Pavel's diary

Builds on GNU/Linux are finished and uploaded.

Changes that are worth testing include:

• child workspace helpcl was included, which means that #i22548# was fixed and I was even able to find the word "záložka" in the help. This will make help translation easier, because you'll be able to search in it.
• child workspace hc680postbeta was integrated, so there are many changes in helpcontent
• I included changes from sb to fix the "decline license crash" issue (#i41130#)
• Czech Contents tree in help is now completely visible after I found, filed and fixed #i41607# in my build system.
• and last, not least, I have identified my problem with KDE plugin build to come from vclcompact child workspace, so I have temporarily disabled VCL visibility fixes and re-enabled KDE plugin again :-)

Builds for Solaris/SPARC and Microsoft Windows will follow soon. -----

The module avmedia contains 18 strings that were not exported from the source code using localize tool (see #i41611# for more details). I fixed this bug in my source tree and prepared new POT files with this module. -----

SRC680_m75 is the third milestone published this week. It includes: adc10, babelfish05, cd04, cd05, dvoqbfix, hc680postbeta, helpcl, impress27, impress32, kso13, kso14, oasisbf4, pbbeta07, qwizardsbf2, scriptingf10, sj14, smoketest4, swqbugfixes13, tbe17 and xmlperf01.

As it contains two helpcontent2 related child workspaces, it will be funny for translators ;-) I have also uploaded new POT files for this milestone. -----

No, this time, I won't be that exhaustive like I was yesterday.

Due to #i41328# I had to update the compiler on my Solaris/SPARC machine. My current versions are:

cc: Sun C 5.5 Patch 112760-15 2004/11/10
CC: Sun C++ 5.5 Patch 113817-10 2004/10/05

So right now, I should be using the same compiler like Hamburg RE is using on Solaris/SPARC and also on GNU/Linux.

The only architecture my build system is missing now is Solaris/x86. Any volunteer to provide access to this architecture? -----

I have uploaded POT files for milestone SRC680_m74s1. Only minor string changes. New strings for context menu on the tool bar. -----

The milestone SRC680_m74s1 is now ready with these child workspaces integrated: configure4, gcj1, systempython, trabugfixingso8 and unxlngs01. Please test carefully! -----

For past ~10 years, I used to work with free and open source software exclusively. Unfortunately, I also have to work with proprietary software sometimes. The culture of programmers there is very different from what I'm used to and expect also from others. And I was (again) not well prepared for such huge difference. People working on proprietary operating systems are somehow automatically prepared to expect the bad quality of the software they use, but I'm not.

The current experience I have now is LanSafe 5 software from PowerWare. In one system, we used the previous version of this software, called LanSafe 3. In the meantime, they produced LanSafe 4 too and we decided to update to the newest version recently.

And there are issues.

Issue nr. 1 is not directly about programmers, but about marketing of such software. It supports several operating systems (e.g. AIX, HP-UX, IRIX, Linux, Mac OS, NetWare, SCO OpenServer, Solaris/SPARC, Solaris/Intel and various variants of Microsoft Windows). The main feature announced on the main page is support for USB communication. Once you read the release notes, you are surprised that USB is not supported on AIX, HP-UX, IRIX, Linux, Mac OS, NetWare, SCO OpenServer, Solaris/SPARC and Solaris/Intel. Yes, if you read carefully, you'll realize that USB is working only on Microsoft Windows. Two ways to announce this feature. They have chosen the bad one and thus lost me as a potential customer for the future.

Issue nr. 2 is interesting from programmers point of view. LanSafe 5 installation process tries to add several lines into /etc/services - the port number is not static but depends on the actual contents of the file before installation... Ugly... The comment that is going to be added is about static port number 3068 though. In our case, the port number was 3097 and the comment about 3068. Interesting.

But this is not everything about /etc/services handling. Before adding the above mentioned lines, it will *delete* the lines containing string "ls", yes you read carefully - the letter "l" followed by letter "s" ("because" previous version used ls3 service name) and thus it will delete definition of the following services: afs3-vlserver, afs3-volser, allstorcns, b-novative-ls, cops-tls, cplscrambler-al, cplscrambler-in, cplscrambler-lg, dicom-tls, dls, dls-mon, dls-monitor, dlsrap, dlsrpn, dlswpn, els, etlservicemgr, f5-globalsite, h323callsigalt, h323hostcallsc, hlserver, hpoms-ci-lstn, hpoms-dps-lstn, htuilsrv, iclpv-nls, ilss, imip-channels, imqtunnels, initlsmsad, ipulse-ics, llsurfup-http, llsurfup-https, lmsocialserver, lsp-ping, lstp, mcs-mailsvr, ndl-als, olsr, pslserver, realsecure, scipticslsrvr, seclayer-tls, sip-tls, slslavemon, sns-channels, sonuscallsig, sqlserv, sqlsrv, ssm-els, stonefalls, tekpls, vlsi-lm and whosells. Only because they all have "ls" in their name. Smile?

Issue nr. 3: when the power is down and system is to be halted in e.g. 5 minutes, it should warn you/all users of the system using the script called PowerMonitor.broadcast which itself works. But no one is calling it...

Last, but not least, when you run Lansafe 5's PowerMonitor with -h, it will tell you, it is LanSafe 3. And in fact, it also looks like very old, unmaintained, crappy proprietary software :-(

Do you have similar experience with proprietary software? -----

Hamburg RE team announce the availability of milestone SRC680_m74 with the following child workspaces integrated: ause016, bm4, dba22, fwkperf01, impress23, svg03, swqbugfixes12, vcl33 and visibility01.

For l10n people, child workspace ause016 is interesting, because with its integration, you do no longer need to edit any files to add/deliver your localized README file.

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

Czech Post Turns to SUSE Linux

Anything more to add? Yes. I still wonder, why this is not on the front page of Slashdot... -----

Jacqueline announced (in art@marketing) that eleven entries were pre-selected for voting by the community. The process of voting is to be announced soon.

Remember my old diary entry about the splash screen? My favourite splash screen (MD5SUM f509db07c9dec4da53cdfb290647ac9d) is in the pre-selection. Thanks!

Unfortunately, there is another splash screen that is even better in my opinion. This time MD5SUM is 075caa55c8fd36b2200fc862822db17c. The voting will be death-match ;-)

P.S. MD5SUM here is not MD5SUM of the image or its name. It is a md5sum(f(image name, image, author)) where f is unpublished (yet) function ;-) -----

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

During this weekend, I committed all patches Kendy and me generated against m71s1 to the child workspace ooo64bit02 and made several additional changes needed there (back merging stuff like LINK in the environment, etc.) I packed the complete tree we use now to OOo_ooo64bit02_source.tar.gz.

Additional patches are needed right now:

• because we decided (for now) to ignore binfilter and also java builds, we need patches to disable pieces of build/final packing (OOo_ooo64bit02_source-64bit-temp-do-not-build-binfilter-testtools.diff, OOo_ooo64bit02_source-64bit-temp-scp2-without-binfilter-1.diff, OOo_ooo64bit02_source-64bit-temp-scp2-without-binfilter.diff, OOo_ooo64bit02_source-64bit-temp-solenv-without-binfilter.diff).
• because of (probably!) broken gcc-3.4.1 optimization (#i31057#), we have to turn it off using OOo_ooo64bit02_source-64bit-i31057-solenv.diff. Kendy's machine uses newer gcc, and he said he is not using it. I'll try to update my compiler soon.

Using these patches and gcc-3.4.1 with enum and visibility patches, we were able to show the splashscreen and almost immediately crash :-) There seems to be an issue in bridges for AMD64 (x86_64). I created *brute force* patch that simply doesn't raise the exception in bridges, and I was able to get a bit further (OOo_ooo64bit02_source-64bit-temp-bridges.diff).

I have uploaded both install tree and also the log from my build. It is really install tree only, not RPMs, because epm (although packaging successfully) haven't created them. But this is minor issue right now.

So: what is working, what is not and how to get the most of it? Before starting the soffice, you have to manually create the directory ~/.openoffice.org1.9.71.1/user/config/soffice.cfg. After that, you can run it and the window "Welcome to OpenOffice.org 1.9.71.1" will be opened. You can click "Cancel" here, and OpenOffice.org will exit successfully. The button "Help" here doesn't work, it SIGSEGVs in SfxObjectShell::GetModel. The same crash will be described below while running Help from UI.

The button "Next" works, of course, so you can go further in the Installation wizard. Now you can Decline the license or accept it. Yes, you really can Decline the license even though you can't decline it in SRC680_m73 (see #i41130# for more details ;-).

After accepting the license, you'll see "User Information" dialog. Do not try to set focus by clicking in any field there, because it will crash in _XIMCountVaList in libX11.so.6 (called from SalI18N_InputContext::SalI18N_InputContext).

When you click "Next" here, you will see the dialog "OpenOffice.org Registration". You can click on radio buttons happily (the first one does not work also in x86 build and it is already known/fixed somewhere).

After clicking "Finish", you should see GUI. Now you can do the following here:

• move the mouse over toolbar and see tips
• undock/dock the toolbar
• lock/unlock the position of the toolbar

Do not try to activate menu items, it will crash.

I could get more progress by copying the user directory configuration from my x86 machine to my AMD64 machine and thus starting OpenOffice.org on AMD64 machine with the user config directory from classic Intel machine. Now, more things are working:

• menu walking
• About dialog (Ctrl+SDT works too)
• you can turn on extended tips
• you can open the "Open file..." dialog, but it will crash almost immediately in _XIMCountVaList
• you can open Tools - Customize and Tools - Options dialogs

But many pieces still do not work (Help, DicOOo etc).

I'd be happy if you can confirm my findings. Please mail me about your testing - either with the install tree I provided or (better ;-) in your own build. Thanks. -----

Hard day, lot of work. I produced POT files for SRC680_m73, translated several strings into Czech, dumped several patches into child workspace to have them verified by others, prepared patch for #i40720#, reported one crasher (#i41130#) in installer dialog on GNU/Linux, tried building Mozilla from source on Solaris/SPARC (thus reopening #i11424# and providing simple fix), re-enabled GTK VCL plugin on GNU/Linux because GTK requirements were normalized again.

I also dumped almost all 64bit patches for AMD64 platform into child workspace ooo64bit02 and will continue to work on it tomorrow. -----

New milestone SRC680_m73 is ready for use. It integrated cfgsmartcache, gtkfpicker2, industrial01, os46, pj16, res32bit2, solarisgcc01, systemboost and xmlsec07 child workspaces. -----

I have just uploaded POT files for milestone SRC680_m73.

Changes include several removed files (*/sdi.pot, see #i33585#). This fix was done in the child workspace fwkbugfix04 which, unfortunately, is not shown in EIS and also was not mentioned in the "SRC680_m73 is now ready" mail from Hamburg RE. What is the reason for this? -----

Finnish team has implemented translations statistics of PO files. How many strings of coming OpenOffice.org 2.0 do you have translated? -----

POT files for SRC680_m72 were uploaded.

Changes include: several new strings in several modules (like basctl and dbaccess). Helpcontent2 already doesn't contain xml-lang tags (see #i40909#) and also comment tags were removed. -----

Hamburg release engineering team has just finished their work on integrating child workspaces ause015, babelfish, cd02, dr31, notesapi, presperf01, tbe15, tbe16 and vclcompact into the master and thus SRC680_m72 is ready for use.

Test very carefully, especially babelfish changes affecting UI. -----

I have uploaded new POT files for SRC680_m71s1. If you want to update from the previous version, please README first. -----

POT files for the recent milestone SRC680_m71s1 are now ready.

Changes from previous version: one typo in English string fixed and one new string (because of a bug fixed in the source code). -----

SRC680_m71s1 has been tagged now and is ready for use. Child workspaces integrated: calc27, dbgdump1, jb01, macosx01, monaleenfilterteam30, pj15, sb19 and sb25.

s1 suffix was somewhat new for me, but after reading OpenOffice.org source control again, I know about it ;-) Thanks JoergB for hint. -----

With the help of Josef, I have turned on the build of Mozilla from source in my build system also for Windows. We will see, how it works ;-)

I'm testing Ivo's patch for removing xml-lang's from the <image tags in the helpcontent2 module. If it works, I'll integrate it into my build system and the next round of POTs won't contain any xml-lang tags. -----

Today, I planned to spend the evening on syncing my tree with Kendy's recent developments on AMD64 porting and syncing it with SRC680_m71 because of weird problems in older milestones. But I forgot that libxmlsec needs Mozilla's nspr.h etc. So...

So I integrated #i11424# ("remove the need for pre-compiled Mozilla tarballs from OOo build environment") into my build system and am using it now on GNU/Linux on both x86 and AMD64. Thanks to _rene_ who fixed issues found almost instantly in the respective child workspace.

I'll keep the old way of building Mozilla on Solaris/SPARC and Windows build for a while to verify the effect of new Mozilla. -----

Hamburg RE team is trying to flood us with newer milestones :-) Thanks!

SRC680_m71 is now ready with these child workspaces integrated: cfgimpex, os45, pbbeta04, sj13 and vcl32. -----

Czech Native-Lang project is now working on translation of OpenOffice.org 2.0 strings. We also have a page that shows the current status of our work.

Right now, we have just passed 61% level (GUI is translated from 99%, help content from 38%). Looks like we need to search for some funds (both human and financial) or for better motivation of translators. -----

SRC680_m70 has been released only two days after m69 (which only integrated one child workspace). m70 contains 8 child workspaces: dba20, fooobuild (finally ;-), html03, localedata2, os44, smoketest3, swqbugfixes11 and vq08. -----

... is exactly 6 months old right now. Happy sleeping now ;-)

We have a lot of fun with him. If you are interesting in some family photos, check out Michal's diary too. -----

Eric Bachard collected and filed MacOS X build issues and posted summary of them to the list.

If you are interested, look here for the complete list. -----

SRC680_m69 has been marked as ready by Hamburg Release Engineering team. The only child workspace integrated is res32bit which fixes several issues in resource manager and format of resources. -----

I have uploaded new POT files updated to the latest milestone SRC680_m69. No new strings (only two in options.po which are used only in Russian version), but I have fixed several typos/issues in my source tree, thus some strings are changed.

You can find them at ftp.linux.cz and also at ooomisc.services.openoffice.org. -----

Just a random observation: my child workspace pj15 which was created 1.1.2005 has cwsid 2005 ;-) -----