Google Earth Look Bad?

While waiting for our Fedora ISO to download, I have been tinkering with some small-ish projects. There have been some calls about the font rendering in Google Earth 4. I brought down version 5 for testing and found the fonts were still not anti-aliased and looked terrible on our GNOME desktop. This post explains that Google seems to ship static QT libraries that do not render well. The post explains how to move the shipped libraries out of the way and then softlink into the QT libraries which exist in /usr/lib. Google Earth now looks wonderful. I'm going to come up with settings that work well with remote display and Compiz and push this all out to users in the coming weeks.

Prepping For OpenOffice 3.1

There are always multiple projects moving ahead concurrently. OpenOffice 3.1 is almost done, so this afternoon I have been prepping the dialog the users will see the first time they launch the newest release. Users won't navigate to web pages, won't read the release notes, and won't read a long message describing everything that has changed. This GUI just gives them a short overview, and hopefully a few people will scroll down and see the new features. It only runs once, and then puts a dot file in $HOME.

We also encourage users to run OpenOffice at home, so we give them a checkbox to request media for the latest version. We will then burn them a copy of the software for Windows and Mac and interoffice it to them. A shot of the dialog is below.



On another note, this blog concerning OpenDocument support in SP2 of Microsoft Office 2007 is a huge win for all of us. As organizations begin to upgrade, we will not longer have to train users to store in .odt here, and then always email as .doc to outside vendors and contacts. My thanks to everyone that worked hard to pressure Microsoft to add this feature. This issue seems to have been driven mostly outside the United States; so my thanks to everyone around the world.

Dear Adobe, Please Stabilize Flash

I continue to test Firefox 3.1 with Flash 10 and have found it to *extremely* unstable:

Program received signal SIGSEGV, Segmentation fault.
0xae6b610d in ?? () from /u/firefox_3100b3/firefox/plugins/libflashplayer.so
(gdb)

Frame rates and sound both seem better than 9, but it crashes constantly during testing. Right now there is no way I could deploy the most recent release.

Adobe: It's great that flash is so widely used, people must like using it for content development. But with this much content on the web, you always need to provide a quality player too. I understand that many companies have made cuts in staff, and I would be happy to help debug it for you. Please think about deploying bugzilla for Acrobat Reader and the Flash Player. Please, help us help you!

PS: Already tested swfdec, it's no where near ready for enterprise deployment.

Update: Having some better luck with FF 3.5B4, not crashing as much as B3; but still crashing more than it should.

Update2: Lots of ideas and things to try from the comments. I'm now testing older builds of Flash 10 to see if they provide better stability (current is r22, I have r15 here too). I'm also going to try using Firefox 3.0.X builds and see if that works better too. I had been using the 3.1/3.5 builds with the anticipation that it would be released the same time as this new server goes live. The video card we have is ATI 9250. I'll test over other video cards, and also test with a new version of pulseD.

Update3: OpenSuse 11.1 seems to have some GTK issues that are making the testing process difficult for me. Current plan is to blow the server and load with Fedora Core 10 and see if that works better. OpenSuse/Stable seems to still have some file manager bugs, with the GUI constantly bouncing around the screen and resizing.

Flash On Flash With Pulse

I think that I have gotten Flash working with pulse now to our thin clients. Here are my work notes in case this helps anyone.

From our old system, this was what the daemon reported when playing Youtube videos:

I: client.c: created 27 "Native client (TCP/IP client from 128.222.233.233:53650)"
I: client.c: client 27 changed name from "Native client (TCP/IP client from 128.222.233.233:53650)" to "flashplayer"
I: module-volume-restore.c: Restoring volume for
I: module-volume-restore.c: Restoring sink for
I: sink-input.c: created 20 "playback" on oss_output.dsp with sample spec s16le 2ch 44100Hz
D: memblock.c: Memory block too large for pool: 1048576 > 16376
D: memblockq.c: memblockq requested: maxlength=132300, tlength=88200, base=4, prebuf=87318, minreq=882
D: memblockq.c: memblockq sanitized: maxlength=132300, tlength=88200, base=4, prebuf=87320, minreq=880

Here is the same page on OpenSuse 11.1

I: client.c: created 31 "Native client (TCP/IP client from 128.222.233.236:38065)"
I: client.c: client 31 changed name from "Native client (TCP/IP client from 128.222.233.236:38065)" to "flashplayer"
I: module-volume-restore.c: Restoring volume for
I: module-volume-restore.c: Restoring sink for
I: sink-input.c: created 22 "playback" on oss_output.dsp with sample spec s16le 2ch 44100Hz
D: memblock.c: Memory block too large for pool: 1048576 > 16376
D: memblockq.c: memblockq requested: maxlength=4194304, tlength=44100, base=4, prebuf=44100, minreq=8820
D: memblockq.c: memblockq sanitized: maxlength=4194304, tlength=44100, base=4, prebuf=44100, minreq=1
D: memblock.c: Pool full
D: memblock.c: Pool full
D: memblock.c: Pool full
D: memblock.c: Pool full
D: memblock.c: Pool full

The sound would play for a short while, and then get choppy. The browser would start to gray out and eventually freeze.

I then started looking around and found that many people were mentioning that libflashplayer was no longer needed and to use the Alsa-Pulse "device" instead to increase browser stability. I straced aplay and could see that it was looking for /etc/asound.conf which it was not finding. I looked in the directory and noticed that asound-pulse.conf is what was loaded instead. I soft-linked it into asound.conf and aplay -Dpulse started to work as expected. I then removed libflashplayer.so and started up Firefox and it's working *much* better now. The pulse daemon is no longer getting hammered with spewage.

Things like this really make you want to put in a little classic Queen. :)

Staging Firefox 3.1 Again

The last few weeks have certainly kept me busy. I have been prepping and testing Evolution 2.24 and OpenOffice 3.1 for deployment here at the City.

My next major project is to get the new Firefox server finished and deployed for the whole City. I got some excellent feedback from people with some parameters to tune to lower the memory footprint of Firefox. I also am testing some new code that extracts out certain settings from the users prefs.js file and then merges those changes back into a global template. This will ensure that everyone is always given a working session each time they click on the icon and can't configure it in a manner which "breaks" it (i.e, disconnecting from the proxy server or whatever).

I also have globally installed FoxTab (shot below) which gives Firefox the features in Safari 4 for displaying thumbnails in 3D of opened pages. No flame wars please - I don't know who did it first. :) But regardless, it's cool and is working well even over remote display.

I also updated to the latest version of Flash, and am now starting to poke at the issue of Firefox crashing the PULSE daemon on the thin clients. It appears that there are some settings where you can tune the priority of applications pushing sound to the daemon, and I think possibly the stock OpenSuse 11.1 settings are anticipating PULSE to be running locally on the same computer; and not being pushed over the network. I'm certain I'll find it.

Anyway, this seems like it's going to be a good upgrade and that people will like it.

Pierre-Eric Rocks 2: Evolution PDF Drop And Drag

Being a Governmental agency, everything that we do is about record retention and long term storage. Citizens can request email messages at any time and we have to provide them in a format that can opened without cost based software. In other words, if someone wants email from 10 years ago we can't just simply give them a binary backup of the Groupwise 4.1 post office and expect them to get it loaded on SCO Unix and working. For ease of retention, we keep email messages in PDF format in project folders. More so in Evolution 2.6, but even in Evolution 2.24 there are a few steps and a requirement that users know how to navigate to the right folders in order to write out PDFs.

In order to make this easier, I starting poking at some ideas. I noted that currently Evolution drops and drags messages in mbox format only. I had some IRC conversations on #evolution with mbarnes, mcrha, chen and xkahn and created a scope of work and placed it on BGO here. Pierre-Eric (pepp_) was interested in the project and we agreed on the scope of work and we allocated some bounty money. An initial patch has landed. Default behavior of course will be to continue to write out mbox files, but with a gconf key change Evolution will now write out PDFs into Nautilus from a drop and drag.

I was only able to fund the areas required for our needs. I'm hopeful that in the future this functionality is expanded and moved into the UI of the Preferences screen.

This one small change will save our users countless hours of time (and frustration) in the future.

Obligatory screenshot

Having Problems With Acrobat Reader 9.1?

If you have started using Acrobat Reader 9.1 on Linux, there is a nasty little bug that took me a while to locate. It's leaking files into /dev/shm that don't delete, and cause problems on the next startup. We were getting calls from people that when they clicked on Reader, their whole Xserver would freeze until the stuck process was removed. After running strace, I was able to see that it was attempting to use /dev/shm right before the lockup; and found the leaking files.

I placed these lines into the script that calls Reader and things are working as expected:

rm /dev/shm/sem*$USER 2> /dev/null