Robert Schwebel <[EMAIL PROTECTED]> writes: > > I noticed that after the release, when compiling with an gcc-2.x under > > Solaris. Appearantly mixing code and declarations (like in C++) is > > allowed by c99 and gcc-3.0. The solution is to move the declaration, > > as you notice. > > What does the ANSI-C standard say?
My understanding is that it's allowed by the brand new 1999 ANSI-C standard, but disallowed by the older 1989 standard, and by most existing C compilers. > My workaround is to add a test to argp-namefrob.h like this: > > #ifdef _LIBC > #ifndef weak_alias > > [...] > > #endif > #endif > > weak_alias ist an ELF feature, so the decision if it shall be used or not > is independend of LIBC. That sounds reasonable to me. > Next problem: HAVE_PROGRAM_INVOCATION_NAME is determined incorrectly, > because the test in configure.in searches for "program_invocation_name" in > errno.h. As uClibc uses the original glibc header files it would be better > to look at the library itself and search if the functions exist there. > I've changed it this way: > > # ARGP_CHECK_VAR(program_invocation_name, [#include <errno.h>]) > # ARGP_CHECK_VAR(program_invocation_short_name, [#include <errno.h>]) > AC_CHECK_FUNCS(program_invocation_name) > AC_CHECK_FUNCS(program_invocation_short_name) It doesn't seem right to use AC_CHECK_FUNC to check for symbols that aren't functions. It would be better to change ARGP_CHECK_VAR (defined in src/argp/acinclude.m4) to use AC_TRY_LINK. Ah, but it already does! So there's something wrong with that test. Please check config.log. > Now these errors remain: > > gcc -g -O2 -ggdb3 -Wall -W -Wmissing-prototypes -Wmissing-declarations > -Wstrict-prototypes -Waggregate-return -Wpointer-arith > -Wbad-function-cast -Wnested-externs -o argp-test argp-test.o > libargp.a > libargp.a(argp-parse.o): In function `argp_default_parser': > /home/robert/embedded/lsh-1.2.5/src/argp/argp-parse.c:144: undefined > reference to `__sleep' > Is there a reason why the __ versions are used here? This does normaly > only make sense if you want to redefine a function from libc and overload > it with your own substiture. Before I understand what I'm doing here ;-) > I've removed the underscores, with the result that everything in argp > compiles fine now and the test works. There should probably be an #ifndef _LIBC #define __sleep sleep #endif in the file that uses sleep. I think some other functions are handled in that way, in argp-help.c. > The next problem is the size test for [int/short/long]. At the moment the > configure script guesses incorrectly - all sizes are set to 0. I have no > idea why... To find out, you have to look in config.log. > > configure checks for a definition of gethostbyname in libnsl, because > > that's needded on Solaris. I've never seen it on any linux or glibc > > system. > > robert@callisto:~/embedded/glibc-2.2.4/nis ! ls *nsl* > libnsl.a libnsl.so libnsl.so.1@ libnsl_g.a libnsl_pic.a You still have to figure out _whether or not you're supposed to use that library_. If I have understood your situation correctly, lsh's configure scripts figures that you want to use -lnsl and adds it to LDFLAGS. And then you get some link errors on conflicts between that library and glibc. In order to fix the configure script, I need to know what the right thing is. To me, it sounds like there's something fishy with your glibc installation. Regards, /Niels
