Greg
animals language
brewing multimedia
food and drink music
fiction health
opinion gardening
photography general
technology history
Stones Road house politics
Greg's diary
August 2001
Translate this page
Select day in August 2001:
Su Mo Tu We Th Fr Sa
1 2 3 4
5 6 7 8 9 10 11
12 13 14 15 16 17 18
19 20 21 22 23 24 25
26 27 28 29 30 31
Select month:
2001 Jan Feb Mar Apr
2001 May Jun Jul Aug
2001 Sep Oct Nov Dec
Today's diary entry
Diary index
About this diary
Previous month
Next month
Greg's home page
Greg's photos
Network link stats
Greg's other links
Copyright information
    
Groogle

Wednesday, 1 August 2001 Echunga
Top of page
next day
last day

The electricians are still here! Today they spent an inordinate amount of time rewiring the network cabling in my office, which kept me away from doing anything useful. They claimed to be finished round 4 pm, but to be sure I went through everything again. Three of four network cables in my office were wrong, and the generator connection to the water pump didn't work. Sheesh.

I did finally find time to finish my photos of the USENIX annual conference in Boston. I made two pages: the general photos and the mug shots taken at the “Kernel Summit”.

On IRC with Warner this evening, we finally found a possible cause for the PCMCIA problems.

We had Hainan chicken rice for dinner tonight: chicken steeped in hot, but not boiling water. Traditional recipes require you to manually maintain the cooking temperature by bringing it almost to the boil several times. I tried instead to maintain the temperature by putting the pot in the oven and keep that at about 80°C. It worked quite well, but required careful monitoring of the temperature. The thermometer only maintained the oven temperature within a range of about 25° between warmest and coldest, though of course the water wouldn't have changed that much. It was at 60° when I took it out, and the meat was very succulent. I need to experiment more.


Thursday, 2 August 2001 Echunga
Top of page
previous day
next day
last day

More work on the PCMCIA code today. Things aren't looking as good as I hoped.

Made the first steps towards getting NetBSD working on the RS/6000. Started by installing NetBSD 1.5 on kimchi and trying to upgrade it to 1.5.1-CURRENT, which proved more work than I thought.


Friday, 3 August 2001 Echunga
Top of page
previous day
next day
last day

PCMCIA development seems to have stagnated. I still have no functional machine, which is sad. Yana's firefly was in such a mess that I could no longer connect to the network, thus making it impossible to fix. On the other hand, the FreeBSD 4.3-RELEASE CDs finally arrived (they were dated April), so removed the disk from the laptop for backup, and plan to install 4.3 on it after that.

Building NetBSD-CURRENT was equally frustrating. After cvs updating the source tree from 1.5-RELEASE to 1.5-CURRENT, nothing would build. Finally forced a build through with -k, but I couldn't configure a kernel. I wonder what I'm doing wrong.


Saturday, 4 August 2001 Echunga
Top of page
previous day
next day
last day

Finally gave up with the -CURRENT and -STABLE PCMCIA code and installed yesterday's 4.3-RELEASE CD-ROMs on firefly. To my surprise, everything worked out of the box, including the Aviators.

Another IRC party today. Stephen Plunkett (sisko) and Daniel O'Connor (Darius) came for lunch, along with Stephen's girl friend Jen. Last time we did it all with wireless on a single laptop; this time Darius brought his laptop with him and tried to get wireless working on -STABLE, experiencing only moderate success. Gave sisko sydney, but ended up being rather more wired than we wanted. Here are a couple of photos.


Sunday, 5 August 2001 Echunga Images for 5 August 2001
Top of page
previous day
next day
last day

Another quiet day. Spent most of the time getting my laptops back to 4.3-RELEASE, which seems to work. I really don't have enough time to work on this any more.

Something is wrong with X on wantadilla. A couple of times in the last day or so the keyboard has frozen up, only in X. Switching to a text mode display frees up the keyboard, but on returning to X the mouse no longer works. I discovered that if I connected a mouse to canberra, my laptop, and used x2x between the systems, it would work fine, and the keyboard freed up again. I hate stopping X, but it looks like the time has come.


Monday, 6 August 2001 Echunga
Top of page
previous day
next day
last day

Spent most of the day setting up a cross-build environment for NetBSD/PPC. It took much longer than I expected.

In the afternoon, breaker on IRC decided to crank up his graphics card to 1920x1440 pixels. It's a Matrox M200, like what I have, so I helped him, and it was successful. Decided that it's been a while since I did any reconfiguration, that I needed to restart X anyway, and that I'm just shy of 10 megapixels, so did some reconfiguration, but that didn't get finished either.


Tuesday, 7 August 2001 Echunga
Top of page
previous day
next day
last day

Continued with reconfiguring X today, first on echunga, where I got the display to 1920x1440 with no trouble. On wantadilla things were somewhat complicated by the fact that I didn't want the displays in the order in which the system detected them, but I was getting past that when Yvonne told me that her system wasn't responding. It turned out to be an NFS hang on wantadilla: the Ethernet card (3C905C) was no longer responding. On closer investigation, it seemed to hang up completely when I started X.

That was strange, because I hadn't changed the hardware configuration, only the resolution. I ended up backing out all the changes I had made, to no avail. Decided to power down the machine and take a look, but I still had a backup of my source disk (27 GB) going. It had been going since the previous afternoon, onto DLT which should 7 GB per hour. Investigated that problem, and discovered that tar was using 100% CPU time. I've seen that problem before: tar has to keep track of files with a link count greater than 1, so that it only makes one copy of the file. GNU tar, which is what FreeBSD uses, does this by storing the information in a linked list. My source tree has a very large number of multi-linked files, and this was causing a heavy load; tar had used nearly 7 hours of CPU time on a 750 MHz Athlon. Its process image had also grown to nearly 40 MB.

While waiting for that to finish, I decided to burn a CD-ROM. Running cdrecord scanbus hung for some reason, but it recovered—with a SCSI bus reset, killing my 18 hour backup. Grrr.

Things didn't get much better after that. Nothing I could do would get the Ethernet card to continue to run reliably while X was running. In the process crashed the system while I was deleting a large number of files on /src; the result was a series of bad problems under fsck, and I didn't have an up-to-date backup.

Messed around with varying Ethernet cards and confirmed that the problem only happened with the 3Com. I had spent too long already, so decided to carry on running with a 10 Mb/s NE-2000 clone, and back to my displays, which then worked fairly well. I now have, from left to right,

As before, they are joined together with x2x. I have now well passed the 10 megapixel mark with 11220528 pixels (previously 9994800). Here's the X configuration for wantadilla.

While looking at the Ethernet problems, I noticed a surprising amount of HTTP traffic. Our web server isn't on the local network: it's hosted by Safeport Network Services, and all we have locally is for our own use. Looking more carefully, I noticed that the requests were going to random addresses on the network. They appeared to be part of the Code Red worm which is infecting Microsoft web servers. Put a block on all incoming HTTP requests, and discovered that we were getting between 10 to 15 requests a second.

Later in the evening tried to fix tar. I did this once before, in August 1995, but the code was never contributed. It didn't take long to compile, but I ran into a bus error I didn't have time to look at, so that'll be one for another day.

As if that wasn't enough, the Australian Government took a census today. There seems to be a theme of Centenary of Federation about the census, which might explain its antediluvial nature: no interactive interface, which would have allowed for consistency checks and easy counting. Instead, everybody gets a form on paper which must be filled out in block capitals suitable for 1970s style optical character readers. I wonder what they will make of my scrawl. Some of the questions showed a lack of thought, too. I had to enter my address, but the form was too rigid to allow me to enter it (“Street number: NONE”). Others were just plain stupid: “Do you speak a language other than English at home? If so, enter just the language you speak most”. Well, we speak English, French and German at home, but mainly English, so that's what I entered. I'm sure that's useful. In general, I note that the census asked very few questions.


Wednesday, 8 August 2001 Echunga
Top of page
previous day
next day
last day

Back to work after yesterday's catastrophe. Well, I had tar to work on. While I was doing that, Yvonne came to me and said she had changed the batteries in her Palm Pilot m100, and wanted to know how to get her address book back. The answer was: she didn't. Somehow the thing had done a full reset, and that before she had done a backup.

Another day of unnecessary work. On Martin Pool's recommendation, downloaded jpilot and installed that. Problem: Yvonne's machine, battunga, wouldn't recognize the serial connection. Spent a lot of time trying things out, but it really looks as if there's something wrong with the machine. sigh.

Got the tar patches working correctly. A backup of a FreeBSD source tree, most of the files of which had multiple links, took 9.5 minutes elapsed time and 2.5 minutes CPU time with the old algorithm, and 8.5 minutes elapsed and 15 seconds CPU time with the new algorithm. I'd assume that the 8.5 minutes are based on maxed out disks; as the tree grows, the difference would get bigger. Unfortunately, the tar process image grew to 110 MB VSIZE, something I should work on reducing.


Thursday, 9 August 2001 Echunga
Top of page
previous day
next day
last day

Another day following up on started work. Mail is picking up, much of it discussing Code Red. I don't understand the press: they went crazy about the “Millennium Bug”, they talk about every new iteration of a mail “virus” as if it was completely unexpected, they announced ahead of time that “Code Red” was going to be a disaster. Now the disaster has happened and they're not talking about it. I'm still getting about 15 connection attempts per second, all of which are harmlessly bouncing off the firewall (as opposed to harmlessly bouncing off something else in the system). The problem is that's going to cost real money. Each packet has a data length of 48 bytes. That makes 720 bytes per second, 2.6 MB per hour, 62 MB per day. At $0.19 per megabyte, that's nearly $12 per day, $360 per month. Who's going to pay that?

More thinking about address books. Before the advent of PDAs, I kept my addresses on my machine in an indexed-sequential file system I had written about 10 years ago, called Monkey. I had a whole lot of tools for updating the files and printing them out with TeX or groff, but it was a relatively large amount of work to maintain the data base, and the last time I updated anything was nearly 2 years ago. Decided to convert the database to Palm format, which was surprisingly simple: Monkey needed recompilation, of course, and because it was written in C++, it ran into some problems with the FreeBSD header files: /usr/include/machine/setjmp.h includes the following:

        typedef struct _jmp_buf { int _jb[_JBLEN + 1]; } jmp_buf[1];
      

This blew the C++ compiler's mind, for reasons I haven't really understood yet. Ended up replacing it with a more traditional definition. After that the only problem was to understand the (undocumented) Palm dump format and write something like it, about 3 or 4 statements. Fortunately the format of the Monkey address book was surprisingly similar to that of the Palm. At the end the only issue I had was a Millennium bug in Monkey, which claimed that this year is 1901. Ah well, we weren't using the dates.


Friday, 10 August 2001 Echunga –> Sydney
Top of page
previous day
next day
last day

More work on the Palm Pilot stuff today. There seems to be a complete lack of consistency in how things get done, and there are also some interesting (and sometimes worrying) bugs in jpilot, which doesn't seem to be compatible with pilot-xfer. Didn't have time to investigate that before leaving for the airport to go to the AUUG exec meeting.

Spent some time in the afternoon on IRC helping Rasmus Lerdorf debug a new Apache module. It eventuated that it needed both the Kerberos and the glib libraries, and that each of these packages define a function called g_strdup. What a mess, especially since the glib version does nothing that plain strdup doesn't.

In Sydney met up with the AUUG committee members and spent far too much time talking, getting to bed really late.


Saturday, 11 August 2001 Sydney
Top of page
previous day
next day
last day

Another AUUG exec meeting. What can I say? It's all in the minutes. The meeting was in the Carlton Crest Hotel, where we will be holding the annual conference, so we did spend a bit of time looking at the facilities and planning our wireless network. We're fortunate in being in line of sight from the University of Technology, Sydney, so we won't have any particular cost involved with setting up the network.

In the evening into the to look for a suitable venue for the Speaker's reception. Round 9 pm this became a problem: they have bouncers outside the pubs, and they objected to my footwear. I feel particularly insulted because they're the kind of place I wouldn't normally want to go to in the first place.


Sunday, 12 August 2001 Sydney –> Echunga
Top of page
previous day
next day
last day

Off to the airport with Warren and David this morning, and spent some time looking at the wireless network in the Qantas club. As usual, there was no other traffic, but I'm pleased to see that they're now doing authentication with https, not http.

Back home, had plenty of work to do to get the call for the AGM out the door. After that, had some friends of Yvonne, Essey and Mark Deayton, in for dinner. Got an early night notwithstanding; Yvonne has a cold and isn't feeling the best.


Monday, 13 August 2001 Echunga
Top of page
previous day
next day
last day

The PC is 20 years old today! Strangely, not many people mentioned it.

Monday is also the day I get my statistics from Telstra. Code Red has really made a difference:

Telstra network traffic

Plenty of administrative work today.


Tuesday, 14 August 2001 Echunga
Top of page
previous day
next day
last day

Another day where I didn't get much done. The trouble is that when I drop everything to get some other job done, mail piles up to the point where it takes me all day to attend to it.

Got a message from Per Abrahamsen. He has included me in the Free Software Celebrities list. I'm currently number 23.


Wednesday, 15 August 2001 Echunga
Top of page
previous day
next day
last day

More admin stuff today. Also started getting round to bringing some of my machines up to date.

In the evening Yvonne arrived with new hardware. monorchid, one of my test machines, is becoming the new battunga, since the old one had problems with the serial ports, and the CPU fan was getting noisy. That machine will join the pool of test boxes, and the new machine will replace monorchid. I chose the cheapest I could find, which cost $595 (Australian; about $300 US). 850 MHz Duron, 128 MB memory, 8 MB Sis 6326 display card (with the somewhat inappropriate name “Sparkle”) and a RealTek 100 MB/s Ethernet card. Only a year or two ago that would have been a high-end configuration.


Thursday, 16 August 2001 Echunga
Top of page
previous day
next day
last day

In catchup mode today, things that I should have done months ago. First into town to upgrade the AUUG Web Site. On 12 May Luigi gave me a 10 GB disk for the machine, and I still hadn't got round to installing it. Made pretty good time: I had parked in a half-hour parking spot and it didn't expire.

Then down to the IBM office and finally read my Notes mail. I've been with the company 3½ months, and this was the first time I got round to it. It's pretty painful across the net: I must install it on my laptop.


Friday, 17 August 2001 Echunga
Top of page
previous day
next day
last day

After yesterday's experience with remote mail access, I decided it was finally time to set up mail and dialup PPP on my Microsoft-based laptop. That took all day and was about as frustrating as I might have expected. At least I got it working, but I can't understand why vendors cater for novices to such an extent that it makes it really painful for experts. I estimate that the user interface alone makes it 10 to 30 times slower reading mail.


Saturday, 18 August 2001 Echunga
Top of page
previous day
next day
last day

A quiet day today. Finally got down to doing some work on my paper for the AUUG2001 Conference next month. Also had a bit of a play round installing OpenBSD on the old machine that used to be battunga, which in an OpenBSD context I have called blowfish. I'd welcome suggestions for a better name. And because that wasn't enough, I decided to start porting Vinum to NetBSD, making at least superficial progress.


Sunday, 19 August 2001 Echunga
Top of page
previous day
next day
last day

More work on my paper. Also decided that it was ridiculous to think of porting Vinum to NetBSD when it's been over a year since I looked at the root file system option on FreeBSD, so started working on that instead.


Monday, 20 August 2001 Echunga
Top of page
previous day
next day
last day

Another day mainly occupied with administrative stuff. I'm finally getting my act together at work.


Tuesday, 21 August 2001 Echunga
Top of page
previous day
next day
last day

Yet another day with administrative stuff. It's been so long since I have worked for a Real Company that I had forgotten the different pace. It's also been a long time since I've had to work with Microsoft. I wish it could have stayed this way. It's really frustrating to have the computer tell you what it will let you do, rather than you telling the computer what to do. At least I seem to have finished it now; another few hours for my expenses and I'll be done.


Wednesday, 22 August 2001 Echunga
Top of page
previous day
next day
last day

Finally getting rid of the admin stuff; it only took a week.

In the afternoon back to the NetBSD cross build environment. It works a lot better than I thought, though we're not there yet.

Got my paper for AUUG2001 finished. Ahead of time, too.


Thursday, 23 August 2001 Echunga
Top of page
previous day
next day
last day

Winter is coming to an end. A rather surprising event brought that home; at breakfast we saw Pebbles, our old Haflinger/Thoroughbred mare, lying down on the ground with a couple of crows pecking at her rump. She didn't look to be in any pain, but considering her condition and the fact that crows are carnivorous, we checked anyway. Pebbles is losing her winter coat, the crows are building nests, so they helped her shed her coat and used it for the nest.

Spent some time working on the NetBSD PPC port today, and finally ended up with a 32 bit userland and kernel which may run on an Apple Mac. Now that I have things in place, I need to decide on where to go next. At the moment it looks like:

  1. Modify the code for the RS/6000 environment.
  2. Boot from NFS.
  3. Boot from disk.
  4. Modify for 64 bit mode.
  5. Port to FreeBSD.

One way or another, it looks like I'll have plenty of work ahead of me.


Friday, 24 August 2001 Echunga
Top of page
previous day
next day
last day

More work on NetBSD PPC today. After a talk with Paulus, got down to comparing the Linux and NetBSD initialization code. Linux can build a kernel which can run on both Macs and RS/6000s, and that's clearly the way we need to go in BSD as well.

Jorge de Moya and the Yeardleys came in the evening with Estrella Negra; Chris and Yvonne have done a deal, and Estrella is now back with us. Good dinner, but I had to go to bed early.


Saturday, 25 August 2001 Echunga Images for 25 August 2001
Top of page
previous day
next day
last day

Up in the middle of the night (well, 7 am) this morning to go to this year's Installfest. Took me until after I got there to properly wake up.

The Installfest fell far short of last year's event. We're still trying to work out why, but the most likely culprit is the weather: it rained heavily all day, and only about 40 installations were performed, compared to well over 100 last year, though we had about the same number (90) of pre-registrations. I suppose people would seriously reconsider carrying their precious hardware through the pouring rain.

Did my BSD talk, which seemed well received. Considering the number of attendees, the 25 people for the talk sounded like quite a number.

Today was the 10th anniversary of the announcement of Linux, and somebody baked a cake. Since I was the only certified kernel hacker at the event, I was asked to cut the cake. I was about to protest that I was a BSD hacker, not a Linux hacker, until I realized the potential joke value:


This should be penguin-4.jpeg.  Is it missing?
Image title: penguin 4          Dimensions:          1600 x 1200, 576 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 August 2001, thumbnails          All images taken on Saturday, 25 August 2001, small
Diary entry for Saturday, 25 August 2001 Complete exposure details

 

 


Sunday, 26 August 2001 Echunga
Top of page
previous day
next day
last day

Finally the pressure is letting up a bit, so I didn't do much computer work, apart from building a couple of new images on sydney. PCMCIA seems to be picking up, which is a good thing in view of the fact that 4.4 is about to come out. Also another NetBSD/PPC system. There's a bug in the compiler that causes it to die horribly trying to compile sbin/fsck_ffs/pass5.c.

Apart from that, played the bassoon until my arms hurt, did some Indian cooking and generally tried to pretend that I have a life.


Monday, 27 August 2001 Echunga
Top of page
previous day
next day
last day

More work on the NetBSD port today. NetBSD seems to have a maze of twisty little PPC ports, all the same. Sent out a message to the lists and asked for feedback. The only reply I got was a personal one from within IBM.


Tuesday, 28 August 2001 Echunga
Top of page
previous day
next day
last day

More work on the PPC port today. About the only feedback I got was a question about why I was porting to NetBSD and not to FreeBSD. The answer is clear, of course, but in view of the fact that the NetBSD community is completely silent, I wonder if it wouldn't make more sense to port to FreeBSD instead. Decided to do both at the same time, and spent the day putting comments into locore.s, with the advantage that I learnt more both about what it was doing and about PPC assembler.


Wednesday, 29 August 2001 Echunga
Top of page
previous day
next day
last day

I was planning to go to the ADUUG lunch today, but didn't feel too well and cancelled after all. Spent the rest of the day with locore.s, which is liable to keep me busy for some time to come.

In the evening with Yana to a rehearsal of the Mount Barker Concert, basically a brass band. The bassoon looks a bit funny in there, but they had parts for it, and although I offered an Eb clarinet for next time, it's not clear that they'd prefer that. Yana enjoyed herself greatly.


Thursday, 30 August 2001 Echunga
Top of page
previous day
next day
last day

More work on the PPC ports today. Spent a lot of time analysing the code differences and came to the conclusion that there was enough dubious code in the fledgling FreeBSD port that it would be wise to do the NetBSD port first after all. At least I know why now.

People are talking about cheap satellite Internet access lately. Considering that I probably won't be able to get DSL in any reasonable time frame, it's an interesting alternative, especially since I'm now downloading significant quantities of code. Downloaded the remainder of binutils and gcc today, which kept my modem busy.


Friday, 31 August 2001 Echunga
Top of page
previous day
next day
last day

Once again I have some source tree manipulation to do, both the NetBSD PPC work and also the binutils and gcc stuff. Years ago I wrote a program, mklinks, that did various things to replicate source trees (and thus causing the tar performance problems that I fixed a few weeks ago). Unfortunately, mklinks had its own issues, one of which was the ability to get stuck into a loop when unravelling a directory symlink pointing to a directory up the path from where it was located. Also, it contained a maze of twisty little options, all different, and even after years I kept wondering which ones to use. Decided to coalesce things and create three “macro” commands, clone (which creates a new hierarchy with file links to the old one), update (which updates files in the new hierarchy) and delete (which removes files in the destination hierarchy if they're present in the source hierarchy). That seems to work, and I've at least partially solved the symlink loop issue, but I'm not done yet.


Top of page Previous month Greg's home page Today's diary entry Next month Greg's photos Copyright information

Valid XHTML 1.0!

$Id: diary-aug2001.php,v 1.21 2016/05/04 00:08:35 grog Exp $