The Goldfish Chronicles

Noticing that AtlasCamp is on in a couple of weeks (blog announcement). Typical Aussies have found somewhere to surf while they’re up in the US

JIRA 4.0 also appears to be out. Looking at the June Atlassian Summit presentation, it looks like the major topics are JQL (do more interesting searches), Gadgets (aka AJAX based portlets by the look of it and the 4.0 install I did the other week), Actions from the Issue Navigator (hopefully pluggable actions too, and a new Browse Project HUD that may or may not be anything to write home about.

I’m looking forward to both using and updating my plugins for the new Gadget world. I also keep meaning to write a pair of Issue actions to move a project up or down a version.

There’s probably a better way, but thought I’d share a quick hacked up script I created to take an SVN changelog and filter out revisions that I didn’t want there. I used it while merging Commons Collections generics branch into trunk to create specific submit messages showing the bugs fixed on each file in its commit message. I filtered out the common and not interesting revisions for fixing tabs, or my shifting of the license header so the merging went easier.

# ARGV contains a list of filters to ignore
# stdin contains a file

svn_separator = "------------------------------------------------------------------------n";

entries = $stdin.read.split(svn_separator);

ARGV.each do |revision|
  match = "^r#{revision}"
  entries = entries.select do |entry|
    entry !~ /#{revision}/
  end
end

if(entries.length > 1) then
  puts entries.join(svn_separator);
  puts svn_separator
end

JIRA 3.13 is out and so I’ve spent an evening fixing my FilterList plugin due to a change in API. That’s happily doneand a 1.1.1 release is out there.

It’s because of a nice new feature called favourites whereby users can signify the filter they like by clicking on a star, rather than have a billion filters rammed down their throat. My filterlist plugin allows similar, but as always it’s a bit of paint on the top while the new change is at the core of things.

More importantly imo is the addition of favourites to dashboards. You can now go searching for others dashboards (if they’ve shared) and select it as a favourite so it appears in your screen.  It’s a bit iffy in that when you start managing your dashboard you get a copy of the default and not just the default as a favourite - bit of a legacy user experience I suspect and I still need to test whether you can get back on the default dashboard whilst still also hooking up to another dashboard (say a product release dashboard etc).

Atlassian also extended their personal license to the JIRA space. At last. We used JIRA to move from KY to WA and it was a happy experience (it was in a private corner of the osjava JIRA… don’t tell Mike et al). Now we can use JIRA to fix up the new house and have such components as ‘Bathroom (Our)’ and release versions of ‘2008-09-14′ - ie) a release each week. It also gives me somewhere to test my plugins, which was getting painful on the 30 day demos.

I started to put both Confluence and JIRA on the machine, but it’s OS X and that means installing a database which is a pain in the arse on OS X. So just JIRA for now on HSQL. Random request to the Atlassian guys - give me a JIRA/Confluence personal license download which integrates the two in a single Tomcat/HSQL. I’m sure it’s not hard to do myself… just lazy

Lastly - there’s an AtlasCamp in November. I’m bummed that I can’t go, but it clashes with ApacheCon and I was already signed up (which thankfully means I don’t have to choose between the old community I belong to, and the new community I like to code in).

I’m finally getting near the end of my every now and then hacking at a static version of a JIRA.

The basic story is that I want to turn off the computer that is running in my attic doing nothing but hosting a JIRA. The slicehost I pay for isn’t a great place to try and squeeze in the memory hungry JIRA/JVM combination, and as the project has mostly moved to Google Code it seemed like a good idea to get the JIRA gone too. Google Code doesn’t have any simple methods for uploading the data (basically you get to do HTTP and have no control over dates etc), so I decided to dump JIRA as XML, and XSLT a nice look to it. Maybe one day I can migrate it to Google Code or elsewhere, but I doubt it matters.

Pulling the XML down is easy - in fact I don’t even implement that. I assume you have used the IssueNavigator in JIRA to get a dump of all your issues as RSS. That’s right - probably will make a big JIRA explode. Ah well. I also pull down attachments for you. Just not the audit history.

Then I use XSLT (stunning architecture eh?) to put together a simple static site that shows users, issues, projects and versions (couldn’t be bothered with components though I should add them some day). The nice part here is that as I have the xml, I can restyle and make it better if I ever want to. It will support the project and issue URLs out of the box, while version and user urls will require a mod_rewrite rule, however no one goes directly to those links anyway.

Here’s the site:

http://www.osjava.org/issues/

I need to redo it from the latest xml dump, and then maybe it will be done and that light hum from above can go quiet.

If you want the code to do this (you’ll have to modify it yourself as it’s got little hacks in for osjava specific bits) it’s public domain and available at:

https://svn.atlassian.com/svn/public/contrib/jira/jira-outlet/trunk/mig4jira/static

Just realized that I should have pointed out that James put out a Quartz 1.6.1 release candidate. I got involved with Quartz while at SourceLabs and have since kept a bit involved.

Anyway, consider yourself informed

If you’re using Ruby 1.8.6, try the following:

irb
require 'rexml/document'
REXML::Document.new("<foo x='&amp;'/>").to_s

In 1.8.5 I get an answer of "<foo x='&amp;'/>", in 1.8.6 it is "<foo x='&amp;amp;'/>, ie) entities are doubly escaped when they shouldn’t be.

About 10 days ago the GoDaddy server we use to host our websites ($30 being cheaper than the $100 we used to pay and the $200+ that colocation would be around here) decided that they weren’t going to run Tomcat webapps anymore. They’ve always been crappy, but this time they took it to a new level with the Tomcat regularly falling over with an OutOfMemoryError before it’s had a chance to get comfortable.

Almost definitely GoDaddy’s fault btw rather than the JVM, Tomcat or Roller/JIRA - whatever Xen/VMWare style thing they run does not like Java. Running ‘java -version’ on the command line will get a not enough object heap error and when I tried to use a standard Tomcat install it could never get any memory. In the end I chose to use the one that GoDaddy supplied and it stood up somewhat happily for 6 months, then something must have changed and things died.

So, three options jump to mind. Number 1, complain to GoDaddy. I’ve read various complaints online concerning Java on GoDaddy machines and am really not in the mood for such things. I ought to send them an email, but haven’t yet. Number 2, go back to colocation. I’ve got two 1U servers sitting upstairs unused, but colocation is a pain. It means physically installing the machines, and as we live near downtown, that either means paying a lot or driving further away from downtown to find a good price. That means taking a morning or two off. Sucky. Number 3, ditch Java on the server. I was running a couple of Rollers and a JIRA. Migrating the JIRA to Code.Google has been a todo for a while, so I’ll just do that (at some point), which leaves the Rollers.

We’ve enjoyed using Roller, hopefully we will again. It has features that we’ll definitely miss; and I definitely feel safer using products that I’m confident are going to remain open source. We were using Moveable Type before moving to Roller, and the one I’ve chosen, WordPress, seems to be accused of a similar direction (Google for ‘wordpress security’).

Migrating is fun though, and it’s more fun than hassling support to get Java running again on a low-memory virtual system, or doing physical machine migration. So first up was choosing new software. Moveable Type was out - they haven’t released their return to open source version yet, and even if they did there’s a lack of trust issue to deal with there. Other Java blogs are out too. As are the free hosted services, we’re used to more control. The one that Carrie suggested via happy friends was WordPress. When this later matched with the one that coders on IRC were recommending, it was time to look at it.

Most important is data migration. I took a script from somewhere (and now can’t figure out where) that you setup as a template in Roller and executed. The idea being that it would dump your blog as wordpress.xml format; however that’s only for older Rollers (probably still JRoller) so I had to rewrite it as a JSP to get it to work on the newer Roller version. If not for the script, I suspect I would be migrating to a new physical box by now. It gave me the guts that even an evening of fighting with Velocity wasn’t enough to slow me down and make me want to drive out to Bothell with a 1U.

Next up - installing the software and importing the data. WordPress itself seemed nice and shiny, until my confused self wondered where the blogs were. Seems that it’s a uni-blog. Installing 3 or 4 seemed daft, but that was seeming to be my only hope before I wandered across WordPressMu - the software that runs wordpress.com. It looks like a big hack on top of WordPress, but I gave it a shot.

It has quirks. They turned various features off in the code. I turned one back on (in-browser theme editing). They support *.foo.domain.com, we want various different domains. This lead to some late night hacking of their login code to support the cookies. That might have been a deal-breaker usually, but we’d just finished getting the l&f right, moving to the real domains and setting up the rewrites when I discovered that we could no longer log in so rolling back was the alternative. After these changes, I’m not looking forward to the first upgrade.

I liked the feature whereby I can define the permalink structure. This lets me continue to use the Roller style url and maintain existing urls. That’ll also let us migrate back someday if I get the platform.

We’re going to miss Roller’s feature whereby comments would close after N days. I’m going to have to write a cron-job that hits the database for that I think - though always possible that one of the many plugins I hear of does that. That’s a negative of WPMu, not all plugins work.

I’ve a suspicion that spam will continue to be a problem. It’ll be interesting to see whether it’s better or worse than Roller.

License is of course an irritation. Less so for the server, but even the default theme is GPL. That sucks mightily as it might infect the themes we’re migrating as the default is the obvious example to base your new theme on. Fortunately I doubt very much that we would pull anything of value out of the theme when migrating to another server someday, instead we’ll continue to take the part we created, the theme customisations themselves.

I’m not sure how this’ll work out. It’s been an irritation to have to do this, but fun to play with something new(ish). Now I just need to avoid the other Roller committers lynching me

Tim Ellison with Geir and his flag at JavaOne via my admittedly crappy phonecam.

So the big question from Apache’s point of view with regard to JavaOne is whether we’re being listened to concerning the inability to properly license Harmony as Java(tm). Geir guessed right with his first guess, as far as I can interpret the statement was that they would sort out the TCK for open source Java (in the context of OpenJDK) via a series of steps etc etc usw usw.

I don’t think that Rich Green announced anything about it being open source per se, just that it would be useable by open source. Not that he defined whether ‘open source’ is Sun’s projects or the community.

New versions of the Current User Status plugin (1.0.1) and the Release Status plugin (2.3) were released tonight. Both were bugfix releases based on user feedback. I know, it should have been 2.2.1.

I also released the Improved HTML plugin. It’s not very snazzy, but it was something I needed as the default JIRA ones didn’t quite do what I wanted.

Which was to have a message for anonymous users that didn’t get in the face of logged in users on the Struts JIRA. Yes - all 6 of my plugins (8 portlets in total) are now deployed into the Struts JIRA install for your bug viewing pleasure. If that continues to go well, my next aim will be to replace the main ASF JIRA with something very similar to the Struts one.