Greg's diary--October 2003

Wednesday, 1 October 2003	Echunga	Images for 1 October 2003
		Top of page
		next day
		last day

In the morning, investigated further why remote kernel debugging no longer works in FreeBSD. As I suspected, the problem isn't in the serial interface code: I was able to return to ddb from serial gdb, but then the system hung on returning from ddb, which it doesn't do when you enter ddb directly. It looks as if something is locking out interrupts and not resetting them.

Got more mail from Shimokawa-san about getting the remote memory access fixed; it looks as if the new version of gdb may require a library change. Unfortunately, Dan Shearer was working on installing exim on zaphod, and so I couldn't continue crashing the system. Instead continued trying to build a NetBSD kernel with Vinum, which proved more difficult than expected.

Thursday, 2 October 2003	Echunga
		Top of page
		previous day
		next day
		last day

Somehow my energy of earlier in the week has dissipated. Spent most of the day trying to rebuild the system on daikon, not made any easier because of the fact that the kernel build directory had moved, and I was left with an old version. That took all day, but was finally successful.

Apart from that, didn't do too much. Various administrative things managed to get in the way again.

Friday, 3 October 2003	Echunga
		Top of page
		previous day
		next day
		last day

Gradually my work is coalescing, though things are still strange. There's definitely something funny going on with FreeBSD's remote kernel debugging. While reading the code, discovered that a detach will bring you back to the ddb prompt, and that it still works. In fact, a lot of things work, but at some point or another everything grinds to a halt. Managed to get a dump which showed that the system was completely idle. It wasn't responding to interrupts, but there wasn't anything in the dump which suggested that any had happened. At first I picked on an unmatched interrupt disable command, but the processor registers indicate that the interrupts are still enabled. More investigation and code reading necessary. It's really difficult to debug a kernel debugger.

On the NetBSD side finally managed to get a kernel with Vinum up and running, and even managed my first commit, but there are still some strangenesses there. Had planned to commit the Vinum code as well, but didn't understand the instructions for committing vendor branches which are necessary in NetBSD, so instead spent some time wondering why all access to Vinum volumes was rejected with an I/O error.

Saturday, 4 October 2003	Echunga
		Top of page
		previous day
		next day
		last day

Quiet day. Did a little work on the NetBSD issues, none on the FreeBSD stuff.

Dan Shearer now has exim working well enough for me to use it as a front end to postfix. I don't have any particular reason to want to keep postfix, but exim's configuration files were sufficiently non-intuitive (i.e. different from postfix and sendmail configuration files) that I wasn't able to set it up correctly as a primary mailer. The log messages are also very different, and though it's thankfully less verbose than postfix, it takes a bit of getting used to. I wonder what the down sides of exim are: most people I know, even Linux users, prefer postfix. Maybe it's just a performance issue, which is definitely not the problem here. In any case, it's doing a good job of checking the senders, and it rejects a lot of mail that way, though unfortunately not enough to allow me to stop spamassassin.

Sunday, 5 October 2003	Echunga	Images for 5 October 2003
		Top of page
		previous day
		next day
		last day

An even quieter day today. Spent a bit of time addressing the odd anomaly with mail, but on the whole things went pretty well. Apart from that, spent most of the day brewing beer. Cleaning bottles is a real time-consumer.

Monday, 6 October 2003	Echunga
		Top of page
		previous day
		next day
		last day

Finally mail is more or less back to normal, and my downlink traffic has dropped to acceptable levels:

0090BC014DB0,13.52007293701172,18.605719566345217,card1

That's only 33 MB of data instead of up to 80 MB a couple of weeks ago. I still don't understand exim, but certainly the idea of callouts to verify the sender is very useful. There are still problems, though: some MTAs, including, it would seem, version 1 of postfix (but not version 2) accept any name until the data phase, thus disabling the callout mechanism.

I'm off to Singapore to a conference of the Singapore Computer Society at the end of the month, and as usual I don't have my slides ready, so spent most of the day doing that. Despite all the nasty things I say about Open Office, I find that I always have to fight groff as well. Finally finished the slides, but they kept me going all day, and I still need to revise the paper itself.

Tuesday, 7 October 2003	Echunga
		Top of page
		previous day
		next day
		last day

More tidying up with my presentation today, and didn't get much else done. At least I now have a clean slate. Hopefully I can get back to my Vinum work soon.

Wednesday, 8 October 2003	Echunga
		Top of page
		previous day
		next day
		last day

Finally time to get back to looking at my various Vinum projects. It wasn't easy: one of those days where everything seems to go wrong.

Bruce Evans has taken a look at the current problems with remote kernel debugging, and he sent some patches. Decided it would be a good time to bring zaphod up to date before applying them, and first to check whether the problem has gone away. For a number of reasons, building the system has become a real time sink, and it took all day.

In the meantime, looking at daikon, my NetBSD -CURRENT machine, which for some reason decided that it didn't want to perform I/O to Vinum volumes. Put the serial kernel debugger on that with sydney, and sydney promptly decided to hang up. I've seen this problem before a few months ago, and it seemed related to the Ethernet cards I was using, so spent much of the afternoon mixing and matching to find something that would work. Finally dropped back to FreeBSD 4.8, which worked, but the version of gdb on 4.8 doesn't work with NetBSD kernels. sigh

Thursday, 9 October 2003	Echunga	Images for 9 October 2003
		Top of page
		previous day
		next day
		last day

More work on the two Vinums and kernel debugging today. Tried out Bruce Evans' patches to the kernel debugger, but they didn't have any effect. It seems to be an SMP-related issue, though, so put zaphod back to build a single user kernel. sydney was also building another world, so used zaphod to debug daikon, which mercifully worked, and showed only a minor problem. At least I now have Vinum working on NetBSD again:

    === root@daikon (/dev/ttyp3) /home/grog 6 -> uname -a
NetBSD daikon.lemis.com 1.6ZC NetBSD 1.6ZC (DAIKON) #6: Wed Oct  8 15:20:19 CST 2003  grog@daikon.lemis.com:/usr/src/sys/arch/i386/compile/obj/DAIKON i386
    === root@daikon (/dev/ttyp3) /home/grog 7 -> df /mnt
Filesystem       1048576-blocks     Used     Avail Capacity  Mounted on
/dev/vinum/first             96        0        92     0%    /mnt

Hopefully I'll be able to check the stuff in tomorrow.

Friday, 10 October 2003	Echunga
		Top of page
		previous day
		next day
		last day

Finally I had Vinum ready for committing to NetBSD, so spent most of the day doing that. Somehow there must be an easier way to update the CVS repository: NetBSD doesn't use cvsup, since it isn't available for all platforms. A normal CVS update seems to take forever.

I had submitted my patches a couple of months ago, but I still managed to surprise people when I committed, and we had quite a long discussion about the merits of additional flags for newfs. That looks like keeping us busy for a while.

Saturday, 11 October 2003	Echunga
		Top of page
		previous day
		next day
		last day

Quiet day. Spent some time thinking about how to handle NetBSD's newfs, which in the end turned out to be relatively simple: there just needs to be an ioctl to determine the size of the file system. Sent a message out about that, and to my surprise got no response before evening.

Building software doesn't seem to be getting any faster. Ten years ago it took all day. In those days my fastest machine was a 486/66. Nowadays I use an Athlon 850 for my test machines, probably 20 times as fast, but it still takes all day. It's quite a problem with the work I'm doing at the moment.

Yana finally brought the exim book that Dan Shearer gave her on Thursday. Another author who wrote an O'Reilly book with Andy Oram as editor. The book looks good, and I more or less linearly read the first couple of chapters, something I don't often do.

Sunday, 12 October 2003	Echunga
		Top of page
		previous day
		next day
		last day

Yet another quiet day. For some reason I have a guilty conscience when I'm not working seven days a week, something I need to work on.

Today spring was in the air, it wasn't raining, and it was dry enough to mow the lawn, which was desperately in need of it: in some places the grass was up to 1.5m high.

Apart from that, continued reading the exim book, something I haven't been able to do for some time, and also reading a C't article about using computers as a digital TV recorder. It looks as if analogue TV is on its way out in Europe. I must investigate whether the Australian TV standard is compatible.

Monday, 13 October 2003	Echunga	Images for 13 October 2003
		Top of page
		previous day
		next day
		last day

Got a phone call in mid-morning from an ISP in rural South Australia with a machine which they thought had been compromised, and which they couldn't get back up again. Proved to be a FreeBSD 3.3 installation with Vinum, apparently installed by Mark Newton. Somehow they had managed to lose the contents of their /modules directory, which meant that Vinum couldn't load, and it made recovery very difficult, especially as there was no technical staff on site.

Spent half the afternoon sorting that one out: in the end we copied the modules from another system via floppy, since none of the standard tools were available without the /usr file system, and we couldn't mount a CD-ROM because of the missing module. Yet another reason not to have a separate /usr file system, though I suppose with a modern system with a root file system on Vinum as well, things would be even more complicated. Still, that's not an argument against modules: they might just as easily have deleted the kernel, in which case things would have been even more difficult.

This work is tiring, and after it I didn't feel like any mental activity, so out and continued mowing the lawn. Spring is gradually (and very belatedly) showing its face.

Tuesday, 14 October 2003	Echunga
		Top of page
		previous day
		next day
		last day

Got another apparently forged message from eBay today, demanding payment and offering neither explanation nor a method of contacting them. Confirmed that the URLs do exist, and spent nearly an hour trying to find a way to contact a human at eBay. It seems impossible. What a load of incompetence! About the only way to contact them was to report abuse of their system, and since I couldn't completely rule out abuse (how could I without a way of confirming?), sent them an abuse report. It'll be interesting to see whether anything comes of it.

In the afternoon, round to Diane Saunders' place to pick up some old furniture, including a fridge which I propose to use for brewing beer. It doesn't look very well suited to that task: the thermostat seems to be broken, and the highest temperature I can set is round 5°. For brewing I need something closer to 10°, which is normally not a problem. It's surprising how often fridge thermostats fail by dropping the range of temperatures which can be set. I wonder what the background is.

Later Yvonne came home with a tortoise; it seems there are plenty around here, and Diane found this one crossing the road. It took happily to our pond. I wonder if we'll ever see it again.

Somehow didn't get much done in the day. Time to get back to real work.

Wednesday, 15 October 2003	Echunga
		Top of page
		previous day
		next day
		last day

No reply from the eBay abuse department. I bet that they're going to be a lot faster to condemn me as a bad financial risk if I don't pay, so I ended up paying my $3 or so without knowing what it is. There's something terribly wrong with this current anonymous mentality. Anybody who wants money from you should be prepared to explain why. I'm not happy.

Spent some time messing around with Vinum, and finally managed to get kernel debugging working on zaphod. The problem seems to be related to SMP: running on only one processor, there's no problem.

The conversion to exim has still not finished. One great disadvantage with exim seems to be a lack of a basic configuration file which you can tweak; at any rate, I found the supplied configuration file much harder to understand than that of postfix. The one I have insists on forwarding to the primary MX (it doesn't want to be that itself), but there's no other way for the mail to get there. Previously I had an MX (wantadilla) firewalled off, but that obviously takes too long to time out. Ended up having to install two different kinds of DNS to get things to work.

Things weren't helped by my old laptop, sydney, which had generated a mail loop for the domain worldwide.lemis.com with the exim running on echunga and filled up its disk. Reconfigured DNS so that wantadilla was the primary MX for worldwide.lemis.com, which I also degraded to a host, and ended up sending round 15,000 messages across, where spamassassin spent an average of 30 seconds looking at each of them. There must be an easier way.

Accepting all those mail messages kept wantadilla 100% busy for quite some time, and not surprisingly the CPU temperature rose. The reported chipset temperature didn't, though:

CPU temperature graph

As the summary at the top of the graph shows, it considered the temperature to be 0° C. It was already unlikely to be only 18° at the beginning of the graph, but 0° is obviously wrong. I wonder what kind of bug is causing that. I'd guess it's the BIOS. That begs the question, of course, whether the reported CPU temperature is correct.

Thursday, 16 October 2003	Echunga
		Top of page
		previous day
		next day
		last day

The weather has been gradually getting better, and as a result the level of water in the pond has been dropping. It looked too low in comparison with just a week ago. We have an automatic level adjustment (water feed with ball valve), but it's hidden in the reeds on the other side of the pond, and tends to get jammed up in them. Set off to take a look, but couldn't find it. Went and got the whipper-snipper, which had the wrong fittings on it (string instead of blade; the reeds would laugh at that), and spent a good 30 minutes trying to change the thing over. In the end, gave up: the instructions are anything but clear, and my attempts to do what seemed right led nowhere. Yvonne off to town to get expert advice, and found that I had been on the right track, but that the thing was so clogged with dried grass that I couldn't insert the pin.

Off again chopping down the reeds looking for the pipe and ball valve. Spent another 30 minutes doing that, chopping down a surprising number of reeds before finally finding it pretty much where we thought, just further into the pond: the level is still far too high. sigh What a waste of time.

Somewhat unmotivated back to work, and did a bit of messing around, but nothing of great importance.

Friday, 17 October 2003	Echunga
		Top of page
		previous day
		next day
		last day

Somehow I seem to be getting nothing done. Bruce Evans committed a new patch for SMP machines to do serial debugging correctly, and I spent too long building that, but then ran into trouble with firewire support, and wasn't able to do any testing. Also a lot of discussion on the NetBSD lists about process. The code may look very similar to FreeBSD, but the projects are significantly different in flavour.

Saturday, 18 October 2003	Echunga → Olivaylle
		Top of page
		previous day
		next day
		last day

Up early and off to Olivaylle for a weekend relatively far from computers. Took La Tigre, who is to stay a couple of months and be bred, and Darah, whom I wanted to ride. Yvonne found it rather strange that I should take a horse with me, since they have about 80 of them, including a number of eminently ridable ones, but then I don't ride Darah enough, and this sounded like a good opportunity.

The trip was fairly uneventful—it's only 340 km—but the weather is getting a lot warmer. In the afternoon rode around the property, where the temperature really became apparent: it was round 35°, a far cry from the 15° or so we've been having in the last few days.

Sunday, 19 October 2003	Olivaylle
		Top of page
		previous day
		next day
		last day

The temperature dropped significantly in the night, and it rained a lot. By morning it was 12°. With Yvonne down to the homestead to talk to Jorge de Moya, and somehow managed to talk him out of a foal from his Paso Fino mare “Guantanamera”, in exchange for one out of La Tigre. That feat managed to surprise both Chris and Fiona, and since Guantanamera (whom, unfortunately, they have called “Wanda”) was in season, there were some preparations for that.

First, though, off riding again, this time at a faster pace, especially after we came across a pair of red kangaroos, whom Darah chased with delight (but obviously without success). Nice canter country.

In the afternoon, tried out José, a “problem” Paso Peruano gelding. Yvonne had thought it would be a good idea for Darah to have company on the way home, but I suspect she's just trying to collect more horses. Anyway, José didn't seen that bad, so he'll come home with us tomorrow.

After that, bred Guantanamera to Cali, not without significant preparations: Guantanamera is not easy on stallions.

In the evening, dinner with Fiona and Max. Max was up at the pump house, where they have spent a remarkable amount of time installing a computer-controlled irrigation system (using Microsoft!). Interesting setup, with about 10,000 litres each of phosphoric acid, potassium nitrate, ammonium nitrate, another nitrate and “trace elements”. Unfortunately, they didn't have time to explain it to me: the irrigation season has started already, and every minute counts (witness the work on Sunday).

Managed to drag Max down in time for dinner, and had a good time. Max and Fiona collect old kerosene lanterns, and their house is quite interesting.

Monday, 20 October 2003	Olivaylle → Echunga
		Top of page
		previous day
		next day
		last day

Up early this morning wondering whether we should breed Darah as well, but she proved not to be in season, either “after all” or ”at all”. Discovered that we didn't have much else to do—we had already decided not to go riding this morning—so went home. The weather was still relatively cool, but the wind was very strong, and with the horse trailer, the fuel consumption was higher than I have ever experienced. Between Bordertown and Tailem Bend, a distance of 170 km, we used over 40 litres of petrol, a consumption of 24.1 litres per hundred. We hardly got the car out of third gear the whole journey.

At home, the usual 3000 mail messages waiting for me, but I wasn't in the mood and didn't do too much. Maybe I need to take a couple of weeks' holiday some time in the not-too-distant future.

I've recently got some new (second hand, in fact) loudspeakers for my 20-year-old Technics amplifier, which means that I can again use digital computer output. It didn't work: since I last used it, I have changed the motherboard, and the sound hardware is different. In fact, it's not just different, it's newer than the operating system version that wantadilla is running (a 12 month old version of FreeBSD 5-CURRENT), so when I loaded the correct kld, it didn't recognize the chip: had to rebuild the kld from newer sources first. Even then, sound wasn't brilliant. I suspect that interrupt latency is rearing its ugly head again.

Tuesday, 21 October 2003	Echunga
		Top of page
		previous day
		next day
		last day

After yesterday's problems with wantadilla and sound, decided it was probably about time to upgrade the system. I had a spare 40 GB drive hanging around (in fact, from the previous rebuild of wantadilla), and it looks as if I had started an installation on it a couple of months ago, though I can't recall anything, and there's nothing in my diary. It's configured with an 8 GB root file system (including /usr) and a little over 1 GB of swap, enabling me to dump core of a 1 GB machine when necessary. Started an update on that, and tried to remember where I had started a list of what to do to upgrade or clone a machine. Spent some time searching for that, without success.

Apart from that, more catching up on mail. Somebody has recently done a performance comparison of Linux and the BSDs, with interesting results, causing much mail. OpenBSD did particularly badly, and I suspect there are going to be a number of people who will be spending time addressing the issues in the near future.

Wednesday, 22 October 2003	Echunga
		Top of page
		previous day
		next day
		last day

Time for a system upgrade

Topic: technology

Link here

Came into the office today to find that wantadilla had crashed and rebooted, and for some reason X had died on echunga. High time for my upgrade, and spent some time working on that. Finally found the description of to upgrade a system that I had been looking for yesterday, and spent some time copying stuff across in the background.

More Vinum work

Topic: technology

Link here

Apart from that, found some time to work on Vinum and got some minor work done. Much more should follow.

Thursday, 23 October 2003	Echunga
		Top of page
		previous day
		next day
		last day

Came into the office this morning to discover the display wantadilla:0 looking distinctly sick. That's my 2048x1536 display, the one I use for images. Shut down the machine and discovered that the display board had apparently overheated, and I suspect blown a capacitor. Looks like more work to be done.

After that, a call from Internode, who have been having some strange problems with a batch of motherboards they've been getting lately, and we couldn't work out whether they were caused by software or hardware; certainly they've seen a lot of version sensitivity. This box wouldn't run at all on FreeBSD 4.8, and they were getting frequent panics with 4.9 RC3. Looking at the dumps, it eventuated that they appeared to be out of KVA, not that surprising for a 4 processor machine with 2 GB of main memory. The surprising thing was that the obj_zone information has a null pointer for the KVA address:

(kgdb) p *obj_zone
$2 = {
  zlock = {
    lock_data = 0x0
  },
  zitems = 0x0,
  zfreecnt = 0x0,
  zfreemin = 0x2c,
  znalloc = 0x12a86d,
  zkva = 0x0,
  zpagecount = 0x0,
  zpagemax = 0x0,
  zmax = 0x0,
  ztotal = 0x17288,
  zsize = 0x5c,
  zalloc = 0x1,
  zflags = 0x10,
  zallocflag = 0x2,
  zobj = 0x0,
  zname = 0xc04278fc "VM OBJECT",
  znext = 0x0
}

They were all the same in this respect, though the panics themselves were the result of zalloc() trying to assign memory from this pool. Spent some time pondering that, though the important thing was to get the machine back up and running. Built a kernel with more KVA and waited for it to crash again.

Didn't get much other work done.

Friday, 24 October 2003	Echunga
		Top of page
		previous day
		next day
		last day

Double trip to Adelaide today. Lunch was with Mark Prior's ADUUG (Adelaide UNIX User Group), spent mainly discussing licenses for TV and DVD viewing.

Both before and after lunch picked up a hop plant: Pride of Ringwood from Thomas Hamann at Grumpys before lunch, and a Hallertauer from Gavin Scarman after lunch. Hopefully they'll get far enough this year to be able to bear reasonable fruits next autumn.

Also managed to get hold of a thermostat for the fridge at Jaycar: I should be able to both raise and lower the temperature depending on what I want to do.

In the evening, to the annual dinner of the Australian Democrats, to which I had been invited by Ian Gilfillan. Interesting evening; I was surprised how many people I knew there, and also how many were interested in the concept of Open Source in Government. Looks like we'll get a keynote for the Open Computing in Government conference in January.

One thing that I didn't mention at the time was one of the people at my table, who told me that he had found a way to use water as a fuel for his car, and that he had driven his car all over Australia with no other fuel. It seems he was well known for that: I mentioned it to Ian later, and he knew all about it.

Saturday, 25 October 2003	Echunga	Images for 25 October 2003
		Top of page
		previous day
		next day
		last day

After the “excitement” of the last few days, didn't do much today. Changed the display card in wantadilla: the old one looked decidedly the worse for heat:

https://lemis.nyc3.digitaloceanspaces.com/grog/Photos/20031025/big/display-card-1.jpeg

Image title:	display card 1	Dimensions:	2048 x 1536, 368 kB
Make a single page with this image		Hide this image
Make this image a thumbnail		Make thumbnails of all images on this page
Make this image small again		Display small version of all images on this page
All images taken on Saturday, 25 October 2003, thumbnails		All images taken on Saturday, 25 October 2003, small
Diary entry for Saturday, 25 October 2003		Complete exposure details

There are clear discolourations above the 69C4 on the lower label, and also to the left.

Setting up the new card (an ATI Rage 128) was fairly straightforward, and it came up at 2048x1536 with no problems. Tried higher resolutions, but it didn't work; I suspect the chip doesn't understand them.

Apart from that and a bit of messing around, didn't do much.

Sunday, 26 October 2003	Echunga
		Top of page
		previous day
		next day
		last day

Daylight Savings Time started today, and it's always a bit disorienting. It's also the day that DST ends in much of the Northern Hemisphere, so there's a total shift of two hours relative to them.

Today we got a new spammer (rootwars.org) with a difference: we weren't able to block him, at least not on the AUUG machine. It worked fine on my machine, which was running version 2 of postfix, while AUUG was running version 1. Updated postfix, but the problem still existed. Spent an inordinate amount of time to discover that the configuration files allow you to put unrelated keywords in sections: we had

smtpd_recipient_restrictions =
        permit_mynetworks,
        check_recipient_access hash:/usr/local/etc/postfix/always_accept_to,
        check_client_access hash:/usr/local/etc/postfix/client_access,
        check_sender_access hash:/usr/local/etc/postfix/sender_access,
            reject_unknown_sender_domain,
        reject_maps_rbl,
        check_relay_domains

That didn't help much. Looking at the sample-smptd.cf configuration file, I found that I had it in a separate section:

smtpd_sender_restrictions =
        check_client_access hash:/usr/local/etc/postfix/client_access,
        check_sender_access hash:/usr/local/etc/postfix/sender_access

The description in the config files is still pretty confusing, but it seems that the first version defers the checks to the RCPT processing, whereas the latter does it when establishing connection. Both should work, but the first one definitely didn't. I'm beginning to think that conventional table-driven MTA configuration is not adequate. Of course, if somebody could get rid of the spammers, that would be even better.

Monday, 27 October 2003	Echunga
		Top of page
		previous day
		next day
		last day

Finally got some time to work on Vinum again, with the intention of looking at Poul-Henning Kamp's changes to the bio layer. Decided that the best thing I could do would be to finally start writing some regression scripts (a thing that Poul-Henning has been asking for forever), and made some progress on that, by no means helped by the permanent failure of another power supply to my Sun disk array, along with a transient failure of yet another. I now have about six defective power supplies, and if things carry on like this, I soon won't have enough to continue using the trays. I wonder where I can get them repaired.

I had thought that the problem with the power supplies was related to leaving power on to the trays while they're turned off, so now when I turn them off I disconnect all power. That doesn't seem to help. My best bet is that they must suffer damage during the power-on surge.

As if that wasn't enough, had a number of panics. A couple of camcontrol functions no longer work correctly: camcontrol rescan should rescan a SCSI bus, but I can perform a rescan after pulling a drive, and it doesn't notice. Subsequently the system hangs when I try to access the disk. On the machine that is to become wantadilla, camcontrol devlist, which should list devices, panics the system instead. sigh.

Despite these problems, made some progress with scripts to create Vinum objects and perform testing on them. Somehow the performance doesn't look as good as it once was.

Tuesday, 28 October 2003	Echunga	Images for 28 October 2003
		Top of page
		previous day
		next day
		last day

Back to the conference call schedules again, this time for the Linux in Government conference planned for January. Looks like we'll have an interesting programme.

Mail from Simon Hackett of Internode: their POP server has crashed again a couple of times, with the same syndrome as last week. Looks like I'll have to set a memory access breakpoint to catch the thing. Given that it's a production machine, and down time should be kept to an absolute minimum, it looks like it'll have to be custom code which calls panic() and dumps when the situation occurs.

Apart from that, spent some time preparing for my journey to Singapore tomorrow, and Jorge de Moya showed up somewhat unexpectedly in the evening, so didn't get much else done.

Wednesday, 29 October 2003	Echunga → Singapore	Images for 29 October 2003
		Top of page
		previous day
		next day
		last day

Up earlier than planned this morning: Jorge left at 6 am, as planned, but we had locked the gate, and I had to get up to open it for him.

Up again relatively late, and almost immediately off to the airport to head to Singapore. Qantas has managed to lose my frequent flyer profile again—this must be the third time—so I ended up next to screaming kids at the back of the plane. Stopped in Darwin on the way, made all the more annoying by the fact that there was no access to the Qantas Club from the transit area, and then on to Singapore, where I was out of the airport and in a taxi to town in record time (for Singapore), only about 20 minutes. In town, down to Clarke Quay for some satay. I think this place is more or less past it: only tourists would appreciate it. I won't go back again.

Thursday, 30 October 2003	Singapore	Images for 30 October 2003
		Top of page
		previous day
		next day
		last day

Up early this morning to look for a local breakfast, but this part of town, round the Convention Centre, isn't really designed for that, and wasn't able to find any nasi lemak. Ate some mee instead, then upstairs to the convention, which was a lot lower key than in the past—not exactly an unusual circumstance in the current economy. It was only a single session, and the talks were interesting, dominated by embedded systems and license issues. Things ran a little late, and it was after lunch before I delivered my paper, which I thought was probably a little off-topic for the audience, but it was taken well.

After that, had some time to get through my mail, investigating the hotel Internet connection in the process: they have a “broadband” connection, which looks like it must be Ethernet in disguise; the interface ends up in 10 Mb/s mode, which is more portable, of course, and easily fast enough. Connected up for a brief period of time and was astonished at the speed with which my mail disappeared: it all got delivered within ten seconds. It's a pity they have these time-based charges.

In the evening, less of a get-together than last time. Off with Lim Swee Tat, Sanjeev Gupta and Lim Kim Yong and had some food at a somewhat tired food court nearby. Swee Tat brought his girlfriend Christine with him; sounds like a way to strain the relationship to me. Kim Yong was complaining vociferously about my choice of curry to go with nasi lemak, but I never quite got out of him what he thought should be used instead. I notice that the hotel I'm in serves a fried chicken wing and boiled egg, neither of which have enough juice to go with the dish. Late to bed.

Friday, 31 October 2003	Singapore	Images for 31 October 2003
		Top of page
		previous day
		next day
		last day

One with the conference this morning, with another talk about embedded Linux. The Japanese, in particular, are very interested in this topic. It's funny to think back 30 years or so and to note that the original intention of microprocessors was for embedded work. The whole idea of desktop computers caught the industry by surprise, and it looks like that era might be drawing to a close due to saturation.

After the break, I represented my talk on Why I hate OpenOffice. Not surprisingly, Danese Cooper of Sun, who is strongly involved in both OpenOffice and StarOffice, disagreed. Well, she thought she did: in fact, she just saw the same things from a different perspective. Sure, OpenOffice is freedom for the masses, but so was communism for the Russians. We need to think beyond the first improvement in status.

The conference finished after lunch, and I went back to the hotel to catch up on my mail. There's another conference on next week, confusingly also abbreviated to AOSS, and I have been invited to participate. That means staying on here to Tuesday evening.

In the evening, met up again with Lim Kim Yong and went out on a vague search for a new mobile phone for me: mine is 4½ years old, and the battery is getting rather faint. In the process found a new battery for my old phone, which rather took the wind out of my sails regarding a new one. Off on a roundabout way, passing by a Buddhist temple, a Hindu temple next door, and a heavily guarded synagogue, to Woodlands in Serangoon Road for a Southern Indian Vegetarian Thali. I've been going there since I knew about it having arrived in Singapore, about 8 years (it's an offshoot of the Woodlands Hotel I stayed at in Chennai in May 1967), and in the past I've always been very pleased with the food. Not so today; it was pretty boring, and Kim Yong, who had never eaten anything like it before, was not exactly encouraged to try again. Looks like it's time for a change there as well.

Walked back from there, a little tiring I fear for Kim Yong, stopping for a beer at the market on the corner of Rochor Canal Road and Serangoon Road, and later, after bumping in to a couple of Swedes we had met at the conference, into the Long Bar of the Raffles Hotel. Somehow it's just too touristy: in particular, there were women and children there, which would never have been tolerated in colonial times. The prices had to be seen to believed: $23.50 for two beers. Kim Yong couldn't believe it, and carried on complaining to me about the cost until I dropped him near the MRT station. He didn't have to pay, but he was indignant that anybody could ask that much money.