darwin clang linkage fix

[Jack Howarth]

On Tue, Apr 05, 2011 at 06:53:41PM -0400, Jack Howarth wrote:
>    The change...
> 
> --- acinclude.m4.orig	2011-04-05 18:29:17.000000000 -0400
> +++ acinclude.m4	2011-04-05 18:30:04.000000000 -0400
> @@ -1941,7 +1941,7 @@
>  esac
>  
>  cat >conftest.c <<EOF
> -const int foo = 123;
> +const int foo[] = { 1,2,3 };
>  EOF
>  echo "Test program:" >&AC_FD_CC
>  cat conftest.c >&AC_FD_CC
> 
> ,with regeneration of configure to utilitze it, is sufficient to solve the linkage
> problems with clang from Xcode 4.0.1 under darwin10. This causes config.m4 in the 
> build directory to contain...
> 
> define(<RODATA>, <      .section        __TEXT,__const>)
> 
> instead of the offending...
> 
> define(<RODATA>, <      .section        __TEXT,__literal4,4byte_literals>)
> 
> The resulting build fails the make check as...
> 
> libtool: link: clang -O2 -pedantic -m64 -mtune=core2 -march=core2 -o .libs/t-sub t-sub.o  ./.libs/libtests.a /sw/src/gmp-5.0.1/build/.libs/libgmp.dylib ../.libs/libgmp.dylib
> make  check-TESTS
> PASS: t-bswap
> PP_INVERTED == 21cfe6cfc938b36b, but pp_inverted_calc == 2255c453a87eca4f
> 
> After gmp.h,
>   ULONG_MAX  not defined
>   LONG_MAX   not defined
>   UINT_MAX   not defined
>   INT_MAX    not defined
>   USHRT_MAX  not defined
>   SHRT_MAX   not defined
> 
> ULONG_MAX      FFFFFFFFFFFFFFFF
> ULONG_HIGHBIT  8000000000000000
> LONG_MAX       7FFFFFFFFFFFFFFF
> LONG_MIN       8000000000000000
> UINT_MAX       FFFFFFFF
> UINT_HIGHBIT   80000000
> INT_MAX        7FFFFFFF
> INT_MIN        80000000
> USHRT_MAX      FFFF
> USHRT_HIGHBIT  8000
> SHRT_MAX       7FFF
> SHRT_MIN       8000
> 
> Bits
>   long           64
>   int            32
>   short          16
>   unsigned long  64
>   unsigned int   32
>   unsigned short 16
>   mp_size_t      64
> /bin/sh: line 1: 28615 Abort trap              ${dir}$tst
> FAIL: t-constants
> PASS: t-count_zeros
> PASS: t-gmpmax
> PASS: t-hightomask
> PASS: t-modlinv
> PASS: t-popc
> PASS: t-parity
> PASS: t-sub
> ====================================
> 1 of 9 tests failed
> Please report to gmp-bugs at gmplib.org
> ====================================
> make[4]: *** [check-TESTS] Error 1
> make[3]: *** [check-am] Error 2
> make[2]: *** [check-recursive] Error 1
> make[1]: *** [check-recursive] Error 1
> 
> but I believe this may be due to the remaining warnings...
> 
> /bin/sh ../libtool --mode=compile --tag=CC ../../mpn/m4-ccas --m4="m4" clang -c -DHAVE_CONFIG_H -I. -I../../mpn -I.. -D__GMP_WITHIN_GMP -I../.. -DOPERATION_`echo com | sed 's/_$//'`    -O2 -pedantic -m64 -mtune=core2 -march=core2  `test -f 'com.asm' || echo '../../mpn/'`com.asm
> /tmp/cc-611GcV.s:10:2: warning: ignoring directive for now
>  .type ___gmpn_add_nc, at function
>  ^
> /tmp/cc-611GcV.s:14:2: warning: ignoring directive for now
>  .size ___gmpn_add_nc,.-___gmpn_add_nc
>  ^
> /tmp/cc-611GcV.s:17:2: warning: ignoring directive for now
>  .type ___gmpn_add_n, at function
>  ^
> /tmp/cc-i1zwl9.s:10:2: warning: ignoring directive for now
>  .type ___gmpn_sub_nc, at function
>  ^
> /tmp/cc-611GcV.s:86:2: warning: ignoring directive for now
>  .size ___gmpn_add_n,.-___gmpn_add_n
>  ^
> /tmp/cc-i1zwl9.s:14:2: warning: ignoring directive for now
>  .size ___gmpn_sub_nc,.-___gmpn_sub_nc
>  ^
> /tmp/cc-i1zwl9.s:17:2: warning: ignoring directive for now
>  .type ___gmpn_sub_n, at function
>  ^
> /tmp/cc-i1zwl9.s:86:2: warning: ignoring directive for now
>  .size ___gmpn_sub_n,.-___gmpn_sub_n
>  ^
> 

Comparing the config.m4 files for an Apple gcc-4.2 and clang build
shows...

--- build/config.m4	2011-04-05 19:02:31.000000000 -0400
+++ build.clang/config.m4	2011-04-05 18:36:44.000000000 -0400
@@ -11,9 +11,9 @@
 define(<GLOBL>, <.globl>)
 define(<GLOBL_ATTR>, <>)
 define(<GSYM_PREFIX>, <_>)
-define(<RODATA>, <.text>)
-define(<TYPE>, <>)
-define(<SIZE>, <>)
+define(<RODATA>, <	.section	__TEXT,__const>)
+define(<TYPE>, <.type	$1,@$2>)
+define(<SIZE>, <.size	$1,$2>)
 define(<LSYM_PREFIX>, <L>)
 define(<W32>, <.long>)
 define(<ALIGN_LOGARITHMIC>,<yes>)

It appears that the current acinclude.m4 confuses the warning...

configure:26474: checking for assembler .type directive
configure:26485: clang -c -O2 -pedantic -m64 -mtune=core2 -march=core2  conftest.s >conftest.out 2>&1
configure:26488: $? = 0
/tmp/cc-68DYYR.s:5:2: warning: ignoring directive for now
 .type sym, at function
 ^
configure:26508: result: .type  $1,@$2
configure:26512: checking for assembler .size directive
configure:26522: clang -c -O2 -pedantic -m64 -mtune=core2 -march=core2  conftest.s >conftest.out 2>&1
configure:26525: $? = 0
/tmp/cc-naVqNU.s:5:2: warning: ignoring directive for now
 .size sym,1
 ^
configure:26542: result: .size  $1,$2

in config.log as actual support for these features. Hopefully if we fix this issue,
the make check results may improve.
              Jack

> throughout the build. I suspect this is a known issue with darwin clang
> but will check upstream.
>          Jack
> _______________________________________________
> gmp-bugs mailing list
> gmp-bugs at gmplib.org
> https://gmplib.org/mailman/listinfo/gmp-bugs