Spent most of the day working on the XML Schema import feature. Log type creation by XML Schema is happening. It's ugly and it's flat, but it's happening. I created a container field type near the end of the day and will tomorrow begin work on supporting complex element types. I may do a screencast to show where I'm at before I do that though.
Very cool stuff to be working on.
This evening, from about 5:00PM until 10:45PM, I worked on various fixes that people have requested.
For Phil in Australia I worked on the trackback admin interface. Evil spammers were hitting him with trackbacks that had open html tags. This screwed up the admin interface so I just replaced all left brackets with <, which effectively kills that method. Now Phil should be able to delete these ugly trackbacks. If not, I can always delete at the database level for him.
Also worked on a timezone issue for Phil. He sets the site and user timezones to "Australia/Sydney." But a bug crept in recently that messed up his timezones. When he was posting it was actually recording in GMT or EST time. Fixed that. And went a little further to refactor a key class called Entry.java. It was holding too many objects in memory... too many things to prevent it from being robust. It was one of the first objects I wrote when I converted the system from MS to Java. It's still not perfect, but it's better. Actually, I should say that it's getting better... I won't conclude it's better until it's been running in production for a while and thousands of people hit it without incident.
Craig requested some work on the homepage image and comment counts. He was right... the caching mechanism was freezing the file counts so that when you added a file to an entry it wasn't immediately updating the count. It's working on the test servers and should be fixed in production now.
Joel found another fallout from the caching mechanism. He had an entry mysteriously dissappear from the homepage. Not good. I've been seeing some null pointer exceptions in the error log that are caused by the caching mechanism, OSCache, that I'm using. I posted to their forum early in the day but after only three reads I haven't gotten any response yet. I believe the errors are not a huge issue, but I'd like to have cleaner code without the try/catches that I'm using right now. Apparently they're not synchronizing an empty cache properly in their code. So I don't have an authoritative answer for Joel, but I'm looking into it.
My father reported an issue with the New Log page. I saw the error on the error log, but wasn't able to duplicate the issue. Worksforme is a horrible way to answer a bug submission but I may need to get his help to duplicate it. If I can't duplicate it I can't fix it.
The SMTP listener is sometimes failing to bind to port 25 of the web server. It sends out an email message each time it fails... about once every minute or so. But that email message is going to a hard-to-configure-by-admin email address. So, two issues. First, I need to make sure it can bind to the port... generally this is simply a Tomcat restart. Second, I need to make the email address more easily configured.
Heather, my lovely wife, is having some trouble uploading pics with her camera phone. The picture is uploaded properly and attaches to the entry, but the thumbnail fails to build so she ends up with those ugly default File icons. I've seen a similar issue on Anna's site, although with a BMP. Generally bmps will have thumbnails created. So something's wrong, but I'm not sure what.
Stewbeaf was having some trouble with the Static Pages page. Unable to find a Save button. I see it on the test servers and on production accounts, but this could be because I have super-admin privileges. Could be a bug. I'll look into it more closely soon.
Those are some of the critical issues happening right now. I still have an older message from Craig with a few issues to get to. And a few other things hanging back there in the email inbox. Hoping to get to them soon.
Oh, and about 500 issues in the more formal issue tracking system. Yikes!
Thanks to everybody for the patience. And big thanks for the feedback! Keep it coming!
Update 11:12PM: Just got a response from the OSCache forum. It's a long one:
This is fixed in cvs.
Which is good news. CVS stands for Concurrent Versioning System. Basically, it holds the source code for projects. When they say it's been fixed in cvs they mean that the bug has been fixed but it hasn't been released yet. So I'll reply and try to figure out when the next release is.
We use Subversion (SVN) for all of Reger.com's stuff. A little newer but generally accepted as the CVS-killer. Some growing pains for it and some robustness to grow into, but it suits our needs perfectly. And the integration that IntelliJ gave to Idea 5.0 is glorious.
Update 11:17PM: Now I'm seeing the homepage dissappear thing that Joel mentioned. Happened on this post, in fact. Gotta look into it. Not good.
Update 11:22PM: Aw, crap. The issue isn't with the cache. It has to do with the month value. For some reason it's decrementing it by four when you create an entry. And then when you correct it it's off by one. Damn. Gotta fix this.
Update 11:25PM: I put up a system alert message for all users of reger.com so that they're aware of the issue. Good usage of that system alert feature! Bad code.
Update 3:20AM: Just about fixed but not likely to launch until later this morning.
Phil says via email: Gee this problem solving thing must be fun, kinda like trying to hit gophers as they come out of their holes. Insert image of you as Bill Murray in Caddyshack with golf club in hand.
Update: Ok, date stuff is fixed. Also re-worked some other back-end pieces that were causing some errors... nothing critical though. And while I did get the image counts working on the homepage for Craig, I missed the message count updates of the cache. Small change but needs to be done. Phil pointed out the same thing last night.
Update: The guy from OSCache says:
I have been focused on a major rewrite and just don't have the motivation to polish this release enough to get it out the door. If someone wanted to contribute some time to eliminate the remaining problems, it could probably be done fairly quickly.
This is the good and the bad of the open source community. It's great that the guy actually doing the coding can reply directly to twice within a twelve hour period. But he's one person and likely has other responsibilities. Even larger open source projects suffer from this issue as people quasi-own sections of code. It's only when projects get big enough for multiple people to understand each code segment that open source sheds this single-threaded issue. Commercial software is often just as problematic, as I'm proving with this post.
So it looks like we'll have to wait a little bit for the cache bug fixes. That said, I error trap the heck out of them so they never propogate to the user front-end. And when a retrieval from the cache fails it'll just try to refresh the object. It's just a little annoying on the back-end error log... and gives me that "not everything is right" feeling.
Sunday, February 7, 2010