Pavel's diary

This diary entry describes the current status of OpenOffice.org 64bit porting to AMD64 architecture, effort of several people. You can read previous status here.

Little progress this weekend. I have identified and provided workaround or fixed several issues:

• #i43612# - 64bit: gccinstlib.pl and "dual" gcc.
• #i43601# - 64bit/AMD64: testtool and spadmin crashes
• #i43604# - 64bit: support x86_64 RPM target for RPM packages
• #i43592# - 64bit: Make help working
• #i43598# - 64bit: Events are 32bit or 64bit?

To sum up, the following works comparing to the last month status: Martin fixed crash in _XIMCountVaList (#i43501#), help pages work, only images in help pages do not show (only Read-Error is flashing), installer creates RPM packages, testtool displays its UI and the same for spadmin.

You still have to run mkdir -p ~/.openoffice.org1.9.71.1/user/config/soffice.cfg before starting OpenOffice.org for the first time as described in previous status.

How to crash it? There are several ways to do so ;-)

1. after opening the main window, move it. OpenOffice.org will crash after throwing com::sun::star::uno::RuntimeException in framework/source/accelerators/moduleacceleratorconfiguration.cxx
2. in help: select flashing "Read-Error" image and press Ctrl+C, the same crash as above
3. after Ctrl+N in the main window, the same crash as above
4. ... to be continued ...

Do you want to try it? I will upload everything I have at ftp.linux.cz once new clean build is finished.

How to help? Try to compile OpenOffice.org on AMD64 yourself. Take one crash from the above list and debug it and provide fixes. Try to find more places where OpenOffice.org crashes and describe them.

Still a lot of work. Bridges expert needed... -----

Second post-beta milestone is now ready. Integrated child workspaces include: ause022, calc29, dbwizard3, dr32, fwk1, fwkpostbeta02, iha08, kendy03, luic, moresystemstuff, mozab05, nsplugin5, pj19, pj20, sb29, sj15, swqbugfixes14, swqcore05 and swqcore06.

Build is almost without problems (my first build was almost successful, except missing dependency in module linguistic - #i43496#). -----

Something unbelievable happened here in CZ.

The business environment here in CZ is very unfriendly: unbelievable bureaucracy, complex and opaque tax system etc. It is really hard to do your own business here. Czech government is doing everything they can to make it even worse. But they do not want to confirm that. But very interesting thing happen today.

Prime minister's wife told this sentence to press: "You simply can't do business in this country." I hope that Ms. Šárka Gross can tell this also to her husband... -----

Post-beta milestones are here. SRC680_m80 contains following child workspaces: aw026, dba23, dr33, dr35, extras17, hr11, hsqldb2, icontool06, impress31, impress33, os51, pasf4ooo, perform01, rpath1, sdk14, slideshowgcc, tablestrings, visibility02, vq11 and winfpickerfixes.

If you build on Windows/tcsh, you do no longer need to use configure parameter --with-use-shell=tcsh because it should be the default now.

Update: see #i43297# if your build fails in chart2 due to "missing" images. -----

I had a Python day today. Child workspace pyunofixes1 and other fixes brought a new issues into my Windows builds (#i42947#) and much later (because of slow machine), similar problem appeared on MacOS X. Eric Bachard was so kind to report this issue, because he has seen it too (#i43074#).

The problem is that the python binary built as part of OpenOffice.org build process is run and must load dynamic library pyuno.dylib. The problem in Python-2.3.4 was, that on MacOS X/Darwin, the source file dynload_next.c (as in NextSTEP) was used instead of dynload_shlib.c. Second issue (again) was that Python doesn't know .dylib suffix as a valid shared object extension (again: the inventor of .dylib should be punished somehow). Unfortunately, this issue can affect more platforms.

But anyway, I'll start Build-3 of m79 on all platforms with updated GSI/SDF files tonight. I'll also include not-yet-perfect workaround for bus errors of testtool and spadmin on MacOS X so QA on these binaries can start. This has to be fixed correctly though. -----

I have added Danish to the list of languages I produce builds for.

Ause prepared fix for #i42947# and I verified it in my build on Windows. Builds of m79 are ready for GNU/Linux (both RPMs and DEBs) and Solaris/SPARC. Mac OS X builds are almost finished, Windows build will be in Build-2.

pj20 and macosx02 are now in "Ready for QA" status. Cleaning my desk to be ready for beta! ;-) -----

Milestone SRC680_m79 is now ready. Integrated child workspaces are: ause018, ause020, babelfish06, buildcond02, configure5, dmakevercheck, fpic1, kendy01, pj17, pj18, pyunofixes1, vq09 and vq10.

And now, we are close to beta. pj19 with thesaurus needs to be integrated and python build issue (import os reusing os.pyc file from solver *while* still building module python) on Windows/tcsh (at least) have to be solved (#i42947#).

Right now, there are 577 issues with target 2.0. There are at least five issues about thesaurus that have to be verified once pj19 is integrated. -----

Starting from Build-4, my build system will be able to produce DEB packages when building on GNU/Linux.

I have also added MD5SUMs so people are able to check downloaded packages. -----

I have switched my primary OpenOffice.org installation from 1.1.4 to m78 to help beta-testing once we are in beta phase. Hope you are the next one! There are only 595 bugs with target 2.0 as of now, so we have a lot to do now. -----

News from today: m78 released, thesaurus works in pj19, Czech extras are in pj20 for testing, a lot of work on MacOS X builds (have you ever had four terminals opened on one machine and building four modules in parallel?), fixed some typos and oversights in helpcontent2 and a quick IZ session on bugs targeted to 2.0 (looks good now :-).

Cloph in one of his mails adviced that you can add your custom queries to your IZ pages. Nice and can save *a lot* of time! -----

SRC680_m78 milestone is now ready. Integrated child workspaces: chartbeta1, cmcfixes1, impressbeta1, impressbeta2, impressbeta3, impressbeta4, impressbeta5, os53, preselectionfix, swqbeta001, swqbetterbeta, sysui04 and tbe20.

Caolan's child workspace integration means that you should not need any patches for gcc 3.3.x now.

POT files are uploaded too. -----

I was able to get the list of MyIssues down to 63! Almost half of one of my new year wishes finished ;-) -----

Sun's Hamburg release engineering team prepared SRC680_m77 milestone for general use. Integrated child workspaces are: alienwarning, configuretypofix, fwkbetashowstop01, imprec01, mmeeks04, twobeta and validationbeta.

I have also uploaded new POT files for this milestone. -----

SRC680_m76 is now ready. Integrated child workspaces: ause017, aw023, aw025, calc28, frrtf01, fwkbeta03, fwkbugfix05, iha06, impress20, impress30, jre5issues, mav15, nativefixer3, os47, presbeta01, sdksample, vcl34, vcl35, wordmlfilter01 and xsltfilter06.

I have also uploaded POT files for SRC680_m76. -----