glibc 2.2.93 make check failure

["Steve Munroe" <sjmunroe at us dot ibm dot com>]

Any one else seeing the tfformat failure?

GCONV_PATH=/home/sjmunroe/work/build32-2.2.93/iconvdata LC_ALL=C
/home/sjmunroe/work/build32-2.2.93/elf/ld.so.1 --library-path
/home/sjmunroe/work/build32-2.2.93:/home/sjmunroe/work/build32-2.2.93/math:/home/sjmunroe/work/build32-2.2.93/elf:/home/sjmunroe/work/build32-2.2.93/dlfcn:/home/sjmunroe/work/build32-2.2.93/nss:/home/sjmunroe/work/build32-2.2.93/nis:/home/sjmunroe/work/build32-2.2.93/rt:/home/sjmunroe/work/build32-2.2.93/resolv:/home/sjmunroe/work/build32-2.2.93/crypt:/home/sjmunroe/work/build32-2.2.93/linuxthreads
 /home/sjmunroe/work/build32-2.2.93/stdio-common/tfformat  >
/home/sjmunroe/work/build32-2.2.93/stdio-common/tfformat.out
make[2]: *** [/home/sjmunroe/work/build32-2.2.93/stdio-common/tfformat.out]
Error 1
make[2]: Leaving directory `/u3/sjmunroe/work/glibc-2.2.93/stdio-common'
make[1]: *** [stdio-common/tests] Error 2
make[1]: Leaving directory `/u3/sjmunroe/work/glibc-2.2.93'
make: *** [check] Error 2

Error in line 4012 using "%.1a".  Result is "0x1.0p+1"; should be:
"0x1.0p+4".
Encountered 1 errors in 3996 tests.

I am seeing this on powerpc64 build so I pulled glibc-2.2.93 tar ball to
see what was different and saw the same failure.

binutils and gcc 3.2 where pulled form the cvs HEAD on 08/28/2002

gcc -v
Reading specs from
/opt/ppc32-20020828/lib/gcc-lib/powerpc-linux/3.2.1/specs
Configured with: /home/toolchain/gcc-3.2/gcc-20020823/configure
--build=powerpc-linux --host=powerpc-linux --target=powerpc-linux
--enable-languages=c --enable-__cxa_atexit --disable-nls
--prefix=/opt/ppc32-20020828
Thread model: posix
gcc version 3.2.1 20020823 (prerelease)

Running on a IBM 7026-B80
4xPower3 (630+) 375MHz