Re: libtool 2.2.6

[Ralf Wildenhues]

Hello Dann,

thanks for the bug report.

* Donn Washburn wrote on Sun, Aug 30, 2009 at 12:05:11PM CEST:
> From a sorce compile from SuSE 11.2 Milestone 6 - gcc
> gcc (SUSE Linux) 4.4.1 [gcc-4_4-branch revision 149935]

>    4: libtoolize.at:201  libtoolize config files serial update
>       libtoolize

> --- experr    2009-08-30 04:20:34.000000000 -0500
> +++ /media/sdc1/archive/l/libtool-2.2.6/tests/testsuite.dir/at-stderr 
> 2009-08-30 04:20:34.000000000 -0500
> @@ -1,2 +1,3 @@
> +libtoolize: serial numbers `2009.04.28.21; # UTC' or `2009.04.28.21; # UTC' 
> contain non-digit chars

> 4. libtoolize.at:201: 4. libtoolize config files serial update 
> (libtoolize.at:201): FAILED (libtoolize.at:242)

Yeah, Automake changed the serial number format in the scripts and we
only fixed libtoolize post-2.2.6.  This is fixed in the git tree, but
mostly harmless for you (as the newer serial number format means you
do have a new-enough install-sh and config.{guess,sub} scripts anyway.

> #                             -*- compilation -*-
> 26. fail.at:27: testing ...
> ./fail.at:49: if $LIBTOOL --mode=compile $CC $CPPFLAGS $CFLAGS -c a.c; then 
> (exit 1); else :; fi
> stderr:
> a.c:1: error: expected '=', ',', ';', 'asm' or '__attribute__' before 'me'
> stdout:
> libtool: compile:  gcc -DHAVE_MMX -DHAVE_MMX2 -DHAVE_3DNOW -march=athlon-xp 
> -mfpmath=sse -m32 -c a.c  -fPIC -DPIC -o .libs/a.o
> ./fail.at:50: test -f a.lo
> ./fail.at:60: if $LIBTOOL --mode=compile $CC $CPPFLAGS $CFLAGS -no-suppress 
> -c a.c; then (exit 1); else :; fi
> stderr:
> stdout:
> libtool: compile:  gcc -DHAVE_MMX -DHAVE_MMX2 -DHAVE_3DNOW -march=athlon-xp 
> -mfpmath=sse -m32 -c a.c  -fPIC -DPIC -o .libs/a.o
> libtool: compile:  gcc -DHAVE_MMX -DHAVE_MMX2 -DHAVE_3DNOW -march=athlon-xp 
> -mfpmath=sse -m32 -c a.c  -fPIC -DPIC -o a.o
> ./fail.at:60: exit code was 1, expected 0
> 26. fail.at:27: 26. Failure tests (fail.at:27): FAILED (fail.at:60)

Ouch, -fPIC -DPIC is used for the non-PIC object a.o because of
--with-pic that you passed.  I'm checking in the patch below to
avoid this failure, and also a similar one when -prefer-pic is
passed in $CFLAGS.  Also, adding you to THANKS.

> #                             -*- compilation -*-
> 73. cmdline_wrap.at:28: testing ...

This is a followup failure of the failure of test 26.

Cheers,
Ralf

    Fix failure test in the presence of --with-pic or -prefer-pic.
    
    * tests/fail.at (Failure tests): Run non-PIC failure test only
    if pic_mode is 'default' or 'no', rather than 'default' or 'yes'
    and also -prefer-pic has not been passed in $CFLAGS.
    * THANKS: Update.
    Report by Donn Washburn.

diff --git a/tests/fail.at b/tests/fail.at
index fb05f8c..cec954b 100644
--- a/tests/fail.at
+++ b/tests/fail.at
@@ -50,15 +50,20 @@ FAIL_CHECK([$LIBTOOL --mode=compile $CC $CPPFLAGS $CFLAGS 
-c a.c])
 AT_CHECK([test -f a.lo], [1])
 
 # non-PIC compile failure
-case $pic_mode in default | yes)
-  case $build_old_libs,$pic_flag in yes,*-DPIC*)
-    AT_DATA([a.c], [[
+case $pic_mode in default | no)
+  case " $CFLAGS " in
+   *\ -prefer-pic\ *) ;;
+   *)
+    case $build_old_libs,$pic_flag in yes,*-DPIC*)
+      AT_DATA([a.c], [[
 #ifndef PIC
   choke me
 #endif
 ]])
-    FAIL_CHECK([$LIBTOOL --mode=compile $CC $CPPFLAGS $CFLAGS -no-suppress -c 
a.c])
-    AT_CHECK([test -f a.lo], [1])
+      FAIL_CHECK([$LIBTOOL --mode=compile $CC $CPPFLAGS $CFLAGS -no-suppress 
-c a.c])
+      AT_CHECK([test -f a.lo], [1])
+      ;;
+    esac
     ;;
   esac
   ;;