Hello,

on hppa and powerpc, the libreoffice package fails due to a failing smoketest [1]:

[_RUN_____] (anonymous namespace)::Test::test
./unotest/source/cpp/officeconnection.cxx:77:(anonymous namespace)::Test::test
setUp() failed
- equality assertion failed
- Expected: 0
- Actual  : 4

(anonymous namespace)::Test::test finished in: 6ms
officeconnection.cxx:77:Assertion
Test name: (anonymous namespace)::Test::test
setUp() failed
- equality assertion failed
- Expected: 0
- Actual  : 4

Failures !!!
Run: 1   Failure total: 1   Failures: 1   Errors: 0

After some investigation and discussions with upstream [2], I figured out that the problem
is the missing wrapper script instdir/program/soffice for the soffice.bin binary. This script
is normally generated from "desktop/scripts/soffice.sh via makefile targets in desktop/
CustomTarget_sofice.mk and desktop/Package_soffice_sh.mk" (quoting Stephan Bergmann from [3]).

I therefore copied the script manually and rerun the smoketest which immediately succeeded:

glaubitz@perotto:~/loffice/libreoffice-26.2.0~beta1$ cp -av desktop/scripts/soffice.sh instdir/program/soffice
'desktop/scripts/soffice.sh' -> 'instdir/program/soffice'
glaubitz@perotto:~/loffice/libreoffice-26.2.0~beta1$

(sid_powerpc-dchroot)glaubitz@perotto:~/loffice/libreoffice-26.2.0~beta1/smoketest$ make subsequentcheck V=1
rm -rf /home/glaubitz/loffice/libreoffice-26.2.0~beta1/workdir/CustomTarget/smoketest
mkdir -p /home/glaubitz/loffice/libreoffice-26.2.0~beta1/workdir/CustomTarget/smoketest/user
cp /home/glaubitz/loffice/libreoffice-26.2.0~beta1/qadevOOo/qa/registrymodifications.xcu /home/glaubitz/loffice/libreoffice-26.2.0~beta1/workdir/CustomTarget/smoketest/user
[CUT] smoketest
S=/home/glaubitz/loffice/libreoffice-26.2.0~beta1 && I=$S/instdir && W=$S/workdir &&  mkdir -p $W/CppunitTest/ && rm -fr $W/CppunitTest/smoketest.test.user && cp -r $W/unittest $W/CppunitTest/smoketest.test.user &&    rm -fr $W/CppunitTest/smoketest.test.core && mkdir $W/CppunitTest/smoketest.test.core && cd $W/CppunitTest/smoketest.test.core &&   (   MAX_CONCURRENCY=4 MOZILLA_CERTIFICATE_FOLDER=dbm: SAL_DISABLE_SYNCHRONOUS_PRINTER_DETECTION=1 SAL_USE_VCLPLUGIN=svp LIBO_LANG=C      LO_RUNNING_UNIT_TEST=1     env LD_LIBRARY_PATH=${LD_LIBRARY_PATH:+$LD_LIBRARY_PATH:}"$I/program:$I/program":$W/UnpackedTarball/cppunit/src/cppunit/.libs    $W/LinkTarget/Executable/cppunittester $W/LinkTarget/CppunitTest/libtest_smoketest.so --headless "-env:BRAND_BASE_DIR=file://$S/instdir" "-env:BRAND_SHARE_SUBDIR=share" "-env:BRAND_SHARE_RESOURCE_SUBDIR=program/resource" "-env:UserInstallation=file://$W/CppunitTest/smoketest.test.user"    "-env:UNO_TYPES=file://$I/program/types.rdb file://$I/program/types/offapi.rdb"  "-env:UNO_SERVICES=file://$W/Rdb/ure/services.rdb" -env:URE_BIN_DIR=file://$I/program -env:URE_INTERNAL_LIB_DIR=file://$I/program -env:LO_LIB_DIR=file://$I/program -env:LO_JAVA_DIR=file://$I/program/classes --protector $W/LinkTarget/Library/unoexceptionprotector.so unoexceptionprotector --protector $W/LinkTarget/Library/unobootstrapprotector.so unobootstrapprotector   -env:arg-soffice=path:$I/program/soffice -env:arg-user=$W/CustomTarget/smoketest -env:arg-env=LD_LIBRARY_PATH"${LD_LIBRARY_PATH+=$LD_LIBRARY_PATH}" -env:arg-testarg.smoketest.doc=$W/Zip/smoketestdoc.sxw  "-env:CPPUNITTESTTARGET=$W/CppunitTest/smoketest.test" )  2>&1
[_RUN_____] (anonymous namespace)::Test::test
(anonymous namespace)::Test::test finished in: 13398ms
OK (1)
[SCK] smoketest
S=/home/glaubitz/loffice/libreoffice-26.2.0~beta1 && I=$S/instdir && W=$S/workdir &&  mkdir -p $W/Module/subsequentcheck/ && touch $W/Module/subsequentcheck/smoketest
(sid_powerpc-dchroot)glaubitz@perotto:~/loffice/libreoffice-26.2.0~beta1/smoketest$

Thus, we need to figure out why the build doesn't actually generated the wrapper script
for hppa, powerpc (and probably m68k) in order to fix the smokeest on these targets.

Adrian

Hi,

Am 10.12.25 um 08:40 schrieb John Paul Adrian Glaubitz:
That's powerpc, yes.
(experimental_powerpc-dchroot)rene@perotto:~/libreoffice-26.2.0~beta1$ find instdir/ -name "soffice*"
instdir/sdk/examples/DevelopersGuide/ScriptingFramework/ScriptSelector/ScriptSelector/soffice.gif
instdir/program/soffice
instdir/program/sofficerc
instdir/program/soffice.bin
instdir/share/config/soffice.cfg
(experimental_powerpc-dchroot)rene@perotto:~/libreoffi

Source: libreoffice
Version: 4:26.2.0~beta1-1
Severity: normal
User: debian-powerpc@lists.debian.org
Usertags: m68k,hppa,powerpc
X-Debbugs-Cc: debian-hppa@lists.debian.org,debian-68k@lists.debian.org,debian-powerpc@lists.debian.org

Hello,

ce-26.2.0~beta1$ head -n 1 instdir/program/soffice
#!/bin/sh
(experimental_powerpc-dchroot)rene@perotto:~/libreoffice-26.2.0~beta1$

But indeed, that works, interesting.


Regards,


Rene

retitle 1122336 libreoffice: Smoketest fails on hppa and powerpc

thanks


Hi,

Am 11.12.25 um 02:53 schrieb Rene Engelhard:
rebuilt again (you shouldn't use parallel=1...), after it fails I tried this:

mv instdir/program/soffice instdir/program/soffice.old
cp desktop/scripts/soffice.sh instdir/program/soffice
diff -u instdir/program/soffice.old instdir/program/soffice

This results in

(experimental_powerpc-dchroot)rene@perotto:~/libreoffice-26.2.0~beta1$ diff -u instdir/program/soffice.old instdir/program/soffice
--- instdir/program/soffice.old 2025-12-11 01:55:33.611241288 +0000
+++ instdir/program/soffice     2025-12-11 06:08:10.728584683 +0000
@@ -37,7 +37,7 @@
  # uncomment line below if you encounter problems starting soffice on your system
  # SAL_NO_XINITTHREADS=true; export SAL_NO_XINITTHREADS

So, it turns out that the wrapper script wasn't copied because hppa, m68k and powerpc
are listed in OOO_NOGUI_ARCHS which the following patch fixes:

--- debian/rules.orig   2025-12-11 15:56:17.424880233 +0000
+++ debian/rules        2025-12-10 11:19:19.696634221 +0000
@@ -775,7 +775,7 @@
 $(eval $(call gen_no_archs,OOO_REPORTBUILDER_ARCHS))
 OOO_FIREBIRD_ARCHS := $(filter-out m68k,$(OOO_BASE_ARCHS))
 $(eval $(call gen_no_archs,OOO_FIREBIRD_ARCHS))
-OOO_NOGUI_ARCHS := $(filter amd64 i386 arm64 armhf s390x ppc64 ppc64el,$(OOO_ARCHS))
+OOO_NOGUI_ARCHS := $(filter alpha amd64 i386 arm64 armhf s390x hppa m68k powerpc ppc64 ppc64el sparc64,$(OOO_ARCHS))
 $(eval $(call gen_no_archs,OOO_NOGUI_ARCHS))
 OOO_PDFIUM_ARCHS := $(OOO_ARCHS)
 $(eval $(call gen_no_archs,OOO_PDFIUM_ARCHS))

While this change makes sure the script is generated, the script itself still missing the
executable flag meaning that the smoketest will still fail.

Running

	chmod 755 instdir/program/soffice

manually allows the smoketest to pass.

So, while this is fixed halfway, more investigation is necessary.

Adrian

Hi,

Am 11.12.25 um 17:03 schrieb John Paul Adrian Glaubitz:

But the file is there as perotto shows.

See my last reply.

Your "patch" adds them where you say it was listed before. $(filter is not $(filter-out here. This setting
also is supposed to control whether LO is built twice, one time with --disable-gui (*-nogui packages) and one time with GUI.

Don't think you want that on those archs.

See above. This would result in [1]. Not what you want. and the other archs etc just running the basic tests are not there either.

Which would also be solved by removing the JITC_PROCESSOR_TYPE thingy and relying on cp which would copy the +x instead

of relying on sed which dosn't create it +x.


Regards,


Rene

[1]

index ef8c2b4e0..b222415db 100644
--- a/control
+++ b/control
@@ -272,7 +272,7 @@ Standards-Version: 4.0.1
  Rules-Requires-Root: no
  Vcs-Git: https://salsa.debian.org/libreoffice-team/libreoffice/libreoffice.git -b debian-experimental-26.2
  Vcs-Browser: https://salsa.debian.org/libreoffice-team/libreoffice/libreoffice
-Bugs:
+Bugs: mailto:debian-openoffice@lists.debian.org
  Homepage: http://www.libreoffice.org

  Package: libreoffice
@@ -381,7 +381,7 @@ Description: office productivity suite (metapackage)
    * afdko-bin: Handling of embedded fonts during import/export/...

  Package: libreoffice-nogui
-Architecture: amd64 arm64 armhf i386 ppc64 ppc64el s390x
+Architecture: alpha amd64 arm64 armhf hppa i386 m68k powerpc ppc64 ppc64el s390x sparc64
  Section: metapackages
  Depends: libreoffice-base-nogui,
           libreoffice-calc-nogui,
@@ -534,7 +534,7 @@ Description: office productivity suite -- arch-dependent files
   LibreOffice.  See the libreoffice package for more information.

  Package: libreoffice-core-nogui
-Architecture: amd64 arm64 armhf i386 ppc64 ppc64el s390x
+Architecture: alpha amd64 arm64 armhf hppa i386 m68k powerpc ppc64 ppc64el s390x sparc64
  Built-Using: ${cargo:Built-Using}
  Static-Built-Using: ${cargo:Static-Built-Using}
  Depends: fontconfig,
@@ -713,7 +713,7 @@ Description: office productivity suite -- word processor
   This package contains the wordprocessor component for LibreOffice.

  Package: libreoffice-writer-nogui
-Architecture: amd64 arm64 armhf i386 ppc64 ppc64el s390x
+Architecture: alpha amd64 arm64 armhf hppa i386 m68k powerpc ppc64 ppc64el s390x sparc64
  Depends: libreoffice-base-core (= ${binary:Version}),
           libreoffice-common (>= 1:7.0.0~alpha~),
           libreoffice-core-nogui (= ${binary:Version}) | libreoffice-core (= ${binary:Version}),
@@ -755,7 +755,7 @@ Description: office productivity suite -- spreadsheet
   This package contains the spreadsheet component for LibreOffice.

  Package: libreoffice-calc-nogui
-Architecture: amd64 arm64 armhf i386 ppc64 ppc64el s390x
+Architecture: alpha amd64 arm64 armhf hppa i386 m68k powerpc ppc64 ppc64el s390x sparc64
  Depends: libreoffice-base-core (= ${binary:Version}),
           libreoffice-common (>= 1:7.0.0~alpha~),
           libreoffice-core-nogui (= ${binary:Version}) | libreoffice-core (= ${binary:Version}),
@@ -797,7 +797,7 @@ Description: office productivity suite -- presentation
   This package contains the presentation component for LibreOffice.

  Package: libreoffice-impress-nogui
-Architecture: amd64 arm64 armhf i386 ppc64 ppc64el s390x
+Architecture: alpha amd64 arm64 armhf hppa i386 m68k powerpc ppc64 ppc64el s390x sparc64
  Depends: libreoffice-common (>= 1:7.0.0~alpha~),
           libreoffice-core-nogui (= ${binary:Version}) | libreoffice-core (= ${binary:Version}),
           libreoffice-draw-nogui (= ${binary:Version}) | libreoffice-draw (= ${binary:Version}),
@@ -836,7 +836,7 @@ Description: office productivity suite -- drawing
   This package contains the drawing component for LibreOffice.

  Package: libreoffice-draw-nogui
-Architecture: amd64 arm64 armhf i386 ppc64 ppc64el s390x
+Architecture: alpha amd64 arm64 armhf hppa i386 m68k powerpc ppc64 ppc64el s390x sparc64
  Depends: libreoffice-common (>= 1:7.0.0~alpha~),
           libreoffice-core-nogui (= ${binary:Version}) | libreoffice-core (= ${binary:Version}),
           libreoffice-uiconfig-draw (= ${source:Version}),
@@ -872,7 +872,7 @@ Description: office productivity suite -- equation editor
   This package contains the equation editor component for LibreOffice.

  Package: libreoffice-math-nogui
-Architecture: amd64 arm64 armhf i386 ppc64 ppc64el s390x
+Architecture: alpha amd64 arm64 armhf hppa i386 m68k powerpc ppc64 ppc64el s390x sparc64
  Depends: fonts-opensymbol (>= 2:102.11),
           libreoffice-common (>= 1:7.0.0~alpha~),
           libreoffice-core-nogui (= ${binary:Version}) | libreoffice-core (= ${binary:Version}),
@@ -926,7 +926,7 @@ Description: office productivity suite -- database
   This package contains the database component for LibreOffice.

  Package: libreoffice-base-nogui
-Architecture: amd64 arm64 armhf i386 ppc64 ppc64el s390x
+Architecture: alpha amd64 arm64 armhf hppa i386 m68k powerpc ppc64 ppc64el s390x sparc64
  Section: database
  Depends: libreoffice-base-core (= ${binary:Version}),
           libreoffice-base-drivers (= ${binary:Version}),
diff --git a/tests/control b/tests/control
index 4fdfe79c8..ccd91bade 100644
--- a/tests/control
+++ b/tests/control
@@ -35,12 +35,12 @@ Restrictions: superficial
  # the installed LO. works standalone, though, so it's patched out of the
  # smoketest. Do it here separately.
  Tests: test-extension
-Architecture: amd64 arm64 i386 ppc64
+Architecture: alpha amd64 arm64 i386 m68k powerpc ppc64 sparc64
  Depends: libreoffice-core-nogui, libreoffice-common, libreoffice-java-common, libreoffice-smoketest-data, default-jdk
  Restrictions: allow-stderr, superficial, skippable

  Tests: test-extension-shared
-Architecture: amd64 arm64 i386 ppc64
+Architecture: alpha amd64 arm64 i386 m68k powerpc ppc64 sparc64
  Depends: libreoffice-core-nogui, libreoffice-common, libreoffice-java-common, libreoffice-smoketest-data, default-jdk
  Restrictions: allow-stderr, needs-root, superficial

@@ -103,12 +103,12 @@ Depends: libreoffice-dev, libreoffice-dev-doc
  Restrictions: superficial

  Tests: odk-build-examples
-Architecture: amd64 arm64 armhf i386 ppc64 ppc64el s390x
+Architecture: alpha amd64 arm64 armhf hppa i386 m68k powerpc ppc64 ppc64el s390x sparc64
  Depends: @builddeps@, libreoffice-core-nogui, libreoffice-dev, libreoffice-dev-doc, python3-uno
  Restrictions: allow-stderr, skippable

  Tests: odk-build-examples-java
-Architecture: amd64 arm64 i386 ppc64
+Architecture: alpha amd64 arm64 i386 m68k powerpc ppc64 sparc64
  Depends: @builddeps@, libreoffice-core-nogui, libreoffice-dev, libreoffice-dev-doc, libreoffice-java-common, libofficebean-java
  Restrictions: allow-stderr, skippable

Hi,

Am 11.12.25 um 17:21 schrieb Rene Engelhard:

see https://git.libreoffice.org/core/+/refs/heads/master/desktop/CustomTarget_soffice.mk#38

Regards,


Rene

Hello,

Whether it sets "export JITC_PROCESSOR_TYPE=6" or not in the script shouldn't make any
difference unless I'm missing something. This just seems to be an environment variable
that is required when using a JDK provided by IBM but should be safe to ignore otherwise.

Adrian

Am 11.12.25 um 17:21 schrieb Rene Engelhard:

But not +x probably, didn't check that. Which the cp fixes.

See my last reply to that.


That should fix powerpc. The question is what happens with m68k and hppa.

(Note --disable-gui doesn't affect that  magic "X11" conditional either, even if your patch were correctly around and m68k and hppa and
powerpc needed to be removed from it, especially since the checks are ran against a "normal" build)

Regards,

Rene

Hi,

Am 11.12.25 um 17:35 schrieb John Paul Adrian Glaubitz:

OK, was wondering whether OpenJDK choked on it. Then the fix would be to either

a)

$ cat remove-JITC_PROCESSOR_TYPE-setting.diff
diff --git a/config_host.mk.in b/config_host.mk.in
index 3452ad8fde5d..ece7d18116b5 100644
--- a/config_host.mk.in
+++ b/config_host.mk.in
@@ -390,7 +395,6 @@ export JDK=@JDK@
  export JDK_FOR_BUILD=@JDK_FOR_BUILD@
  export JDK_SECURITYMANAGER_DISALLOWED_FOR_BUILD=@JDK_SECURITYMANAGER_DISALLOWED_FOR_BUILD@
  export JFREEREPORT_JAR=@JFREEREPORT_JAR@
-export JITC_PROCESSOR_TYPE=@JITC_PROCESSOR_TYPE@
  JOM=@JOM@
  export JQ=@JQ@
  export JVM_ONE_PATH_CHECK=@JVM_ONE_PATH_CHECK@
diff --git a/configure.ac b/configure.ac
index c1baba596656..04d0db7bd7d3 100644
--- a/configure.ac
+++ b/configure.ac
@@ -9101,14 +9140,6 @@ AC_SUBST(LD_PLUGIN)
  AC_CHECK_FUNCS(posix_fallocate, HAVE_POSIX_FALLOCATE=YES, [HAVE_POSIX_FALLOCATE=NO])
  AC_SUBST(HAVE_POSIX_FALLOCATE)

Hello,

I think it should be safe to drop the complete handling for JITC_PROCESSOR_TYPE_EXPORT.

There is no JDK by IBM for 32-bit anymore, so all of this has become cruft.

Yes, but we can just drop all of this. I don't think that upstream is interested in keeping this.

Adrian

Hi,

Am 11.12.25 um 17:45 schrieb Rene Engelhard:
 >  git diff desktop/CustomTarget_soffice.mk

a) would cleanly fix it (and we don't care about IBM JDK anyway, do we?)

b) would als fix it minimal-invasive but would still have the difference of soffice in the instdir/ tree

(in the installed libreoffice-common it's the same everywhere anyway without that setting anyway. so we already don't care about IBMs JDK)


Trying b) now on perotto.  (hppa has problems with the build-deps so not yet on panama..)


Regards,


Rene

Hi,

Am 11.12.25 um 17:53 schrieb John Paul Adrian Glaubitz:

OK.


Regards,


Rene

clone 1122336 -1

retitle 1122336 libreoffice: Smoketest fails on powerpc due to non-executable soffice

retitle -1 libreoffice: Smoketest fails on hppa

tag -1 - pending

thaniks


Hi,

Am 11.12.25 um 18:00 schrieb Rene Engelhard:
And submitted upstream: https://gerrit.libreoffice.org/c/core/+/195487


For hppa we need another fix (whatever the cause is).


Regards,


Rene

Hello,

I just tried building libreoffice from unstable with qemu-user for hppa
and that worked without problems and without any modifications to the
package:

+------------------------------------------------------------------------------+
| Summary                                      Fri, 12 Dec 2025 06:55:51 +0000 |
+------------------------------------------------------------------------------+

Build Architecture: hppa
Build Type: any
Build-Space: 8934496
Build-Time: 38954
Distribution: unstable
Host Architecture: hppa
Install-Time: 3284
Job: /home/glaubitz/loffice/libreoffice_25.8.3-1.dsc
Machine Architecture: amd64
Package: libreoffice
Package-Time: 42520
Source-Version: 4:25.8.3-1
Space: 8934496
Status: successful
Version: 4:25.8.3-1
--------------------------------------------------------------------------------
Finished at 2025-12-12T06:53:47Z
Build needed 11:48:40, 8934496k disk space

I will retry the package on the buildds.

Adrian

Hi,

Am 12.12.25 um 10:46 schrieb John Paul Adrian Glaubitz:

Tried on panama.

After 5 days of building_do_ get the failure (and another one in the sal tests wrt tempfile, which looks familiar, I think I saw it inside pbuilder, but not outside....)


Regards,


Rene

Hi,

It builds fine for me with qmeu-user, so it could be a timing problem.

see: http://ftp.ports.debian.org/debian-ports/pool-hppa/main/libr/libreoffice/

Adrian

Did you manually built on panama (logged in as debian developer) ?

or a problem which only triggers on physical hppa machines and not in
qemu buildds?

Helge

Hi,

Am 16.12.25 um 23:01 schrieb Helge Deller:
Yes. (nice 19'ed to be nice)


Regards,


Rene

A new rebuild on the qemu hppa buildds will trigger in 30 minutes.
Let's see if it succeeds.
Note that it might be that Adrian has a newer qemu-user binary than we
have on the buildds. If so, I will update the qemu package in the next
few days.

Helge