ldexpf undefined in Python 2.4.3 build on SCO OSR 5.0.5

edcdave edcdave at gmail.com
Mon Jul 24 15:37:17 EDT 2006


My (gcc 2.95.2) build of Python 2.4.3 is failing with:

         gcc  -Wl,-Bexport -o python \
                        Modules/ccpython.o \
                        libpython2.4.a -lnsl -ldl    -lm
Undefined                       first referenced
 symbol                             in file
ldexpf                              /usr/ccs/lib/libm.so
python: fatal error: Symbol referencing errors. No output written to
python
collect2: ld returned 1 exit status
*** Error code 1 (bu21)
UX:make: ERROR: fatal error.

This is hardly surprising since config.log contained:

configure:18218: checking for hypot
configure:18269: gcc -o conftest -g -O2   conftest.c -lnsl -ldl  -lm
>&5
Undefined                       first referenced
 symbol                             in file
ldexpf                              /usr/ccs/lib/libm.so
conftest: fatal error: Symbol referencing errors. No output written to
conftest
collect2: ld returned 1 exit status
configure:18275: $? = 1

Which led to pyconfig.h containing:

/* Define to 1 if you have the `hypot' function. */
/* #undef HAVE_HYPOT */

So, even without HAVE_HYPOT, ldexpf is still referenced somehow.

Any suggestions? Should I step back to Python 2.3 source? Thanks,
Dave Harris

BTW, I'm the guy trying to get MoinMoin running on SCO OpemServer 5.0.5
(http://groups.google.com/group/comp.lang.python/browse_thread/thread/9321890929d5431d/fcc5a79144c195df?lnk=gst&q=getaddrinfo&rnum=2#fcc5a79144c195df).
The configure script correctly identifies getaddrinfo() as absent.

/* Define if you have the getaddrinfo function. */
/* #undef HAVE_GETADDRINFO */

Thanks for the suggestion to check it, Martin. There's still room for
some confusion since I'm trying to build Python 2.4.3 now and my
original binary was Python 2.3.4. But I'm satisfied that the Python
script is correct.




More information about the Python-list mailing list