This week really, really sucked. Where to begin?
It actually started in a good way. I did the 50-mile ride last weekend with Elise and Terry and felt great at the end. But I was a little dehydrated and I really strained my back. By the end of Monday, I was feeling like I was sunburned on the inside and my lower back started spasming. I’ve had that happen before, and it can get really bad. The worst time, I couldn’t stand up straight for two days. It was really horrible.
But that passed, and I took it easy for a couple of days. I’ve taken my Revive for lunchtime spins through downtown San Antonio, but that’s about it.
Work was very, very hard. We’ve had a lot of layoffs this week. Nobody seems to know why, which gives everyone a sense of paranoia. It did not help that there are several very nasty worms running around the Internet which are bringing all sorts of systems to a halt. We had to replace our mail system this week, and the new system kept dropping messages and rejecting logins, which in the light of the layoffs was extremely nerve-wracking. I can’t log in for my mail – what, was my account terminated? AAAGGH!
On top of that, my regression tests failed me, and I managed to deploy at least three show-stopping bugs this week. That was horribly embarassing.
I’ve never really told you much about my job. Currently, I’m the QA person for a software project that is used to track information about my company’s clients and the systems that they lease from us. Whenever one of the other programmers has new code to put into the live system, I first test it out to make sure it’s not going to introduce new bugs in the system. This is what they call regression testing.
Our software project uses a web-based interface, so testing it consists of loading up the site in the browsers that we support, and making sure that all the pages come up correctly. Simple, no? Well, it would be, except that there are thousands of pages in the application. If I hit them all manually, it would take me days to roll out even a simple set of changes. So I use a tool that lets me record a session in my web browser and then play it back later. I can mark up the resulting test session to include checks for specific text in the web pages that the application sends back to me. This is all well and good, but I fell prey to a very simple mistake this week.
I was only checking for error messages in the major pages. We’ve got a single page that displays a condensed view of an account, for example. That’s an important page, so I marked up my test session to report an error if that page contained the text “ERROR HAS OCCURRED”. Well, the problem is that one of the change sets I rolled out this week had an error in it which caused an error in the menu system for that page. And as far as a web browser is concerned, the menus are a different page, even though they show up on the same screen as the account view.
To sum up, my tests were inadequate, and as a result, I broke a very important system three times this week at the same time that a major worm attack was slowing down the network and people were being told to clean out their desks.
Y’all excuse me while I leave the county for a couple of days.