RepairMetatda Resurfaces

If you're wondering where I've been for the past three days, I was attending YAPC::NA in Toronto. Today Mike tested the repairMetadata fix I submitted and during the test it filled up the /tmp directory. So the fix was rejected. At first I thought changing from using subtransactions to transactions, but that did not solve the problem. Then Greg suggested the problem might not be with repairMetadata itself, but with the folder copy that was part of the test. Running repairMetadata directly showed that indeed that was the case. So technically the fix is "correct", but it does not solve the problem it was supposed to. I'll have to step back and think of the best approach to the problem, maybe increasing the size of /tmp or looking at the copy code to see if there's any obvious way to cut down on the /tmp disk space used.

I'll be on vacation up through July 11th, see you July 12th.

Badlinks

My bad links checker script was throwing a segmentation error. The problem seems to be that I set a timeout on parsing an individual page and the asynchronous exit the timeout caused left the script in an invalid state. So I removed the timeout, but now face the problem the timeout wasmeant to cure: html pages that take a very long time to parse and hang the script.

This and That

I moved the change to repairMetadata into CVS and wrote up a test plan for it. I tested Leigh's conference registration form. It passed. Then I reworked Mike's metrics scripts to support the POSKey testing.

PAR

I was in late because of a doctor's appointment. I spent the whole day working on my self-asessment.

Transactions

I modified repairMetadata in STCatalogAware.py so that it uses subtransactions and then tested the change. I still need to install the change in cvs and write up the test plan. We tried once more to install the Palm desktop software on the Macintosh and failed again. I'll have to contact the folks at Alphasmart. I answered a user question on how to set up cvs (CNSHD533045).

Palm Phonebook

I pulled the problem report on the LDAP to palm download out of the stack of problems i've been determinedly ignoring. LDAP to PC palm works fine. LDAP to outlook doesn't work and LDAP to Mac Palm is undetermined: I had trouble installing the Palm software on my Mac.

Check It Out

I looked at repairMetadata and decided there was no bug, though it would be nice to be able to switch off recursion. I checked up on my YAPC information to make sure I have all my ducks in a row. There's bus service between the airport and conference center, so I can afford the cost of a taxi. Hooray! I talked with Mike about what turned out to be a rewrite problem. Dan will handle the fix.

Back to the Grind

I'm back at work after a week and a half vacation. I spent a lot of time going through old email. We met and discussed Anne's memo on Zope risk assessment as well as meeting and talking about the proposed poskey error. Greg pointed out a possible problem with repairMetadata that I will have to investigate.

More Poskey

Craig reported my permissions fix didn't pass the test. I checked and it had not been checked out of CVS. We'll need to discuss and document our CVS procedures. I worked with Greg some more on the Poskey problem. We outlined our proposed solution and then started working on a new, more conservative solution.

A Problem solved

I worked with Greg today on two problems. The first was configuring the development and test versions on remedy, which was pretty straightforward. The second was figuring out a way to fix the last stubborn poskey error. We finally hit on a method that copies the contents of the object to a new object, clears the object and copies it back. So now we're ready to fix the production database.