Testing Evolution on Multiple Distros

Lets see: the big hand is on the 1, and the little hand is on the 50 so... Well I'll be. Welcome to post number 150 at talk.bmc.com... not counting the 17 or so I currently have on my personal blog over at on-being-open.blogspot.com of course. "Adventures" kicked off in September of 2005, so clearly there has been no issue finding things to talk about every week here. :)

I have some updates on where we are at with the Linux NAS server that I'll get posted soon, but I wanted to put a diagram with it, and so that means taking the time to draw them with Dia, and that just has not happened yet. The other topic I have been working on here lately is Evolution 2.10.3 against MS Exchange 2003.

Multi-Distro Evolution testing

In my sequential testing, first Mint, then PCLOS, now Fedora 7, all on the Dell D620, I have seen what looks to me to be the same problem, repeating over and over: An instability in the Exchange Connector that requires totally recycling Evolution to recover from.

I have blathered on and on here about Open Source and Community, and people doing what they can. I have finally decided that this is a case of needed to step up and do what I can. I am not a C language coder. Not beyond 'Hello World" anyway. I have written in many languages over the nearly 30 years I have been in IT, but that is not one of them. Truth be told, as much as I like to think of myself as a hacker, I am more of a hacker-arounder. (tm). Being a manager, even a technical manager, for over 20 years and not in the coding trenches means I am not going to be able to fix the problem(s) in Connector myself. Well... I *could*, I just first would have to come up to speed on C++. Then the Connector project, and the WebDAV protocol. Might take a while.

What I can do is try and get the diagnostic information to those who can fix it. I am technical enough to be able to install and configure Linux, and follow instructions for needed diagnostic information. I clearly can find problems with regularity!

My working theories about the Evolution Connector problems are currently:

• Evolution Connector does not get as many development cycles as it needs,
• Perhaps because it is not as easy / interesting to work on as making Evo work against Open Source mail servers or Groupwise.
• Perhaps because the MS Exchange server environment is hard to come by for the currently active Evolution developers.
• Neither the KDE nor Gnome environment makes a difference in the stability of Connector.
• The Connector stability issue is not related to any particular Distro.

The first assumption is not testable. I base it on the fact that most Open Source stuff with problems of this type get fixed far faster than this one. My general observation is that continuing problems in a given area of open source maps to a lack of critical mass. Not enough eyes are looking at it. I am totally guessing. This could also just be a real snakey , twisted problem. Or both.

Last post I was starting to feel like the latter theory was not valid because Fedora 7 had been rock solid stable on Evo Connector. It has been better for whatever reason. It went way over a week before Connector started crashing. When it started crashing, it did not happen nearly as often as Mint or PCLOS did.

It does crash, and when it crashes, it looks like the same problem, at least from the outside looking in.

One advantage of working in R&D Support is that old hardware goes into bone piles, and I can go dumpster diving. I'm not proud. Linux doesn't need all that much hardware to run on either. I am also technical enough to fix broken hardware. To create a way to test my theories, as well as report the issues as they occur, I needed to be able to have several Distros running at the same time, rather than serially on my D620. The dumpster yielded some new victims. It would be optimal if the hardware all matched of course, but that is not possible. I did, for space in my office reasons go with all old laptops. I have to work in here in the daytime!

Combo Burrito

Here is the current lineup:

Dell C400 running Fedora 7: I wrote about this computer a while back on my personal blog, when I was comparing Mint and Fedora. Fedora is happy here, and supports all the hardware out of the box, so no need to rock the boat. 1.2 Ghz Processor, 1 GB RAM, so pretty decent. I have an open bug on Fedora, so this one is waiting should I need to do any additional debugging.

Dell Inspiron 8100 running Mint 3.1: With a 1.2 Ghz processor, 512 MB RAM, and a dim but otherwise lovely 1600x1200 screen, Mint 3.1 (brand spanking new...) installed late one night on the 8100 without issues. With a dedicated graphics card from Nvidia, I was hoping Beryl or Compiz would work, but there is not enough memory on the card to deal with the resolution and the 3d compositing. Nope. No big deal. I was mostly curious. The keyboard on the 8100 is very clunky and metallic sounding. So far so good. No Evolution failures.

Dell Latitude C610 running PCLOS: I liked PCLOS on the D620, and VMware was easy to get going on it. The ACPI reporting oddities I was seeing with PCLOS on the D620 are not an issue on the C610 either. The screen is bright but only 1024 by 768, same as the Dell C400. RAM is smallish at 256MB, but the KDE memory mapping tool shows well over half of it is in use as a disk cache, so it is plenty. The processor is 1.8 Ghz, and the graphics card is a separate Radeon, so at 1024x768 Beryl works well. Keyboard is much better than the 8100's, and this unit (also like the C400) has a TrueMobile Wifi chip, which PCLOS found and configured without issue. Evolution has not crashed here yet (but I already have one PCLOS crash collected from when it was on the D620).

IBM T41: Right now the T41 is running Mint 3.0, but this one I have set aside for Mandriva

Dell Desktop running Mint 3.0: This is where I run many of my central processes, and I don't upgrade this system very often. I do not expect that it will be much different from Mint 3.1 here, because the Kernel and Evolution release levels are the same.

Dell D620 running OpenSUSE 10.3: Fedora 7 has been replaced by OpenSUSE 10.3. Despite running on the newest hardware of the lot, and the D620 having run Mint, Ubuntu, PCLinuxOS, and Fedora 7 with mostly success, OpenSUSE was a pain to get running smoothly. Most of it revolved around the graphics hardware being mis-identified. After some brute force, it was running at 1280x800. KDE's default fonts look pretty bad, so I have Gnome running there for now. Then I had to get rid of SLAB. But now it is up and running, and Evolution has debugging symbols on *all* the modules, so I am ready to fail... Not that I want to. But that is the goal here.

In reserve are my Acer 5610 and IBM X30, should anything act up. Both are running Mint right now, and I am hoping they don't get called into service.

Next Steps

Getting Mandriva 2007 into the mix is my goal for next week. Week after that I fly to Pune India for a week, so this project will pause for a bit then.

In each case, I am running as close to the 2.6.22 kernel as I can, and the 2.10.3 version of Evolution as is available on each Distro. Obviously across distros I can not even come close to controlling for all the variables here. My theory is that the same Connector failure will occur regardless of all these variables. The responses so far on Bugzilla my two failures now marked as a dup of 458322, so I may be on the right track. I'll read through the details on each next week and see what I need to do to debug further. I am loading up all available debugging packages and Gnome's bug-buddy. I saw that one comment said something about also wanting debugging symbols for things like Pango, so I'll have to sort that all out.

This post, like the NAS post before it, is about a project in flight, rather than one where it already done and all the facts are in, conclusions reached, and doc written up. I hope to learn along the way of course, but also just practice what I preach about being open. Besides, I have to have something to post in entry 151! :)