|
| Greg's diary |
| June 2004 |
|
|
||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
|
|
|
|||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| Tuesday, 1 June 2004 | Wantadilla | |
| Top of page | ||
| next day | ||
| last day |
Back to work on my program testing today, and noted how difficult it is to get back into the swing of things after a few days away. Fixed the problems I had had on Friday, so now the commands are functional. That rather caught me by surprise, so now I have to think through what to do next.
In the evening playing around more with my temperature control equipment, and got it mainly to work. Using parallel ports under FreeBSD is less than obvious: /dev/lpt0 is really only for parallel connected line printers, and other equipment, including my relay board, should use /dev/ppi0. The interface is less than obvious: all I/O is performed by ioctl calls. Still, it was relatively trivial to get it to work, and now things seem to be more or less complete. Once I sort out the mechanical issues, I should be ready to control temperatures.
| Wednesday, 2 June 2004 | Wantadilla | |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
Up earlier than intended this morning at 5:45 am: we had a long-lasting power failure. The important systems have UPSs which can handle short power failures, which are never more than 30 seconds, but if the power fails for longer than 30 seconds, it's likely to be out for 2 hours while ETSA send out somebody to look for the problem and fix it. This failure carried on for a minute, so I had to find my way out in the dark (proving the value of putting torches in well-defined places) and fired up the generator. Even so, lost battunga, which was on a UPS. I wonder what happened there. I don't think it can be the duration of the outage up to that point.
The systems in the Mike Smith Memorial Room were a different matter: the battery of the ancient APC UPS I had in there had failed long ago, and it doesn't like the dirty generator power anyway, so I lost all machines. Time for a new UPS.
In mid-morning, Dan Shearer showed up with a satellite dish mount that he had brought back from Canberra for me after the Security Symposium. While he was there, he took a look at filth, the Microsoft-based laptop that I am using for scanning, to find out why it didn't automatically mount Samba shares, and got that to work. He also managed to navigate the minefield of Microsoft printer setup and configure the colour laser printer as a remote printer. Quite impressive. For some reason, he thought I wouldn't like it to be known that I have used Microsoft, and promised not to tell anybody about it.
More work on my project, but I hadn't had enough sleep, and I was really too tired to think my way through things. As a result also missed a reception associated with the IT Council of South Australia. What a nuisance.
| Thursday, 3 June 2004 | Wantadilla | Images for 3 June 2004 |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
Continuing work on my program today, and discovered a number of omissions and bugs that kept me going all day. They're all more indicative of haste than conceptual problems; maybe I should slow down a bit. Certainly the pace I've been keeping has tired me significantly.
In the evening, more work on the fermentation control, and came to the blindingly obvious solution to my connector problems: what I needed were the kind of connectors that are used on just about every PC to connect the front panel to the motherboard:
I had plenty of them, so spent some time putting things together. The rest worked nicely with a serial cable with 25 pin connectors at each end, which I was able to connect relatively cleanly. It's sad that I have to resort to this sort of solution rather than to get standard solutions.
| Friday, 4 June 2004 | Wantadilla | Images for 4 June 2004 |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
How tired I've been lately! Slept for over 12 hours, and was still feeling tired when I was done. Decided that I probably needed to take a step back from what I have been doing and do some thinking rather than implementing. I now have some good ideas in my head, but I'll wait until Monday before implementing them.
In the afternoon over to Jeff Hooley's place, our “next door” neighbours. We've been living next door to each other (in other words, about 500 metres away) for nearly seven years, and now Jeff is moving away. He's arranging a clearing sale tomorrow, and Yvonne thought we could sell our tractor, the one that she was so insistent about buying a few years back. Took the tractor over there and noted that this was the very first time I had been to this house. Jim Couch had named a reserve price for the tractor, but it seems that that's not the way clearing sales work, and in any case the reserve was far too high (but realistic for us): I saw a much better machine there with a reserve price only half as high. Back home with the tractor, wondering what made me spend so much money on the thing.
The air conditioner is icing up again. Looks like we're going to have to change the switch. I wonder how they work. It seems reasonable to assume that the probe contains water which, when it freezes, causes enough pressure to trip the switch. That sounds prone to leakage with time.
In the evening, found the energy to do some work on the fermentation temperature controller, and mounted things in the housing of an old computer I have, a 486 DX/2, 66 MHz, 16 MB RAM, running an equally old version of FreeBSD, one of my test kernels at the very beginning of release 5 of FreeBSD:
=== root@brewer (/dev/ttyp2) /home/grog color="blue">16 -> uname -a
=== root@brewer (/dev/ttyp2) /home/grog color="blue">17 ->
It's connected to the rest of the world by wireless, which I'm beginning to appreciate more and more for this kind of connection. It's now cheaper to install a wireless card than to run cabling.
The results are more functional than pretty:
|
|||||||||||||||||||||||
This one shows the temperature probe assembly. There are no mounting holes on the probe board, so I had to mount it by its 9 pin serial connector. I had already connected to probe cables to a 25 pin connector. I wanted it inside the case, so I had to connect the flat cable to the serial port on the outside of the case (the grey cable going out through another cutout just below the probe board). I need to find some kind of plate that I can use to mount it inside the case.
|
|||||||||||||||||||||||
This shows the 12V connection to the relay board. I mounted it from the top of the cabinet, and the 12V input is from the computer power supply.
|
|||||||||||||||||||||||
This one shows the other side of the relay board with the mains power connections.
|
|||||||||||||||||||||||
A view of the back of the computer. This shows a number of things:
The lower cable goes from the parallel port back inside to the relays. It would be nice to have internal cabling, but I don't know of any parallel ports that connect to a header on the board. They're all connected directly to an external connector.
Above that is the temperature probe cable, as shown before.
Higher and to the right, the flat band cable mentioned previously.
At the top are the relay power outputs (white) and the computer power cable (black). The power supply is in at an angle because it was originally designed for a smaller case, and the internal cables are too short to allow normal mounting. Some time I must buy a proper power supply.
| Saturday, 5 June 2004 | Wantadilla | Images for 5 June 2004 |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
Somehow today was all spent with temperature control. Started by installing my new computer-controlled fermentation temperature controller in the laundry:
|
|||||||||||||||||||||||
|
|||||||||||||||||||||||
Note the position of the temperature sensors:
The external (“room”) sensor is on the side of the fridge, not an ideal place, especially when I replace the power supply, when it'll be in the exhaust area. To be relocated.
The internal (“ambient”) temperature sensor is barely visible in front of the 25 pin connector. It's fastened to one of the bars of the grille.
The wort temperature sensor is taped to the outside of the fermenter. It's covered with some bubble foil to minimize the effects of the ambient air.
The fourth sensor is intended for a second fermenter. It's hanging down in front of the light.
Spent a lot of time fine-tuning the software, which still isn't ready. I'll make it available on the web when it is, but at the moment it's not in any good condition to publish. Still, it works well. By the evening I was working on graph plotting software—how I hate gnuplot—and, with the exception of the ugly plotting, came up with a most gratifying graph:
|
|
||||||||
It looks as if I'll be able to get better than 0.2° accuracy either way.
On other temperature control fronts, still had problems with the air conditioner. Tapping the switch seems to help. It's interesting to note that my temperature control setup could equally well monitor the de-icing problems.
In the evening, Jorge de Moya came along, and brought something interesting to drink:
|
|||||||||||||||||||||||
Olivaylle has had its first commercial olive harvest, and Jorge brought us a couple of litres of unfiltered oil—really virgin.
In the evening to Warrawong, just down the road, for the Dr Deb Bennett clinic dinner. I won't go there again in a hurry. It's nice to see the potaroos and bandicoots outside the restaurant window, and that's where you'd think they would have put us, but instead we were put in another corner, very cold—another issue of temperature control!— where we had to wait over an hour for dinner. The food wasn't bad, but it also wasn't cheap. If we come back, it'll be in the summer during the week with a smaller group of people.
| Sunday, 6 June 2004 | Wantadilla | Images for 6 June 2004 |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
After yesterday's work, I intended to take it easy today, but it didn't turn out like that. The temperature control went fine overnight, so started another brew of beer. This time, mercifully, things went a lot better than last time, but it's still a lot of work, especially since I was formulating different things to change. In the meantime wrote up the description of yesterday's exploits and further played around with the software.
Having grain in the house attracts mice, and this morning Yvonne found one in the rubbish bin under the sink in the kitchen. Brought in our cat Lilac, who's a good mouser, but she couldn't find the mouse and eventually gave up. Later, after starting the mash, I opened the door under the sink and found a mouse—in the empty plastic bag for the crystal malt. It was relatively simple to grab it and close the seal:
|
|||||||||||||||||||||||
Showed it to Lilac, who to my astonishment didn't pay any attention. She obviously couldn't smell it, and it seems that she saw it move and thought “That looks like a mouse, but mice don't come in plastic bags”. I had to hold her attention while I let it out of the bag, after which she was on it in a fraction of a second:
|
|||||||||||||||||||||||
|
|||||||||||||||||||||||
|
|||||||||||||||||||||||
Put her outside to enjoy her meal, and threw the plastic bag back into the rubbish bin—almost on top of another mouse, which unfortunately escaped.
In the evening, Chris and Jorge came back with Deb Bennett, and we went out to Maximilians in Verdun for dinner. Excellent place. This is a place we will return to. Four of us ordered steaks, all blue, and this so impressed Maximilian that he came out after dinner, sat down and started quite an interesting conversation.
After that, off through dark, windy roads to find where Deb was staying, somewhere in Cherry Hill. It wasn't far, but getting there and then back home took an hour. The roads in some parts of the Hills are just too twisty.
| Monday, 7 June 2004 | Wantadilla | |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
Back to work again today, and made some progress on my program, though not as much as I want. Writing programs is something like making ice cream by hand: when you start stirring, it's easy, and you make good progress, but as things progress, it gets stiffer and stiffer, and it becomes really difficult to stir. I'm at about that stage now: every detail I need to change seems to require going back and changing lots of details. Still, I've broken the back of it, I hope, and things might get better from now on.
| Tuesday, 8 June 2004 | Wantadilla | Images for 8 June 2004 |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
Into town early this morning for a breakfast presentation at EDS, concerning “offshoring”. I was pleased to note that they presented my opinion that South Australia stood more to benefit from this phenomenon than to be threatened by it.
After that, IT Council board meeting, and then to Tellurian to pick up a UPS that Dan Shearer had left me. Then to ASIS to look for some chemicals for brewing: calcium carbonate, calcium sulphate and magnesium Sulphate. They had them, of course, but the prices were astronomical. 500 g of calcium carbonate for about $32. The last time I bought calcium carbonate, it was $12 a tonne. I'm obviously looking in the wrong place.
In the afternoon, another Rocksoft conference call, in which yet again I was given doubts about the way I'm approaching my indexing. Looks like there's even more thinking to do. sigh.
My fermentation temperature control stuff has been working really nicely, and it has also generated a fair amount of interest, to judge by the web site hits. I still have some problems with overshoot, but I also have some ideas about how to fix it.
| Wednesday, 9 June 2004 | Wantadilla | |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
Another day spent pondering how to do my indexing, and didn't get very far. Decided to continue along the method that I had thought out, though I have indications that it's suboptimal. At least completing the code this way will make it easier to understand the issues, and it'll also be useful for comparing the performance. Finished my next step (due last Friday), but didn't get round to testing it.
Looks like SCO is raising its ugly head again. I suppose they have to, given the way their stock has been performing. Heard from IDG, who asked me some questions and published some of the answers. I wonder how this thing is going to pan out.
| Thursday, 10 June 2004 | Wantadilla | |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
Spent most of today investigating the indexing status quo that we had been talking about on Tuesday, and made some progress, though after some months still don't really understand the issues. Based on what I find today, though, it looks as if I'm not the only one.
My responses to Computerworld have generated some interest. Almost immediately I heard from Detlef Borchers of C't, and today I got mentioned on a Slashdot article. That always has an effect on the web site hits, and today my SCO pages experienced a hundredfold increase in hits. People seemed particularly amused by my suggestion of sending 20 tonnes of paper to SCO.
| Friday, 11 June 2004 | Wantadilla | |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
Spent most of the day working my way through the current index implementation. Somehow I have a lot of trouble understanding the documentation, and I ended up writing quite an extensive program to investigate the data structures. At the end of it, I found my suspicions confirmed: I wasn't the only person to misunderstand. I did, however, find a possibility which might save me a lot of time and space.
| Saturday, 12 June 2004 | Wantadilla | |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
Three weeks since I went through all my mail! The result was about 7,500 messages to look at, and spent most of the day doing that.
| Sunday, 13 June 2004 | Wantadilla | Images for 13 June 2004 |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
A whole day spend only with beer (and yes, I drank some too). Started off in the morning with another batch of beer, this time somewhat complicated by my intentions to add appropriate minerals. It's interesting to note that most of the anions of interest are heavier than the corresponding cations, which makes a balance quite difficult. More details in my brewing log.
Apart from that, also did some work on software and documentation, which kept me until early evening, when I headed off to Grumpys to hear David Logsdon of Wyeast talk about yeast. Very interesting talk, more details in my report.
| Monday, 14 June 2004 | Wantadilla | Images for 14 June 2004 |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
Today was a public holiday (one of the Queen of England's many birthdays), and I had intended to catch up on things I had left behind while attending to beer things, but somehow it didn't work out that way. I did get some time to look at the proposed amendments to the constitution of the IT council of South Australia, most of which was taken up trying to use Microsoft “Outlook Express” to read the messages and Microsoft “Word” to print them out. What a pain this software is! It's not so much that it's bad as that it's built on broken concepts, that people are too stupid to learn to think.
Spent a lot of time looking at the beer temperature control, though, and made many improvements, some of which seem to have made things worse. On the other hand, I was testing on a very vigorous fermentation, and that won't have made things easier. Still didn't find time to get gnuplot to interpret times correctly, so what we're left with is:
From about 60% into the time, the temperature control gets pretty ragged. Admittedly, it's not as simple as it seems: the first temperature spike was when I poured in 3 litres (1/8 of the total volume) of water at about 24°, and the second was when I taped the temperature sensor to the fermenter more carefully. I may need to look more carefully at how I measure the surface temperature.
| Tuesday, 15 June 2004 | Wantadilla | |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
Back to work today, with interruptions. Started with testing the code I wrote in the middle of last week, but didn't get very far before having to go to Adelaide for an extraordinary meeting of the IT council of South Australia to discuss the proposed changes to the constitution. Looks like we're going to split into an executive board and a less active council with many more members. In many ways, the meeting was like the AUUG meetings on similar topics, and after two hours we were still far from agreed. Still, hopefully Graham has enough to go on to bring out a draft for vote in 3 weeks time.
| Wednesday, 16 June 2004 | Wantadilla | |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
Long, vociferous teleconference this morning; at least they're not as boring as I (or, presumably, the others) had feared. It's also interesting how teleconferences always seem to go on for between 30 and 60 minutes, no matter how much or how little one has to say.
On with my program, and made another milestone: I can now write data, though not very much before blowing the whole thing up. I suppose that's some kind of progress.
| Thursday, 17 June 2004 | Wantadilla | Images for 17 June 2004 |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
On with my program today. It's slow work, but I'm making some headway.
In the evening, racked Brew 32. Considering David Logsdon's recommendations, decided to drop the temperature to 17°. The resultant temperature graph was interesting:
|
|
||||||||
The second drop to 16° was an experiment. It's fairly clear from the gradient of the ambient temperature (green) that it's taking longer to cool down the beer than the fermenter probe shows. Without stirring the beer, I'm not sure what the correct approach here is. It's also interesting to note that the second fermenter (probe 2, brew 30) cools down more slowly than the first, though it contains the same amount of beer. I can't make up my mind whether this is a question of temperature distribution in the fridge or insulation of the temperature probe.
| Friday, 18 June 2004 | Wantadilla | |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
Finally got my program almost working. I can now write files, though it's a little untidy at the end. Still, that's a minor detail. By the end of next week I should have it reading as well.
Earlier this month I took our tractor to Jeff Hooley's place for sale. At the time it appeared to be a waste of time, since we decided we wouldn't get enough for it. One thing is apparent, though: the reason that Jeff sold is because he was offered 3 times what he paid for it 7 years ago. That could happen to us too, and though we're happy here, the prospect of making some big money and living in even more luxury in a cheaper area is attractive. Today set off looking. The first place is one we looked at years ago, opposite both the Kuitpo Forest headquarters (good) and the Kuitpo sawmill (bad). We turned it down then, and we didn't even consider it again today.
Then up to Mount Pleasant, not far from Mount Crawford forest reserve (yes, it's the same URL as Kuitpo forest), where land is cheaper. Just getting there nearly made us turn back: the roads are terrible, and there's lots of traffic. It's 40 km from Hahndorf, making it effectively more difficult to get to than Murray Bridge, and the prices aren't that good. There was one nice plot of land, but I can't see us moving there.
| Saturday, 19 June 2004 | Wantadilla | |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
We now have this bee in our bonnets about moving house, so spent most of the day looking round the area to see what we could find. We've decided that, if we do it, the only way to go is to have a new house built. Ten years ago my father and I came up with a design for the “square house”, built around a central atrium, and that concept has stood the test of time, so that's what we'd be building.
First down to Kuitpo Forest to look at a well-hidden plot of land. It's quite pretty, but very big and quite expensive. I also can't imagine how we'd get an Internet connection there; it's a long way from any telephone exchange and quite low-lying, so we can't get wireless either.
Then via Mount Barker to Callington, where we saw two properties on two different Jaensch Roads. It's a lot dryer, and thus cheaper, out there. The first one was pretty boring except that most of the land was used for growing barley, which had a certain romance about it. The second one is on a hill with a superb view of Mount Lofty, and also with a phone cell tower a few hundred metres away. That looks like a much better choice. Unfortunately, it's a bit too hilly.
On to Murray Bridge, where we saw a horse property across the road from the (general aviation) airport. It already has a house, but it also has two titles, so that might actually work in our favour. Still, it's a long way from any good riding: the closest seems to be in the Monarto Woodlands, which are still 15 km away and coincidentally a lot closer to the second Callington property.
| Sunday, 20 June 2004 | Wantadilla | |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
Another beer day. In the morning bottled 100 bottles of beer, brew 30 and brew 32. After that spent most of the rest of the day packaging my temperature control software, including adding it to the FreeBSD ports collection. And I had planned to do so many other things.
| Monday, 21 June 2004 | Wantadilla | |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
Back to work on my program today, and made further progress, so I can now write programs up to a point; after that, for some reason, I end up with a SIGSEGV in memmove, a likely enough place—except that the parameters passed to it look fine. Spent some time investigating that one; it could be hiding a very interesting and obscure bug.
| Tuesday, 22 June 2004 | Wantadilla | |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
On with my work today, and made further progress. My obscure bug turned out to be difficult to find but easy to fix: the Monkey record offsets are 16 bits, and in the past there's been no problem. Now I have increased the block size to 64 kB, and the offsets can grow into the sign bit, causing some interesting data corruption as the data lands 64 kB from where it should be. Changing everything to unsigned solved that problem. The issue was that the problem showed up in memmove, which has no stack frame, and so my macros were showing the wrong values. Ended up having to dump the top of stack directly on the call instruction, and that showed a negative count being passed.
Michael Pietsch of Sexton and Glover came and took a look at our house today, and was apparently not impressed. The price he thought we could get was well below what we expected, and indeed well below what people are asking for other smaller properties in the area. We're a little confused, but if we don't sell, it'll have the advantage that we'll have more time to do other things.
| Wednesday, 23 June 2004 | Wantadilla | |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
More program work today, and discovered that I had missed out some important functionality. Still, it looks like the basic structure is nearly finished, and I'll soon be able to read what I'm writing. I'm finding the continual creating work very tiring.
Another estate agent visit today, this time Chris and Tom Weston from Weston Raine & Horne. They'll get back to us with some ideas. Then back to look at a property near Kuitpo Forest, but it's basically a chicken farm with two enormous battery sheds. Not our style, I fear.
In the evening to town to an information evening by Solution City, including the SA Minster for IT and the Mayor of Adelaide. Heard an interesting talk about the Adelaide delegation (the only delegation from all of Australia) at the World Conference on IT, mainly the same photos as on the web link. It's interesting to see what the others are doing, but it's rather a long way from my particular areas of activity.
“Solution City” was an idea we had at the time. I don't know if it ever really took off.
| Thursday, 24 June 2004 | Wantadilla | |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
On with my program today, and made some progress, at the same time noting that gdb is really bad at some things. I haven't looked at the new version 6, though I probably should do, but the previous versions have such obvious problems as being unable to directly specify a line in file (you need to find a function in the file first and list that, making that the default file; then you can specify a line number). One of the things that I found particularly annoying was the lack of a “canonical” raw memory dump function such as the kind that hexdump -C does:
Set to writing one, and was surprisingly successful:
This enabled me to write other macros that call the macro for a raw dump:
Somehow I find programming in gdb macro language particularly frustrating, perhaps because the language is (gratuitously) different enough from C to not be obvious, and there appears to be no documentation. There's also some kind of implicit typing in the parameters, with the result that I could make changes, test them and confirm that they work, but after restarting gdb I'd get messages like:
More work to be done. I'm pretty close to being able to read files, though.
| Friday, 25 June 2004 | Wantadilla | |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
In the middle of the night it occurred to me that my storage formats needed to be sorted, which makes little-endian integers a no-no. Spent most of the day fixing that, not completely.
To town to the ADUUG meeting today. Only Grant Ward and Mark Prior showed up. Somehow enthusiasm for software is waning across the board.
After that, finally bought a new UPS for my laboratory machines. The number of power failures recently has been completely unacceptable; we must have had 5 or 6 this month.
| Saturday, 26 June 2004 | Wantadilla | |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
Rather to confirm the correctness of my purchase of another UPS, we had a further power failure this morning, which at least gave me a chance to install the new UPS. I don't like taking machines down when they've been up for a long time, but after a power failure I don't mind. Still, we're a long way from the uptimes I had at the beginning of the year; the longest uptime is flame, with only 180 days.
My brewing temperature control software seems to have a problem with power failures, too. I'm not currently brewing anything, but I'm keeping some yeast in the fridge at 0°. After a power failure, the system comes back and restarts the temperature control software, but it seems that the relays don't switch on. After stopping and restarting, it works normally. Looking at the configuration, it seems that the system was still starting lpd, which uses the same port. I don't have a printer connected, of course, but it's likely that at startup lpd resets the port. I've removed lpd from the configuration; we'll see if things work correctly next time.
Con Zymaris has resigned as editor of AUUGN, and we haven't found a replacement for him. Why do I always end up holding the baby? In any case, given the work that it caused Con (admittedly using OpenOffice), I had been putting it off, at least partially with the excuse that people due to hand in their copy hadn't done so yet. Today I finally had to admit that the end of June is approaching, and got down to work, converting the whole thing to groff—not necessarily a recipe for getting things done quickly. To my surprise, by the end of the day I had about 30 relatively well formatted pages. There's still a lot to do, but most of it involves getting more copy from people. Makes me wonder what the fuss was with OpenOffice.
| Sunday, 27 June 2004 | Wantadilla | |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
Down to the airstrip off Blackfellows Creek Road in Kuitpo Forest this morning for an endurance ride organized by the SAERA, along with Michael Hickinbotham. It was Michael's first training ride, and he's relatively new to riding in general, and fairly soon it was clear that he couldn't keep up, so with his agreement I went on ahead.
It has been a while since I rode Darah any distance, so I only went into the 20 km ride. To my complete astonishment, Darah didn't even manage that: after 10 km she was flagging visibly, and Michael caught up. Towards the end he overtook us, and I had to dismount to let her catch up. Clearly I need to do some preparation before I do another of these rides.
| Monday, 28 June 2004 | Wantadilla | |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
On with my program work today, and though I had limited success testing, it became clear that I needed a better strategy to handle endianness, and ended up starting the implementation of a complete layer to perform changes. Now all in-core stuff is host-endian, and the stuff on disk is big-endian. Coincidentally it makes the code more readable, since a number of things can be implicit.
Somehow hardware was the flavour of the day. In the afternoon, received in the mail 9 36 GB SCSI disks donated by Ade Lovett. They're ex his servers, and they've given him some minor problems in the past, but they're almost certainly good enough for things like performance measurement. Then discovered I had received an almost new (2 month old) Dell Inspiron 9100 laptop—that was the good news—and that the display had failed. Spent some time on the phone with Dell, who were remarkably efficient, and on IRC discovered that the display failure is apparently a known problem with this model.
To round things off, in the evening echunga died again. I'd already had a problem of this nature at the end of February, but this time it happened while I was there: the screens went dead, and the audio CD in the CD-ROM drive stopped dead, something that almost never happens (once you start it playing, it carries on without further activity on the part of the CPU). Spent some time trying to revive it, without success, and ended up putting the disks into zaphod. Had the machine up and running in about 70 minutes, but then took some time longer to configure X (echunga is the new machine with three displays described in my hardware configuration page). Looks like another dead motherboard, and this one was only nine months old.
| Tuesday, 29 June 2004 | Wantadilla | |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
Up this morning to find that echunga had spent hours doing level 2 dump. At first I thought it was because of the slower processors (bzip really takes a lot of processor power), but then I noted:
Fortunately (in this case), I hadn't fsckd my main source disk, /src, which would have dumped 80 GB and completely overflowed the dump disk. Looking at /etc/dumpdates, I saw:
There was the problem: dump goes by file systems, not by mount points. The ASUS BP6 motherboard that zaphod used has four IDE controllers, and the names of the disks had changed to /dev/ad4s1a etc. I had accounted for that in /etc/fstab, but I hadn't thought of it in /etc/dumpdates.
Next, tried to fsck /src. The system froze during phase 2 and wouldn't even let me into the debugger. Rebooted, repeated, froze again during phase 2. And a third time, after which I decided to try doing it on a different machine. beeble is about all that's left, and that turned out to have file system corruption on the FreeBSD 4.10 side. Tried again with FreeBSD 5.2, and had to boot several times before the disk was recognized. When I did, decided that it was flaky enough to try a partial backup of those directories I knew that I had changed since the last backup (on Saturday); the second time round, that worked. Then set to doing an fsck, which, surprisingly, also worked. I was also able to read the entire disk:
That took almost exactly 40 minutes, or 33 MB/s, and reported no errors, so I felt relatively safe.
Still, I didn't trust the disk, and spent some time looking at the SCSI disks that I had received yesterday. The results were discouraging: I couldn't get any of them to come ready. They're IBM DDYS (36.7 “GB”) and IC35L036 (36 “GB”) LVD drives. They were all jumpered for auto spin, and the host adaptor (a Symbios 875 LVD adaptor) recognized them without any trouble. It just couldn't “spin them up”. Tried also with an old DPT host adaptor and an Adaptec 2740 UW, and also tried forcing SE (another jumper on the drives), all without success. At least the Adaptec reported the problem better. Symbios said “Cannot verify disk”. Adaptec reported:
This shows that the CDB (command descriptor block) was a request sense command (first byte 3), and that it was expecting up to 14 bytes back (0E, rather less than the drives can probably return). It seems strange that this should happen across the board, but I can't see anything I'm doing wrong.
| Wednesday, 30 June 2004 | Wantadilla | |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
And another day spent in hardware hell. During the morning it became clear that my /src disk drive was seriously sick, and sent Yvonne into town to pick up some new hardware. In the meantime tried to improve my system upgrade procedure, with only moderate success; did manage to install FreeBSD 5.2.1 on a disk destined to be the new wantadilla, and built a -CURRENT world on it.
Yvonne back with a new computer case and a MSI K7N2 Delta motherboard, the same kind as in wantadilla. I had bought it because I knew it worked in wantadilla. First, though, I connected up the old motherboard to the new power supply. I couldn't even power on. By contrast, the power supply in the old case would run even without the motherboard, so it looks as if the motherboard is well and truly dead. Hopefully it hasn't taken the processor and memory with it. I can't test the latter, since the new motherboards also take new memory. I didn't want to test the former, since it's such a pain to replace processors nowadays.
Installing the new motherboard wasn't as simple as I thought. It proved to be slightly different from the other motherboard I had; in particular, the BIOS allowed setting the process FSB speed correctly: I had had some trouble with that when installing the last one. Also this version comes with an IOAPIC, which I was pretty sure was not present on the old one. Anyway, decided that it would make more sense to build the kernel on the new box (2500XP+ with 1 GB memory) than on zaphod (in this incarnation dual 466 MHz Celerons with 128 MB memory). Only: I couldn't boot. It got as far as trying to start init and then hung.
After some messing around discovered that the problem didn't exist with Linux, but it existed with all versions of FreeBSD I tried (mainly -CURRENT, 5.2.1-RELEASE and 4.10-RELEASE). The first one (really the system disk from beeble, which currently doesn't have a system to run in) had a kernel debugger, and it showed that the processes were ready to run, but that the USB subsystem was interrupting at about 200,000 times a second. Reset, disabled USB, and was able to boot. Then I had trouble with the network interfaces, both a 3Com 3C905C and a Realtek 8139. In each case, they, too, interrupted continuously, depending on the system between 78,000 and 80,000 times a second. Disabling the APIC solved that problem.
In the meantime spent some time trying to recover the /src file system onto the new disk I had bought, which involved a number of hangs. This old disk is well and truly past it. I only bought them less than two years ago.
Do you have a comment about something I have written? This is a diary, not a “blog”, and there is deliberately no provision for directly adding comments. It's also not a vehicle for third-party content. But I welcome feedback and try to reply to all messages I receive. See the diary overview for more details. If you do send me a message relating to something I have written, please indicate whether you'd prefer me not to mention your name. Otherwise I'll assume that it's OK to do so.
| Top of page | Previous month | Greg's home page | Today's diary entry | Next month | Greg's photos | Copyright information |
