Re: findutils-4.2.29 check failure (iregex1)

[hv]

"James Youngman" <address@hidden> wrote:
:On 2/15/07, address@hidden <address@hidden> wrote:
:> [...]
:> FAIL: iregex1, --- find.out     2007-02-15 03:14:22.000000000 +0000
:> +++ cmp.out     2007-02-15 03:14:22.000000000 +0000
:> @@ -1,2 +1,4 @@
:>  tmp
:>  tmp/d
:> +tmp/d/D
:> +tmp/d/D/d
:> child process exited abnormally
:> [...]
:> # of expected passes            63
:> # of unexpected failures        1
:> /src/package/util/findutils-4.2.29/find/testsuite/../find version 4.2.29
:> Features enabled: D_TYPE O_NOFOLLOW(enabled) LEAF_OPTIMISATION
:>
:> This was built on a system originating as RedHat-7.1 which has been
:> progressively bastardised by piecemeal handbuilt upgrades since the
:> RPM database was lost.
:>
:> I'm happy to provide system information, or attempt to debug it (if
:> someone can turn the find_start call in iregex1.exp into an actual
:> command line for me).
:
:$ rm -rf tmp;  mkdir tmp tmp/d tmp/d/D tmp/d/D/d tmp/d/D/d/e; find tmp
:-iregex 'tmp\(/d\)*' -print
:tmp
:tmp/d
:tmp/d/D
:tmp/d/D/d

Ok, I think the problem here is either the gnulib/lib/Makefile* or an
interaction between them and configure: after starting again from a freshly
unpacked tarball and issuing:
  ./configure 2>&1 | tee configure.log
  make 2>&1 | tee make.log
I can see:

% grep re_compile configure.log 
checking for working re_compile_pattern... no
checking for working re_compile_pattern... (cached) no
% ldd find/find
        libc.so.6 => /lib/libc.so.6 (0x40024000)
        /lib/ld-linux.so.2 => /lib/ld-linux.so.2 (0x40000000)
% nm -a find/find | grep re_compile_pattern
         U re_compile_pattern@@GLIBC_2.0
% perl -nle 'print if /^(libgnulib_a_DEPENDENCIES|am_libgnulib_a_OBJECTS)/ .. 
!/\\\s*$/' gnulib/lib/Makefile
libgnulib_a_DEPENDENCIES =  ${LIBOBJDIR}mktime$U.o ${LIBOBJDIR}strftime$U.o 
${LIBOBJDIR}canonicalize$U.o ${LIBOBJDIR}closeout$U.o ${LIBOBJDIR}dirname$U.o 
${LIBOBJDIR}exitfail$U.o ${LIBOBJDIR}filemode$U.o ${LIBOBJDIR}filenamecat$U.o 
${LIBOBJDIR}human$U.o ${LIBOBJDIR}idcache$U.o ${LIBOBJDIR}modechange$U.o 
${LIBOBJDIR}mountlist$U.o ${LIBOBJDIR}quote$U.o ${LIBOBJDIR}quotearg$U.o 
${LIBOBJDIR}savedir$U.o ${LIBOBJDIR}fopen-safer$U.o ${LIBOBJDIR}dup-safer$U.o 
${LIBOBJDIR}fd-safer$U.o ${LIBOBJDIR}xmalloc$U.o ${LIBOBJDIR}xstrtol$U.o 
${LIBOBJDIR}xstrtoul$U.o ${LIBOBJDIR}yesno$U.o 
am_libgnulib_a_OBJECTS = argmatch.$(OBJEXT) cycle-check.$(OBJEXT) \
        basename.$(OBJEXT) stripslash.$(OBJEXT) xalloc-die.$(OBJEXT) \
        xgetcwd.$(OBJEXT) xreadlink.$(OBJEXT) xstrtoumax.$(OBJEXT)
% 

.. and indeed, make.log shows that these are the only files compiled in
gnulib/lib after the 'make all-am' is invoked.

I assume the intention is that if the re_compile_pattern probe fails, the
regex support in gnulib/lib should be compiled and linked in instead of
allowing it to be provided by the installed libraries. The fact that
doesn't happen would appear to be the proximate cause of the test failure.

I don't know much about debugging at this level, but I can't see any way
that the effect of setting "ac_use_included_regex" in configure actually
gets propagated anywhere, and 'grep -i regex config.status' matches nothing.
Suggestions for a next step would be appreciated.

Hugo