libnih version 1.0.3-4ubuntu1 failed to build on armel

[Ricardo Salveti]

libnih version 1.0.3-4ubuntu1 failed to build on armel
Link to failed build: https://launchpad.net/ubuntu/+source/libnih/1.0.3-4ubuntu1/+build/2523639

Direct link to the build log: https://launchpad.net/ubuntu/+source/libnih/1.0.3-4ubuntu1/+build/2523639/+files/buildlog_ubuntu-oneiric-armel.libnih_1.0.3-4ubuntu1_FAILEDTOBUILD.txt.gz

This log snippet might be of interest, since it triggered the matcher 'Purging chroot-autobuild'.
Excerpt 996 lines into the build log:

...with unhandled error beneath context
PASS: test_error
==================================================
14 of 17 tests failed
Please report to http://bugs.launchpad.net/libnih/
==================================================
make[3]: *** [check-TESTS] Error 1
make[3]: Leaving directory `/build/buildd/libnih-1.0.3/nih'
make[2]: *** [check-am] Error 2
make[2]: Leaving directory `/build/buildd/libnih-1.0.3/nih'
make[1]: *** [check-recursive] Error 1
make[1]: Leaving directory `/build/buildd/libnih-1.0.3'
dh_auto_test: make -j1 check returned exit code 2
make: *** [build] Error 29
dpkg-buildpackage: error: debian/rules build gave error exit status 2
******************************************************************************
Build finished at 20110601-0429
FAILED [dpkg-buildpackage died]
Purging chroot-autobuild/build/buildd/libnih-1.0.3

[Dr. David Alan Gilbert (davidgil-uk)]

When fixed please check if it cures bug 791294

[Dr. David Alan Gilbert (davidgil-uk)]

make[3]: Entering directory `/build/buildd/libnih-1.0.3/nih'
/bin/bash: line 5: 24330 Segmentation fault ${dir}$tst
FAIL: test_alloc
/bin/bash: line 5: 24335 Segmentation fault ${dir}$tst
FAIL: test_string
/bin/bash: line 5: 24340 Segmentation fault ${dir}$tst
FAIL: test_list
/bin/bash: line 5: 24345 Segmentation fault ${dir}$tst
FAIL: test_hash
/bin/bash: line 5: 24350 Segmentation fault ${dir}$tst
FAIL: test_tree
/bin/bash: line 5: 24355 Segmentation fault ${dir}$tst
FAIL: test_timer
/bin/bash: line 5: 24360 Segmentation fault ${dir}$tst
FAIL: test_signal
/bin/bash: line 5: 24365 Segmentation fault ${dir}$tst
FAIL: test_child
/bin/bash: line 5: 24370 Segmentation fault ${dir}$tst
FAIL: test_io
/bin/bash: line 5: 24375 Segmentation fault ${dir}$tst
FAIL: test_file

[Matthias Klose (doko)]

testing with gcc-4.6 4.6.0-12 from debian unstable. the package builds. using the same debian gcc with -march=armv7-a -mthumb -mfloat-fp=softfp -mfpu=vfpv3-d16, the package fails in the test.

looks like an armv7/thumb regression in 4.6 compared to 4.5.

[Launchpad Janitor (janitor)]

This bug was fixed in the package libnih - 1.0.3-4ubuntu2

---------------
libnih (1.0.3-4ubuntu2) oneiric; urgency=low

  * Use dpkg-buildflags to get the build flags.
  * Build with the default build flags, don't hard-code -Os. LP: #791315.
 -- Matthias Klose <email address hidden> Wed, 15 Jun 2011 16:45:42 +0200

[Matthias Klose (doko)]

-O2 works on oneiric, -Os fails

[Michael Hope (michaelh1)]

Interesting. I can reproduce this with a libnih-1.0.3-4ubuntu2 with the -Os hacked back in, a natty chroot, and gcc-linaro-4.6-2011.06. I can't reproduce it using a maverick chroot. test_alloc faults before getting to _start which suggests a linker problem.

Running with LD_DEBUG=all /lib/ld-linux.so.3 ./test_alloc shows a segfault while running the init in librt.

[Michael Hope (michaelh1)]

Stubbing out librt doesn't fix the problem.

The problem seems to be outside the compiler. Using the same host, gcc-linaro-4.6-2011.06, and source package I see the fault when built on natty and run on natty or maverick; and don't see the fault if built on maverick and run on natty or maverick.

[Michael Hope (michaelh1)]

Richard, I've built libnih twice under ~michaelh/bugs/lp791315/libnih-{natty,maverick}

Both use gcc-linaro-4.6-2011.06-0 at -Os. The natty one uses binutils 2.21.0.20110327-2ubuntu3. The maverick one uses 2.20.51.20100908-0ubuntu2.

nih/test_alloc segfaults with the natty binutils and passes with the maverick one.

The maverick binutils is in ~michaelh/linaro/toolchains/binutils-maverick. Put $prefix/usr/bin in your path and $prefix/usr/lib in your LD_LIBRARY_PATH.

[Richard Sandiford (rsandifo)]

This is due to a bug in libc's start.S, which causes _start to be misaligned:

    http://sourceware.org/ml/libc-ports/2011-06/msg00023.html

[Michael Hope (michaelh1)]

Sorry, accidently changed the status on gcc as well...

[Launchpad Janitor (janitor)]

This bug was fixed in the package eglibc - 2.13-8ubuntu1

---------------
eglibc (2.13-8ubuntu1) oneiric; urgency=low

  [ Colin Watson ]
  * Backport from upstream:
    - Update UTF-8 charmap from recent Unidata.txt file.
    - Transliterate U20B9 (LP: #799673).

  [ Matthias Klose ]
  * Merge with Debian (r4770).
  * Align _start constant pool to 4 bytes (Richard Sandiford). LP: #791315.
  * Adjust Breaks to gcc-4.x for the Ubuntu versions.

eglibc (2.13-8) unstable; urgency=low

   [ Samuel Thibault ]
   * Add patches/hurd-i386/submitted-ldsodefs.h.diff to fix loading binaries
    with GNU/Hurd-specific extensions, disabled for now.
  * patches/hurd-i386/local-sendmsg-SCM_RIGHTS.diff: Do not call getauth(),
    use the __USEPORT() cache macro instead. This should fix zsh FTBFS with
    duplicate getproc() symbol.

   [ Aurelien Jarno ]
   * Add patches/any/cvs-addmntent.diff to correctly report errors status in
     addmntent(). Closes: #630699 / CVE-2011-1089.
   * Add patches/any/cvs-resolv-different-nameserver.diff to try a different
     nameserver if the first one returns REFUSED. Closes: #535504, #602291.
  * Update patches/svn-updates to revision 14337:
    - Remove any/cvs-glro_dl_debug_mask.diff (merged).
    - Remove i386/cvs-memmove-static.diff (merged).
  * debian/control: clean-up Uploaders: .
  * Add patches/any/cvs-fnmatch.diff to fix an integer overflow in
    fnmatch() (CVE-2011-1659). Closes: #626370.
  * Add an entry to NEWS.Debian about multiarch and passing flags to the
    compiler on pre-multiarch toolchains.
  * Replace sparc/submitted-ifunc2.diff by upstream version
    any/cvs-ifunc.diff.
  * Fix patches/hppa/submitted-nptl-carlos.diff to correctly pass
    --as-needed and --no-as-needed to the linker.
  * Update breaks on pre-multiarch packages. Closes: #631907.
  * libc.preinst: improve and simplify search for old libraries, detect
    broken LD_LIBRARY_PATH. Closes: #630608.
  * libc.postrm: remove support code from Sarge.
  * rules.d/debhelper.mk: install bug files using dh_bugfiles.

  [ Petr Salinger ]
  * kfreebsd/local-sysdeps.diff: update to revision 3501 (from glibc-bsd).
    to fix <bits/siginfo.h>. Closes: #631867.
  * Drop kfreebsd/local-ftw.diff, needed only for pre 8.0 kernels.

eglibc (2.13-7) unstable; urgency=low

  [ Aurelien Jarno ]
  * Update paths to the memcpy wrapper in NEWS.Debian.gz. Closes:
    #630138.
  * Fix preinst script wrt 3.0 kernel. Patch by Colin Watson. Closes:
    #630077.
  * debian/sysdeps/linux.mk: correctly check from multiarch asm/ include
    directory.
  * Add patches/sparc/cvs-ifunc1.diff and patches/sparc/submitted-ifunc2.diff
    to fix multi-arch support on sparc. Update sysdeps/sparc.mk and
    sysdeps/sparc64.mk to re-enable multi-arch support on sparc and sparc64.
    Closes: #625607.
  * debhelper.in/libc.preinst: improve error message when an old copy of ld.so
    is found. Closes: #629983.
  * debhelper.in/libc.preinst: automatically remove leftovers from libc5 to
    libc6 transition (upgrade from Bo to Hamm). Closes: #629534.
 -- Matthias Klose <email address hidden> Sat, 02 Jul 2011 11:05:07 +0200

[Rolf Leggewie (r0lf)]

Maverick has long since stopped to receive any updates. Marking the Maverick task for this ticket as "Won't Fix".

[Rolf Leggewie (r0lf)]

maverick has seen the end of its life and is no longer receiving any updates. Marking the maverick task for this ticket as "Won't Fix".

[Rolf Leggewie (r0lf)]

natty has seen the end of its life and is no longer receiving any updates. Marking the natty task for this ticket as "Won't Fix".

[Rolf Leggewie (r0lf)]

lucid has seen the end of its life and is no longer receiving any updates. Marking the lucid task for this ticket as "Won't Fix".