This is the mail archive of the libc-hacker@cygnus.com mailing list for the glibc project.

Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]

Re: more glibc 2.0.109 fails on linuxppc

On Thu, 31 Dec 1998 16:13:27 -0500, "Jack Howarth" wrote:
>Zack,
>   Here is an example of a segmentation fault in the 2.0.109 libc.log...
>
>egcs -c t-coverage/efgcvt.c -o /usr/src/redhat/BUILD/glibc-2.0.109/build-ppc-l
>inux/testsuite/efgcvt.o
>ld /usr/src/redhat/BUILD/glibc-2.0.109/build-ppc-linux/testsuite/efgcvt.o -o /
>usr/src/redhat/BUILD/glibc-2.0.109/build-ppc-linux/testsuite/efgcvt.x
>PASS: efgcvt.cc
>./efgcvt.x 
>ERROR: child 14022: Segmentation fault
>UNTESTED: efgcvt.1
>UNTESTED: efgcvt.2
>UNTESTED: efgcvt.3
>
>...whereas in the same test in glibc 2.0.108 gives...
>
>gcs -c t-coverage/efgcvt.c -o /usr/src/redhat/BUILD/glibc-2.0.108/build-ppc-li
>nux/testsuite/efgcvt.o
>ld /usr/src/redhat/BUILD/glibc-2.0.108/build-ppc-linux/testsuite/efgcvt.o -o /
>usr/src/redhat/BUILD/glibc-2.0.108/build-ppc-linux/testsuite/efgcvt.x
>PASS: efgcvt.cc
>./efgcvt.x 
>PASS: efgcvt.1
>PASS: efgcvt.2
>PASS: efgcvt.3
>
>so I think there really is a problem for us in 2.0.109 there.

Sorry, I still think this is the compiler.  The iconv code has exposed
compiler bugs on PPC in the past.

>cc1: warnings being treated as errors
>t-coverage/libm-test.inc: In function `check_equal':
>In file included from t-coverage/math-double.c:38:
>t-coverage/libm-test.inc:351: warning: implicit declaration of function `isinf
>l'FAIL: math-double.cc

It shouldn't be referring to isinfl when testing the double precision
functions.  What's on line 351 of libm-test.inc?

>t-coverage/libm-test.inc: In function `check_equal':
>In file included from t-coverage/math-ildouble.c:39:
>t-coverage/libm-test.inc:351: warning: implicit declaration of function `isinf
>l't-coverage/libm-test.inc:357: warning: type mismatch in implicit declaration
> for built-in function `fabsl'

This is normal - PPC has no long double arithmetic.  The error could be made
tidier.

>./printf-float.x 
>ERROR: child 14579: Segmentation fault
>UNTESTED: printf-float.1

Definitely the compiler.  printf hasn't changed in years.  (Hmm, maybe it's
the dynamic linker - but then nothing should've worked.)

zw
Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]