 
-I../sysdeps/alpha/elf -I../linuxthreads/sysdeps/unix/sysv/linux -I../linuxt
hreads/sysdeps/pthread -I../linuxthreads/sysdeps/unix/sysv -I../linuxthreads
/sysdeps/unix -I../linuxthreads/sysdeps/alpha -I../sysdeps/unix/sysv/linux/a
lpha -I../sysdeps/unix/sysv/linux -I../sysdeps/gnu -I../sysdeps/unix/common 
-I../sysdeps/unix/mman -I../sysdeps/unix/inet -I../sysdeps/unix/sysv -I../sy
sdeps/unix/alpha -I../sysdeps/unix -I../sysdeps/posix -I../sysdeps/alpha/alp
haev6/fpu -I../sysdeps/alpha/alphaev6 -I../sysdeps/alpha/alphaev5 -I../sysde
ps/alpha/fpu -I../sysdeps/alpha -I../sysdeps/wordsize-64 -I../sysdeps/ieee75
4/flt-32 -I../sysdeps/ieee754/dbl-64 -I../sysdeps/ieee754 -I../sysdeps/gener
ic/elf -I../sysdeps/generic   -D_LIBC_REENTRANT -include
../include/libc-symbols.h     -D__NO_MATH_INLINES -D__LIBC_INTERNAL_MATH_INL
INES -DNO_LONG_DOUBLE -D_Mlong_double_=double -o
/usr/src/glibc-2.2.3/obj-alpha-bu21090-gcc29685-glc213/math/test-ifloat.o
In file included from ../include/bits/cmathcalls.h:1,
                 from ../math/complex.h:81,
                 from ../include/complex.h:1,
                 from
/usr/src/glibc-2.2.3/obj-alpha-bu21090-gcc29685-glc213/math/libm-test.c:114,
                 from test-ifloat.c:35:
../math/bits/cmathcalls.h: In function `cimagf':
../math/bits/cmathcalls.h:140: warning: function using short complex types
cannot be inline
../math/bits/cmathcalls.h: In function `crealf':
../math/bits/cmathcalls.h:147: warning: function using short complex types
cannot be inline
../math/bits/cmathcalls.h: In function `conjf':
../math/bits/cmathcalls.h:154: warning: function using short complex types
cannot be inline
In file included from test-ifloat.c:35:
/usr/src/glibc-2.2.3/obj-alpha-bu21090-gcc29685-glc213/math/libm-test.c: In
function `cimag_test':
../math/bits/cmathcalls.h:140: warning: can't inline call to `cimagf'
/usr/src/glibc-2.2.3/obj-alpha-bu21090-gcc29685-glc213/math/libm-test.c:1683
: warning: called from here
/usr/src/glibc-2.2.3/obj-alpha-bu21090-gcc29685-glc213/math/libm-test.c: In
function `conj_test':
../math/bits/cmathcalls.h:154: warning: can't inline call to `conjf'
/usr/src/glibc-2.2.3/obj-alpha-bu21090-gcc29685-glc213/math/libm-test.c:1828
: warning: called from here
/usr/src/glibc-2.2.3/obj-alpha-bu21090-gcc29685-glc213/math/libm-test.c: In
function `creal_test':
../math/bits/cmathcalls.h:147: warning: can't inline call to `crealf'
/usr/src/glibc-2.2.3/obj-alpha-bu21090-gcc29685-glc213/math/libm-test.c:1967
: warning: called from here


Here's a gdb look at the test-float failure:
GNU gdb 4.18
Copyright 1998 Free Software Foundation, Inc.
GDB is free software, covered by the GNU General Public License, and you are
welcome to change it and/or distribute copies of it under certain
conditions.
Type "show copying" to see the conditions.
There is absolutely no warranty for GDB.  Type "show warranty" for details.
This GDB was configured as "alpha-redhat-linux"...

warning: core file may not match specified executable file.
Core was generated by
`/usr/src/glibc-2.2.3/obj-alpha-bu21090-gcc29685-glc213/elf/ld-linux.so.2 --
libr'.
Program terminated with signal 8, Floating point exception.
Reading symbols from
/usr/src/glibc-2.2.3/obj-alpha-bu21090-gcc29685-glc213/math/libm.so.6.1...do
ne.
Reading symbols from
/usr/src/glibc-2.2.3/obj-alpha-bu21090-gcc29685-glc213/libc.so.6.1...done.
Reading symbols from
/usr/src/glibc-2.2.3/obj-alpha-bu21090-gcc29685-glc213/elf/ld-linux.so.2...d
one.
#0  __kernel_tanf (x=2.18555272e-08, y=0, iy=-1)
    at ../sysdeps/ieee754/flt-32/k_tanf.c:80
80              w = x+r;
(gdb) where
#0  __kernel_tanf (x=2.18555272e-08, y=0, iy=-1)
    at ../sysdeps/ieee754/flt-32/k_tanf.c:80
#1  0x2000003d33c in __tanf (x=2.18555272e-08)
    at ../sysdeps/ieee754/flt-32/s_tanf.c:45
#2  0x120024a64 in tan_test ()
    at
/usr/src/glibc-2.2.3/obj-alpha-bu21090-gcc29685-glc213/math/libm-test.c:3994
#3  0x1200266d8 in main (argc=1, argv=0xffffffffffffffff)
    at
/usr/src/glibc-2.2.3/obj-alpha-bu21090-gcc29685-glc213/math/libm-test.c:4394
(gdb) p x
$1 = 2.18555272e-08
(gdb) p r
Erroneous arithmetic operation.
(gdb)

test-fenv.out