Bug 90156 - MacOs LO fails to launch with master sources
Summary: MacOs LO fails to launch with master sources
Status: RESOLVED WORKSFORME
Alias: None
Product: LibreOffice
Classification: Unclassified
Component: LibreOffice (show other bugs)
Version:
(earliest affected)
4.5.0.0.alpha0+ Master
Hardware: Other All
: medium normal
Assignee: Not Assigned
URL:
Whiteboard:
Keywords:
Depends on:
Blocks:
 
Reported: 2015-03-22 10:59 UTC by Julien Nabet
Modified: 2015-04-07 06:50 UTC (History)
4 users (show)

See Also:
Crash report or crash signature:


Attachments
bt (53.74 KB, text/plain)
2015-03-22 11:01 UTC, Julien Nabet
Details

Note You need to log in before you can comment on or make changes to this bug.
Description Julien Nabet 2015-03-22 10:59:41 UTC
On MacOs Yosemite 10.10.12 + master sources updated some days ago, LO fails to launch with this error:
warn:desktop.app:48508:1:desktop/source/app/langselect.cxx:125: ignoring Exception "component context fails to supply service 'com.sun.star.configuration.ReadOnlyAccess' of type 'com.sun.star.container.XHierarchicalNameAccess'"
libc++abi.dylib: terminating with uncaught exception of type com::sun::star::uno::DeploymentException
Abort trap: 6

here's my autogen.input:
--enable-64-bit
#--with-system-odbc
--enable-werror
#--enable-python=internal
--enable-debug
--enable-dbgutil
--enable-crashdump
--enable-dependency-tracking
--enable-online-update
#--enable-ext-mariadb-connector
#--without-system-mariadb
#--enable-bundle-mariadb
#--with-lang=en-US it fr de es pt ru cs hu pl da sv el sk is nl
--with-lang=ALL
--without-junit
--with-myspell-dicts
Comment 1 Julien Nabet 2015-03-22 11:01:12 UTC
Created attachment 114242 [details]
bt
Comment 2 Robinson Tryon (qubit) 2015-03-31 17:24:38 UTC
(In reply to Julien Nabet from comment #0)
> On MacOs Yosemite 10.10.12 + master sources updated some days ago, LO fails
> to launch with this error:
> ...

I see current master builds from today available here:
http://dev-builds.libreoffice.org/daily/master/MacOSX-10.10@61/

Does this problem persist? Is it something to discuss on the dev list?

Status -> NEEDINFO
Comment 3 Julien Nabet 2015-03-31 18:28:01 UTC
I haven't spoken about it on dev mailing list since few people use MacOs + debug. (same reason about IRC)
About the daily build, is it a debug build? Indeed, I use debug build and I don't see config file on this daily.

About building with recent sources, I gave it a try about a week ago.
I give a new try from scratch tomorrow.

Finally, I mainly work on Linux with master sources + debug mode but since I've got a Mac, I'm trying to contribute on some MacOs only bugs when I've got time.
In brief, it's not urgent for me :-)

I'll put it back to UNCONFIRMED (or perhaps RESOLVED/WFM!) once I'll got results from new build.
Comment 4 Julien Nabet 2015-04-01 18:56:23 UTC
With master sources updated today, I still got have the same problem.

Alex: do you still build LO master sources? (I don't remember if you used debug mode or not) If yes, any thought about what may fail in my case?
Comment 5 Alex Thurgood 2015-04-01 21:11:26 UTC
Hi Julien, 
I build MacOS debug incremental builds daily from master, and roughly one or two full clean builds per week too. No problems here with starting LO, either via Finder or command line.

Alex
Comment 6 Julien Nabet 2015-04-01 21:17:32 UTC
I use git sources, run "make" then I type "instdir/LibreOfficeDev.app/Contents/program" and finally "./soffice" as I always do on Linux.
Wonder what's the problem :-(

Anyway, thank you for your feedback.
Comment 7 How can I remove my account? 2015-04-01 21:28:53 UTC
Works fine here, although a simpler way to start LO is to give the command:

open instdir/LibreOfficeDev.app

(or double-click on that LibreOfficeDev.app in Finder)

Maybe you have some cruft accumulated in instdir/Contents/user; try make clean?
Comment 8 Pieter van Oostrum 2015-04-01 22:45:23 UTC
I built master on Mac OS X 10.10.2 yesterday without problems. But I use the --enable-python=internal option.
I remember in the past I had also a DeploymentException. That was when I built it without Python, but I don't know if there was a causal connection. Anyway the DeploymentException disappeared when I built with Python.
I also added PYTHON=python3.3 to the autogen.sh command
Comment 9 Julien Nabet 2015-04-02 05:36:21 UTC
I changed autogen.input to have:
--enable-python=internal
+
I tried make clean before building
and tried too "open instdir/LibreOfficeDev.app"

Same result.

I haven't yet tested PYTHON=python3.3 but is it useful since I use now --enable-python=internal ?
Comment 10 Stephan Bergmann 2015-04-02 08:29:19 UTC
looks like you have no instdir/LibreOfficeDev.app/Contents/Frameworks/libconfigmgrlo.dylib or it cannot be loaded (in which case running "SAL_LOG=+INFO.sal.osl instdir/LibreOfficeDev.app/Contents/MacOS/soffice" might give a clue on stderr)
Comment 11 Julien Nabet 2015-04-02 16:51:29 UTC
Stephan:
here are the information you talked about:

$ ls -l instdir/LibreOfficeDev.app/Contents/Frameworks/libconfigmgrlo.dylib
-rwxr-xr-x  1 julien  staff  2669292  2 avr 12:38 instdir/LibreOfficeDev.app/Contents/Frameworks/libconfigmgrlo.dylib

$ SAL_LOG=+INFO.sal.osl instdir/LibreOfficeDev.app/Contents/MacOS/soffice
info:sal.osl.condition:59030:1:sal/osl/unx/conditn.cxx:74: osl_createCondition(): 0x7fc96257d3a0
info:sal.osl:59030:1:sal/osl/unx/module.cxx:314: osl_getModuleURLFromAddress: /Users/julien/lo/libreoffice/instdir/LibreOfficeDev.app/Contents/Frameworks/libuno_cppu.dylib.3
info:sal.osl:59030:1:sal/osl/unx/module.cxx:314: osl_getModuleURLFromAddress: /Users/julien/lo/libreoffice/instdir/LibreOfficeDev.app/Contents/Frameworks/libuno_cppuhelpergcc3.dylib.3
info:sal.osl.condition:59030:1:sal/osl/unx/conditn.cxx:74: osl_createCondition(): 0x7fc96273b610
info:sal.osl.condition:59030:1:sal/osl/unx/conditn.cxx:74: osl_createCondition(): 0x7fc96273b4b0
info:sal.osl.pipe:59030:1:sal/osl/unx/pipe.cxx:243: new pipe on fd 15 '/tmp/OSL_PIPE_501_SingleOfficeIPC_880fd1d5bf168a495522ffb52bc994'
info:sal.osl:59030:1:sal/osl/unx/module.cxx:314: osl_getModuleURLFromAddress: /Users/julien/lo/libreoffice/instdir/LibreOfficeDev.app/Contents/Frameworks/libuno_cppu.dylib.3
Comment 12 Julien Nabet 2015-04-02 16:52:46 UTC
forgot to say that to be sure there's no remaining cruft, I removed my local repo and gave a new try from scratch
--enable-64-bit
#--with-system-odbc
--enable-werror
--enable-python=internal
--enable-debug
--enable-dbgutil
--enable-crashdump
--enable-dependency-tracking
--enable-online-update
#--enable-ext-mariadb-connector
#--without-system-mariadb
#--enable-bundle-mariadb
--with-lang=en-US it fr de es pt ru cs hu pl da sv el sk is nl
#--with-lang=ALL
--without-junit
--with-myspell-dicts
Comment 13 Julien Nabet 2015-04-03 22:11:16 UTC
After having updated again my LO sources, it works now!

I wonder if it could be thanks to:
http://cgit.freedesktop.org/libreoffice/core/commit/?id=514f9fc675eb3d116adb4b0a33c19c3d6e99e98b
CppunitTest_sw_ooxmlexport5: fix Fedora 22 alpha build
09:41 <@dtardon> vmiklos, maybe it's a regression in libxml2. "Fix XPath '//' optimization with predicates (Nick Wellnhofer)" in NEWS looks suspicious So tweak the XPath not to use // for now. 

Indeed, on my other machine, pc Debian x86-64 with master sources updated this morning, it failed to launch with the same message on console.
After having updated my local repo (with this quoted commit included), it was ok.

Perhaps it's a wrong lead since I don't know why a failing test would cause this. Moreover, even after the last local repo update, check tests failed on my Mac with this:
Assertion failed: (nMapNum == 0 || std::abs(n) < std::numeric_limits<long>::max() / nMapNum / nDPI), function ImplLogicToPixel, file /Users/julien/lo/libreoffice/vcl/source/outdev/map.cxx, line 379.

Error: a unit test failed, please do one of:

export CPPUNITTRACE="lldb --"       # for interactive debugging on OS X
export VALGRIND=memcheck            # for memory checking

and retry using: make CppunitTest_sw_globalfilter

so I don't know what to think.

Anyway, let's put this one WFM, thank you all for your feedback! :-)
Comment 14 Stephan Bergmann 2015-04-07 06:50:11 UTC
(In reply to Julien Nabet from comment #13)
> Perhaps it's a wrong lead since I don't know why a failing test would cause
> this.

So in the past a plain "make" did not return with a zero exit code for you?  That would have been a very vital information to track down the root of your problem.  If a plain "make" (vs. e.g. "make check") returns with a non-zero exit code (even due to a failing unit test) the generated instdir/ tree is, in general, in an indeterminate state and cannot be used reliably.