Re: Errors with rpl_realloc(), pthreads calls on hppa2.0w-hp-hpux11.11

[Daniel Richard G.]

On Sat, 2013 Oct 12 11:52-0700, Paul Eggert wrote:
> 
> Thanks, can you look at the generated 'configure' and see why that is?
> It should contain something like this:
> 
>   if test $ac_cv_func_realloc_0_nonnull = yes; then :
>     $as_echo "#define HAVE_REALLOC_GNU 1" >>confdefs.h
>   else
>     $as_echo "#define HAVE_REALLOC_GNU 0" >>confdefs.h
>     REPLACE_REALLOC=1
>   fi
>   if test $REPLACE_REALLOC = 1; then
>     gl_LIBOBJS="$gl_LIBOBJS realloc.$ac_objext"
>   fi
> 
> You're reporting that HAVE_REALLOC_GNU=0, which means REPLACE_REALLOC
> should be 1 and gl_LIBOBJS should contain realloc.o; evidently this is
> not taking place.  Perhaps you can put some 'echo' statements into
> 'configure' to see what's going wrong.

I looked through the configure script. It appears that realloc.o is
added to gl_LIBOBJS if a non-POSIX realloc() is found. However, if
realloc() is non-GNU, realloc.o is added to gltests_LIBOBJS, not
gl_LIBOBJS.

    $ grep realloc configure.out 
    checking whether malloc, realloc, calloc are POSIX compliant... yes
    checking for GNU libc compatible realloc... no

    $ fgrep realloc.o gllib/Makefile 
    gltests_LIBOBJS =  putenv.o realloc.o usleep.o


> > This link needs to be done with -mt to pull in the pthreads library.
>
> Wouldn't -lpthread work as well?  pthreadlib.m4 says it would.

To a degree, yes, but there's also some preprocessor symbols that should
be defined. From the cc(1) man page:

      -mt            Sets various -D flags to enable multi-threading and
                     also sets -lpthread.  For details see HP C Online Help.

> I pushed this patch to try to fix that:

That did address the link error, but -mt would be the better way.


--Danny


-- 
Daniel Richard G. || address@hidden
My ASCII-art .sig got a bad case of Times New Roman.