I read an e-mail this morning stating that we need to ship a build today that corrects problems found during internal test. So I scanned the list of problems in the test summary report. One of the bugs was in the application that I am responsible for. The trouble ticket generated stated that the testers got an error message when they ran a specific report. This error was a REP-0736, which means “There exist uncompiled Program units.”
The tester who reported this problem had not come in yet. The developer who normally takes care of reports had not come in yet. So I decided to get busy on this problem. The first thing I did was fully uninstall the version of the application that I had on my development machine. Then I installed the version that the testers had installed. Unfortunately I was unable to duplicate the problem. I also tried running all the reports in the application. I made sure that I was logging into the same database that the testers use. However I could not replicate this bug.
All morning I kept an eye on the door. Then I saw the tester who reported the bug come in. I gave her a few minutes to get ready for work. I then talked with her about the problem. This is when I received some key facts about the problem. Two of the testers encountered the problem. Two other testers could not reproduce the problem. I had her show me the problem in the application while running on her machine. I proceeded to log into the application myself on her machine and got the error. Finally I tried some of the other reports the applications runs. None of them worked either.
So I reviewed the facts. Not all testers get the problem. I do not get the problem on my machine. But I do get the problem on the tester machine. Both the tester and I decided there must be a configuration problem on the machine where the problem occurs. This theory was given more credibility when the tester confessed that she installed the Reports Server software on both machines where the problem was occurring. The next step was to get the tester to reinstall the Report Server using the same instructions that I used when installing it successfully on my machine.
We are already doing the build for the release of all fixes to test today. So I am hoping that this problem does not require a code change. I do not think it does. If it turns out the correct reinstallation of the Report Server from Oracle fixes the problem, I say it will be time for me to go home as I will have earned my pay for the day. When the reports developer came in, I touched base with her to let her know what I found out and what the tester was trying to resolve the problem. I was a bit disturbed when the reports developer told me they knew about the problem and was hoping a DBA would go reinstall the Oracle client to see if this had any bearing on the problem. This route is akin to asking somebody to reboot hoping that this will fix the problem. Yes you may get lucky once in a while. But it is no way to do organized and logical debugging of problems.
A Little Bit of Crypto - I have been trying to figure out to "collision resistant" some of these standard hash functions are. It is a tough concept to get my head around. I figure...