autopkgtests failures on i386 and s390x for 5.4.0, assumes wrong architecture

Bug #1712873 reported by Olivier Tilloy
12
This bug affects 2 people
Affects Status Importance Assigned to Milestone
libreoffice (Ubuntu)
Fix Released
High
Olivier Tilloy

Bug Description

Looking at autopkgtests failures on i386 and s390x for 1:5.4.0-0ubuntu2 in artful-proposed (logs attached).

The following linker failure:

    /usr/bin/ld: cannot find -ljawt

is due to $JAVA_PROC_TYPE being incorrect ("amd64" instead of "i386" or "s390x"), and thus $LINK_JAVA_LIBS has the incorrect architecture in its path and the linker doesn't find libjawt.so.

This is likely caused by $PROCTYPE being incorrect, which itself is set from $PLATFORMID.

There are other hints that $PROCTYPE has a wrong value in the logs, even though that doesn't appear to make the tests fail:

    […]
    mkdir -p /tmp/tmp.OSVUCMLVqv/out/sdk/LINUXexample.out/misc/CppComponentSample/Linux_x86_64
    […]

which means $UNOPKG_PLATFORM has an incorrect value, set from $PROCTYPE.

Revision history for this message
Olivier Tilloy (osomon) wrote :
Revision history for this message
Olivier Tilloy (osomon) wrote :
Revision history for this message
Olivier Tilloy (osomon) wrote :

PLATFORMID is set in debian/rules:

debian/rules:569:PLATFORMID := $(shell grep PLATFORMID debian/vars.$(DEB_HOST_ARCH) | cut -d"=" -f2)

and unsurprisingly:

./debian/vars.i386:1:PLATFORMID=linux_x86
./debian/vars.s390x:1:PLATFORMID=linux_s390x
./debian/vars.amd64:1:PLATFORMID=linux_x86_64

I'm wondering whether DEB_HOST_ARCH could be always (incorrectly) expanded to amd64 in the context of the autopkgtest runner?

Changed in libreoffice (Ubuntu):
assignee: nobody → Olivier Tilloy (osomon)
importance: Undecided → High
status: New → Triaged
Revision history for this message
Olivier Tilloy (osomon) wrote :

In an x86 VM:

$ uname -p
i686

$ cat /usr/lib/libreoffice/sdk/settings/dk.mk
PLATFORM=linux
PROCTYPE=x86_64
SDKVERSION=5.4
BUILDID=5.4.0.3
LIBO_SDK_ENABLE_DBGUTIL=

This is where the incorrect PROCTYPE comes from. It is being installed by libreoffice-dev-common, supposedly an architecture-independent package.

Olivier Tilloy (osomon)
Changed in libreoffice (Ubuntu):
status: Triaged → In Progress
Revision history for this message
Olivier Tilloy (osomon) wrote :
Revision history for this message
Olivier Tilloy (osomon) wrote :
Changed in libreoffice (Ubuntu):
status: In Progress → Fix Committed
tags: added: block-proposed
tags: removed: block-proposed
Revision history for this message
Launchpad Janitor (janitor) wrote :

This bug was fixed in the package libreoffice - 1:5.4.1-0ubuntu1

---------------
libreoffice (1:5.4.1-0ubuntu1) artful; urgency=medium

  [ Olivier Tilloy ]
  * debian/rules: move usr/lib/libreoffice/sdk/settings/dk.mk from
    libreoffice-dev-common to libreoffice-dev (LP: #1712873)

  [ Rico Tzschichholz ]
  * New upstream release
  * Drop lo-xlate-lang-update.diff, add-libebook-1.2.so.19-in-EApi.cxx.patch,
    examples-fix-SDK_AUTO_DEPLOYMENT-check.diff, applied upstream

  [ Rene Engelhard ]
  * debian/patches/stop-shipping-mimelnk-desktop-files.diff: as name says
    (closes: #872001)
  * debian/patches/java9.diff: backport Java 9 patch from master
  * debian/rules:
    - run dh_strip_nondeterminism
    - set locale to en-US.UTF-8 also for make check, the dbaccess tests
      fail in some frenchy locale...
    - move pagein-{calc,draw,impress,writer} to their respective packages
  * debian/scripts/gid2pkgdirs.sh: fix move of types/*.rdb and services/*.rdb
    to core and move pagein-common there too. They reference libmergedlo.so
    which doesn't exist on all archs (closes: #873443)

 -- Olivier Tilloy <email address hidden> Fri, 01 Sep 2017 07:25:31 +0200

Changed in libreoffice (Ubuntu):
status: Fix Committed → Fix Released
To post a comment you must log in.
This report contains Public information  
Everyone can see this information.

Other bug subscribers

Remote bug watches

Bug watches keep track of this bug in other bug trackers.