Older blog entries for klevin (starting at number 124)

Tracking down issues with the linux code in the hbaapi library that were introduced by the rather major re-write that moved all of the data out of globals and into per adapter structures. Next on the list would be to add some locking, but that can probably wait a day or so.

Today seems to be CM Gopher day for me. Fixed a problem w/ some code from one project getting committed to another project (the developer had copied a directory over from an older project to use as a template and got all the CVS files as well). The last time Rajesh committed code to the hbaapi project he managed to wipe out every change I had previously committed. I'm really not fond of working w/ engineer types when they're using something "new" that they refuse to take the time to understand.

--

Fixed a bug in the new code that handles the "fixed in" field. Ahem. It helps if you actually submit the changes to the database. While running that down, I discovered that the "change multiple bugs at once" functionality of buglist.cgi causes process_bug.cgi to only log the changes to the first bug and not the rest. Loverly. Trying to find an elegant way of fixing it.

Absolutely glorious day so far: cloudy, cool and a light mist. This is my idea of good weather. I've liked cool, damp and foggy days for as long as I can remember. We used to get some awesome fogs off the ocean when I was a kid in California (Santa Maria) and when my family moved to Washington, I just mossed over and became a confirmed cloudy weather person. Living in Kansas has been a real struggle, as I get grumpy when it's sunny for more than a couple of days in a row. My ideal home would be on a rocky cliff jutting out into the ocean somewhere on the northern coast of Washington. Man, I'm really home-sick today. Ah well, I get to go visit for a couple of weeks, starting the 20th.

--

Cleanup time for the changes to Bugzilla. Turns out it's only necessary to splice "fix_ver" into the collist array if the user has a non-default collist cookie set (such as me). Otherwise you end up w/ two fix_ver columns in the bug listing. Cleaned up a few other bugs and added more versions to the fix_vers table for the Fc929 FW project.

Finally wrapped my mind around buglist.cgi. At one point I had the new field displaying (w/ title and everything), but the summary field was empty. Turns out I had to push the fix_ver field onto the front of the collist array (`splice(@collist, 0, 0, "fix_ver")'), as the DefCol() function wasn't setting things up properly. Anyway, it appears to be working.

Bugzilla time: Mucking about w/ adding a field to the "bugs" table to track what official release the fix went into. Everything was just fine until I tried to incorporate the new field into buglist.cgi. As is usual, buglist.cgi is an incomprehensible, steaming pile of ugliness and nothing behaves the way you would think. Not much problem at all to integrate it into the query page, but buglist? Hah, forget it.

Mucking about w/ version data in Bugzilla. Happy day.

I now appear to be, unofficially, the group's research gopher. This is both good and bad. I like doing research, but it can get old after a while of nothing else. Also, that puts me in the position of being the bearer of bad news. Not a good thing around here.

Got a few requests for enhancements to Bugzilla. This could be amusing.

--

Also in my "requests for help" inbox: testing and design issues w/ using the new scsi error handling in the mptscsih code, FC management app UI coding (java app written w/ IBM's VisualAge) and probably a few other things. Here I go and say that I have no clue what I'm supposed to be doing at the moment, and a zillion people volunteer to help me out w/ the matter. And none of them, to my knowledge, read this page. Heck, nobody does except me (it's a useful way to sort out and remember what it is that I'm doing). Shout into the void and listen to the roaring silence.

Other problem w/ new computer: graphics card is a Radeon 8500, but XFree86 support for the 7500/8500 boards isn't complete (i.e. no DRI). I suppose I should have checked that beforehand.

Ok. It looks like the SNMP <-> CIM interface ihas been ruled out by higherups. That leaves me doing, well, no one is quite sure yet. Time to take a look at what changes have gone into the network stack in the 2.5 kernels.

Assembled my new computer over the weekend. The hardware survived 4 1/2 hours of a Cerberus Test Control System "newburn" run before I killed it, so I'm calling it good. Named the system "suil" (Irish for "eye"), in line w/ my other system names of "watcher" and "gargoyle". I'm thinking about getting my own domain and using a Dynamic DNS service. Problem: what to call the domain; my first choice of "acies" is already taken in .com, .org, .net and all the other "available to anyone top level domains."

Got side tracked at work w/ a quest to implement FC Abort messages through the ioctl driver. Due to various factors (single threaded ioctls, userland app doesn't know the MessageContext values, the driver is heading towards another GCA test cycle "RSN", etc), it doesn't look like this is going to happen. Have to talk to Rajesh about what impact this will have on the hbaapi libraries.

Woohoo! The two delinquent packages showed up in FedEx's computers last night and on my doorstep about 1:30 today. Assuming UPS delivers the power supply and case fans today, looks like I'll be assembling a new system tonight. Have to wait for my sister, though, as she wants to help put it together (she's never mucked about in the guts of a computer before).

I seem to have bad effects on power supply fans. The fan in my old system at home has been on its way out for some time (third time), and the PS fan on my work computer started dying today. Took the system back into one of the labs and, w/ Jason's help, replaced the fan.

Research time re: doing an interface between a CIM OM and a SNMP agent/entity.