Bugzilla Bug 158040

When building sys-apps/attr on ROOT!=SYSROOT for a final rootfs, it fails with
this error:

./configure --prefix=/usr --host=arm-carel-linux-gnu --mandir=/usr/share/man
--infodir=/usr/share/info --datadir=/usr/share --sysconfdir=/etc
--localstatedir=/var/lib --disable-gettext --libexecdir=/usr/lib --bindir=/bin
--build=i686-pc-linux-gnu
checking for arm-carel-linux-gnu-gcc... arm-carel-linux-gnu-gcc
checking for C compiler default output... a.out
checking whether the C compiler works... yes
checking whether we are cross compiling... yes
checking for suffix of executables...
checking for suffix of object files... o
checking whether we are using the GNU C compiler... yes
checking whether arm-carel-linux-gnu-gcc accepts -g... yes
checking for arm-carel-linux-gnu-gcc option to accept ANSI C... none needed
checking for gmake... /usr/bin/gmake
checking for glibtool... no
checking for libtool... /usr/bin/libtool
checking for tar... /bin/tar
checking for gzip... /bin/gzip
checking for makedepend... /bin/true
checking for awk... /bin/awk
checking for sed... /bin/sed
checking for echo... /bin/echo
checking for sort... /bin/sort
checking whether ln -s works... yes
checking for rpm... no
checking for an ANSI C-conforming const... yes
checking how to run the C preprocessor... arm-carel-linux-gnu-gcc -E
checking for egrep... grep -E
checking for ANSI C header files... yes
checking for sys/types.h... yes
checking for sys/stat.h... yes
checking for stdlib.h... yes
checking for string.h... yes
checking for memory.h... yes
checking for strings.h... yes
checking for inttypes.h... yes
checking for stdint.h... yes
checking for unistd.h... yes
checking for mode_t... yes
checking for working alloca.h... yes
checking for alloca... yes
configure: creating ./config.status
config.status: creating include/builddefs
config.status: creating include/config.h
=== include ===
gmake[1]: warning: jobserver unavailable: using -j1.  Add `+' to parent make
rule.
rm -f attr
ln -s . attr
=== libmisc ===
gmake[1]: warning: jobserver unavailable: using -j1.  Add `+' to parent make
rule.
/usr/bin/libtool --mode=compile arm-carel-linux-gnu-gcc -O -ggdb -fno-inline -O
-ggdb -fno-inline -DNDEBUG -funsigned-char -fno-strict-aliasing -Wall
-DVERSION=\"2.4.28\" -DLOCALEDIR=\"/usr/share/locale\" -DPACKAGE=\"attr\"
-I./include -D_GNU_SOURCE -D_FILE_OFFSET_BITS=64  -O -ggdb -fno-inline -DNDEBUG
-funsigned-char -fno-strict-aliasing -Wall -DVERSION=\"2.4.28\"
-DLOCALEDIR=\"/usr/share/locale\" -DPACKAGE=\"attr\" -I../include -D_GNU_SOURCE
-D_FILE_OFFSET_BITS=64   -c quote.c
libtool: compile: unable to infer tagged configuration
libtool: compile: specify a tag with `--tag'
gmake[1]: *** [quote.lo] Error 1
make: *** [default] Error 2

the problem is that the libtool used it's the system's one, that is not
tailored for the correct CHOST.

I've worked this around by emerging libtool in SYSROOT with -O (to avoid it
taking up autoconf as runtime dependency, that in turn requires perl, that is
not crosscompilable), and then setting
LIBTOOL=/usr/arm-carel-linux-gnu/usr/bin/libtool .

I think crossdev should consider libtool as an extra stage that might be
needed.

Another problem is that acl fails to build against libattr even with the
libtool hack:

=== getfacl ===
gmake[1]: warning: jobserver unavailable: using -j1.  Add `+' to parent make
rule.
arm-carel-linux-gnu-gcc -O -ggdb -fno-inline -O -ggdb -fno-inline -DNDEBUG
-funsigned-char -fno-strict-aliasing -Wall -DVERSION=\"2.2.34\"
-DLOCALEDIR=\"/usr/share/locale\" -DPACKAGE=\"acl\" -I./include -D_GNU_SOURCE
-D_FILE_OFFSET_BITS=64  -O -ggdb -fno-inline -DNDEBUG -funsigned-char
-fno-strict-aliasing -Wall -DVERSION=\"2.2.34\"
-DLOCALEDIR=\"/usr/share/locale\" -DPACKAGE=\"acl\" -I../include -D_GNU_SOURCE
-D_FILE_OFFSET_BITS=64    -c -o getfacl.o getfacl.c
arm-carel-linux-gnu-gcc -O -ggdb -fno-inline -O -ggdb -fno-inline -DNDEBUG
-funsigned-char -fno-strict-aliasing -Wall -DVERSION=\"2.2.34\"
-DLOCALEDIR=\"/usr/share/locale\" -DPACKAGE=\"acl\" -I./include -D_GNU_SOURCE
-D_FILE_OFFSET_BITS=64  -O -ggdb -fno-inline -DNDEBUG -funsigned-char
-fno-strict-aliasing -Wall -DVERSION=\"2.2.34\"
-DLOCALEDIR=\"/usr/share/locale\" -DPACKAGE=\"acl\" -I../include -D_GNU_SOURCE
-D_FILE_OFFSET_BITS=64    -c -o user_group.o user_group.c
/usr/arm-carel-linux-gnu/usr/bin/libtool --mode=link arm-carel-linux-gnu-gcc -o
getfacl    getfacl.o user_group.o   ../libacl/libacl.la /usr/lib/libattr.la
../libmisc/libmisc.la
mkdir .libs
arm-carel-linux-gnu-gcc -o .libs/getfacl getfacl.o user_group.o 
../libacl/.libs/libacl.so -lattr /usr/lib/libattr.so ../libmisc/.libs/libmisc.a
/usr/libexec/gcc/arm-carel-linux-gnu/ld: skipping incompatible /lib/libattr.so
when searching for /lib/libattr.so
/usr/libexec/gcc/arm-carel-linux-gnu/ld: cannot find /lib/libattr.so
collect2: ld returned 1 exit status
gmake[1]: *** [getfacl] Error 1
make: *** [default] Error 2

!!! ERROR: sys-apps/acl-2.2.34 failed.
Call stack:
  ebuild.sh, line 1546:   Called dyn_compile
  ebuild.sh, line 937:   Called src_compile
  acl-2.2.34.ebuild, line 42:   Called die

!!! (no error message)
!!! If you need support, post the topmost build error, and the call stack if
relevant.

I'll try to modify crossdev tonight, and see what happens.

Created an attachment (id=103976) [details]
Ebuild patch

This patch replace AC_PATH_PROG calls with AC_PATH_TOOL, so that it will use
${CHOST}-libtool if present (see bug #158067), and allows attr to be
crosscompiled easily.

I'm going to attach a patch and an ebuild patch that solves the issue for me (I
sent the patch upstream, waiting to see what they think of it).

Created an attachment (id=104781) [details]
Ebuild patch

Created an attachment (id=104782) [details]
attr-2.4.32-libtool.patch

this patch doesn't seem to be applied in the ebuild, cross compiling is still
nerfed..

Any word on this? Please please please put it in there, if it will help
crosscompiling to work, period. I can't build anything for i686 (on my amd64
host), as everything that tries to be compiled starts at the first step, with
attr and the infamous bug:

libtool: compile: unable to infer tagged configuration
libtool: compile: specify a tag with `--tag'

What needs to happen to get this committed ASAP? It's been long enough. :)

*** Bug 191363 has been marked as a duplicate of this bug. ***

for me the patch applied from Diego Pettenò works, thanks for that. 
But again, problem appears for acl.

Isn't it better to put this problem more in general terms?
I mean Is there some other bug open?

(In reply to comment #8)
> 
> Isn't it better to put this problem more in general terms?
> I mean Is there some other bug open?
> 

See bug 158068 for ACL patch.

attr-2.4.39 should handle this