|
| Greg's diary |
| April 2005 |
|
|
||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
|
|
|
|||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| Friday, 1 April 2005 | Echunga | |
| Top of page | ||
| next day | ||
| last day |
As usual, Friday's a meeting day, and didn't get much done. Managed to get some graphs of my I/O performance testing done; they showed things completely at variance to what I had seen earlier: stream I/O was the slowest, system call I/O with O_DIRECT was the fastest, and aio was in between. With the exception of the last, this is approximately what I had expected, but it's upsetting that I don't know why I came up with such completely different results earlier. Apart from that, these tests showed effectively no change in throughput for differing numbers of threads or concurrent I/O operations. There's something funny there. More testing required.
On the way home, dropped into Grumpys to pay for some malt and hops that Yvonne had picked up. It was pretty quiet; only one bloke there—the German/Canadian one from the Neanderthal whom we had met in William Creek and Coober Pedy last week. Sure, I expect to bump into people again and again in the Outback, but it was a little surprising here.
| Saturday, 2 April 2005 | Echunga | Images for 2 April 2005 |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
Brew day again today, which meant that I didn't get much else done. Tried a few new things:
For mashing, instead of heating between rests, started with a smaller quantity and added boiling water. This was not a success. The initial mash was far too thick, and after adding the boiling water the temperature looked too high; but by the end of the mash it dropped too low, presumably because the malt took its time to cool things down. The resultant wort was fully 10% weaker than by my normal mash method.
In the second brew, used the first of my home-grown hops. This too proved to be an issue: the hops clog up the racking tube. I'll have to think of a solution to this by the next brew day.
In addition had problems with my temperature control stuff. One of the sensors seems to have contact problems, and the software is not handling them well:
It's the red line in the graph, of course. Time for some programming.
| Sunday, 3 April 2005 | Echunga | |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
Today was supposed to be a quiet day, but it didn't quite work out that way. Spent some time looking at the problems with my temperature control system, and finally found that the processor on the temperature monitor board was badly seated in the socket. This is after nearly a year of operation, so I suppose the relatively moist environment is to blame—and the choice of using a socket in the first place, of course. Also spent some time working on the software to ensure that it can handle such problems in a better manner.
| Monday, 4 April 2005 | Echunga | Images for 4 April 2005 |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
Back to work on obelix today, and spent an inordinate amount of time installing Fedora Core 3 on it. Why do I always have trouble with CD-ROMs on this platform? It has happened at least once every time I've tried it, with at least four different drives, disks from multiple sources, and different versions of Fedora. Today it ran into problems just before the end of the installation: it got an error 5 (presumably EIO, but it didn't say that) on a specific package. It gave me only one choice: retry. After doing that about 10 times, I gave up and discovered that I could remove the CD and copy it. Did that, and got the same error on the new one. Presumably there's an error pointing out of the address space of the CD. It wouldn't have been an issue if I had been able to ignore it, but as it was, I had to restart the install.
Why is it that if you tell Fedora to install everything, it uses CDs 1, 2 and 3, but if you use a standard install it uses all four?
Got obelix up running Fedora; I think I'm going to have to rename it, though. Otherwise I'll run into all sorts of problems with ssh keys.
After that, returned to my issues of last week with obelix and 8 GB memory. Somebody had suggested memory problems, so started running Memtest86+, a standalone (bootable) memory test program. That was well worthwhile: I was running without ECC, and it found thousands of errors even with only 4 GB. Running with all four chips (8 GB) seems to be beyond the capacity of the system. Even with ECC enabled, got spontaneous reboots out of the memory test program. To add to the confusion, some people claimed that the parts weren't even correct. Photos here.
| Tuesday, 5 April 2005 | Echunga | Images for 5 April 2005 |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
Ben Henderson, the gardener, along today and did some more work on the flower beds we're having put in the garden. One result was that we have to move a sprinkler, in the process exposing the cables and pipes that went past the old location:
|
|||||||||||||||||||||||
On the work front, continued with my memory testing today. By morning I had hit about 40,000 correctable errors, and tried again with 8 GB. That didn't get off the ground: the machine kept rebooting spontaneously. It looks as if the motherboard just can't handle the memory. Did some discussion on IRC, but came to no good conclusion. Finally inserted the memory which I had bought for this machine, and it worked perfectly. So I don't seem to have a serious problem, but the moral of the story is, I suppose, “don't buy the MSI K8T Master2-FAR”. I was warned about the motherboard before I bought it, but at the time it was the only game in town. Now I know of other places with (hopefully) better boards.
Also evaluated the no-load tests on the 14 disk array. They were pretty much exactly the same as under load:
| User time, no load | User time under load |
|
|
| System time, no load | System time under load |
|
|
| Elapsed time, no load | Elapsed time under load |
|
|
In each case, the number of threads seemed completely irrelevant, suggesting that the requests to the disks were being serialized. Checked the debug output from vinum info and found confirmation, though it's still not clear what's causing it.
After that, on with trying to run the same tests under Linux. That took some time too: had trouble getting the kernel to recognize two Symbios host adaptors, and ended up running with only 7 disks. And even then I couldn't get the aio stuff working—probably require to download a package—and when I ran it, it crashed in the thread code:
Program received signal SIGABRT, Aborted. [Switching to Thread 182894173856 (LWP 5021)] 0x0000003320e2e4dd in raise () from /lib64/tls/libc.so.6 (gdb) bt #0 0x0000003320e2e4dd in raise () from /lib64/tls/libc.so.6 #1 0x0000003320e2fc8e in abort () from /lib64/tls/libc.so.6 #2 0x0000003320e67868 in malloc_printerr () from /lib64/tls/libc.so.6 #3 0x0000003320e68406 in free () from /lib64/tls/libc.so.6 #4 0x0000003320c0d4e9 in _dl_deallocate_tls () from /lib64/ld-linux-x86-64.so.2 #5 0x0000003321f05756 in __deallocate_stack () from /lib64/tls/libpthread.so.0 #6 0x0000003321f06fcb in pthread_join () from /lib64/tls/libpthread.so.0 #7 0x0000000000400cff in run_test () at pluto.c:137 #8 0x0000000000400da5 in main (argc=4, argv=0x7fbffff228) at pluto.c:167
Plenty of fun.
| Wednesday, 6 April 2005 | Echunga | |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
Orkut has a new service: a CSV-separated list of “friends” and their (real) email addresses. That sounded like it was useful, so I used firefox to download it. The result shows me yet again how terrible the modern “Desktop” paradigm is. After downloading, I got the following window:
I'm getting to hate this abuse of the existing term open to mean “display”. In this case I wanted to save it to disk, but just out of interest, I selected Browse. What I got was:
In other words, the application is looking for executables in my home directory. I have no idea what the yellow date (written the wrong way round) is, but it seems completely unaware of the concept of a PATH environment variable, even though Microsoft had it 20 years ago. It also separates the directory name part and the file name part of the pathname, and it displays in a window so tiny that not a single line can be completely displayed, and only displays three lines. This is on a monitor with 2048x1536 resolution. It is possible to specify the pathname in the File name field, but of course it needs to be expanded to be even remotely usable.
I wanted to save to disk, of course, so I selected that option and got:
Where did it put the file? ..., it would seem. Expanding the (again) too small window showed:
Well, that's something, but not very much. Where's the pathname? Where is Desktop? The answer is: nowhere. I don't have a directory or file called Desktop:
=== grog@wantadilla (/dev/ttypc) /home/grog 79 -> locate /Desktop
/home/grog/Documentation/Data-acquisition/ds1820k/Desktop
/usr/X11R6/include/Xm/DesktopP.h
/usr/X11R6/share/gnome/apps/Settings/Desktop
/usr/X11R6/share/gnome/apps/Settings/Desktop/.keep_me
/usr/X11R6/share/gnome/control-center/Desktop
/usr/X11R6/share/gnome/control-center/Desktop/.keep_me
I did have sufficient distrust to check that it didn't store the file in the /usr/X11R6 hierarchy, but it didn't.
So where did the file Contacts-1.CSV go? Used ktrace to help me there:
1351 firefox-bin NAMI "/home/grog/contacts.CSV" 1351 firefox-bin RET access 0 1351 firefox-bin CALL access(0xaa67ac8,0) 1351 firefox-bin NAMI "/home/grog/contacts-1.CSV" 1351 firefox-bin RET access 0 1351 firefox-bin CALL access(0xaa67ac8,0) 1351 firefox-bin NAMI "/home/grog/contacts-2.CSV" 1351 firefox-bin RET access 0 1351 firefox-bin CALL access(0xaa67ac8,0) 1351 firefox-bin NAMI "/home/grog/contacts-3.CSV" 1351 firefox-bin RET access 0 1351 firefox-bin CALL access(0xaa67ac8,0) 1351 firefox-bin NAMI "/home/grog/contacts-4.CSV" 1351 firefox-bin RET access -1 errno 2 No such file or directory 1351 firefox-bin CALL open(0xaa67ac8,0xe01,0x180) 1351 firefox-bin NAMI "/home/grog/contacts-4.CSV" 1351 firefox-bin RET open 22/0x16 1351 firefox-bin CALL close(0x16)
So it put it in the home directory. While that's a good idea from my perspective, how can I possibly know from this ridiculously obfuscatory window? Instead of a window in the first place, which I have to then explicitly select with the mouse and then close, wouldn't the following simple message on the status line be so much better?
File saved as /home/grog/contacts-4.CSB
Of course, as kdump shows, the file name wasn't Contacts-1.CSV at all. firefox gets its list backwards, so the name it chooses is placed at the top of the list. I really can't understand why so many people use this reverse listing. I wonder how long it's going to be before we're going to have to start reading pages in books and newspapers from bottom to top.
Spent some more time testing my test program on asterix, and spent far too long discovering the cause of the error, which despite my expectations had nothing to do with 64 bit processor: it was a simple pointer address calculation error. Got very different results from FreeBSD, though: here we did get significant throughput improvements from an increased number of threads. It's too early to post results, though: first I need to understand what is going on on FreeBSD. And for that I need to upgrade to a newer version of the system, not made easier by a decision to make source updates from older versions impossible: instead I need to do two back-to-back updates.
| Thursday, 7 April 2005 | Echunga | Images for 7 April 2005 |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
Spent more time than I should getting asterix up to date on Linux, and didn't get round to doing any testing as a result. Also thinking about the hardware issues. Tim Stoakes has been having his own: his new ultra-fast SCSI RAID array is much slower than the IDE array—some of the time. Spent some time looking at that, but didn't come to any useful conclusions.
| Friday, 8 April 2005 | Echunga -> Murray Bridge -> Avoca Dell landing | Images for 8 April 2005 |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
Off on a long weekend this afternoon, so spent the morning wrapping up the week's work—not too spectacular—and preparing for the afternoon. Then down to Long Island Marina at Murray Bridge to pick up the houseboat “Breakaway” and head upriver for the weekend.
That wasn't as easy as it sounded. We arrived at 15:00, but spent nearly 2 hours mainly waiting for the people to get round to doing the formalities. They didn't have any paperwork, and we didn't even get a receipt for the cash deposit of $500 that we had to leave behind. Got some reasonable information on the boat and navigation, but also a lot of information we didn't need—how to use a gas oven, for example. Also discovered a number of things wrong with the boat: the ignition for the barbecue didn't work, the gas heater was missing a battery, and the fridge worked so badly—we were told—that we should put everything in the freezer, which didn't really freeze.
Finally got away after a test run round the river, and off north to Avoca Dell camping place, all of 8 km away, discovering on the way that the CD player didn't work properly. Managed to make do with the DVD player, which wasn't connected to the stereo speaker system. Made it by nightfall, when the DVD player and TV failed: we discovered that the inverter-supplied power points didn't work: a connection to the inverter, which Gary had said wasn't needed, had to be connected.
Managed to have dinner and early to bed, somewhat tired, and not helped by the noisy inhabitants of one of the other boats, who continued surprisingly loud TV programs until 1 am.
| Saturday, 9 April 2005 | Avoca Dell landing -> Mannum | Images for 9 April 2005 |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
Up this morning after very little sleep to discover that all our food was frozen, along with the beer I had put in the “freezer”. So much for Gary's advice.
Set off round 8 am with intention to get to Mannum by lunchtime. Made very poor headway upstream. We had been told yesterday that the boats would do 10 to 12 knots; nevertheless they had expected it to take 4 hours to get to Mannum, 40 km away, which makes it sound more like 10 km/h or about 5½ knots. In fact, though, we couldn't even make this speed; at the recommended 2800 rpm my GPS receiver showed at one point that our speed had dropped to 1.5 knots, or less than 3 km/h. I attributed that to the wind, but on continuing with the wind behind us, it didn't seem to make much difference. Came to the conclusion that the boat is completely underpowered (only a single 90 hp motor).
Somehow the whole thing seems a little slow, clumsy and unexciting. I'm remained of the “Nautical Yarn” that I learnt decades ago, and which for some reason doesn't seem to be available on the web. I don't have them written down, but I recall:
I sing of a captain who's well known to fame,
A naval commander, Biil Jinks was his name,
Who sailed where the Murray's clear waters do flow,
Did this fresh-water shellback with his “Yo-heave-a-yo”.To the port of Wahgunyah his vessel was bound,
When night came upon them and darkness around.
Not a star on the water its clear light did show,
But the vessel sped onwards with a “Yo-heave-a-yo”.“Oh captain, oh captain, let's make for the shore,
For the winds they do rage and the waves they do roar”.
“Nay, nay” said the captain, “though fierce winds may blow,
I will stick to my vessel with a Yo-heave-a-yo”.“Oh captain, oh captain, the waves sweep the deck,
Oh captain, oh captain, we'll soon be a wreck,
To the river's deep bosom each seaman will go”.
But the captain laughed loudly with a “Yo-heave-a-yo”.“Farewell to the maiden, the girl I adore,
Farewell to my friends, I will see them no more.
The crew shrieked with terror, the captain he swore:
They had stuck on a sandbank, so the men walked ashore.
Wahgunyah is a long way away. In fact, apart from the strong north wind, the river is slow-moving and gentle, and the only waves are caused by passing speedboats. And though it contains a number of sandbanks, there are none between Murray Bridge and Mannum.
Despite the slow headway, it was tiring work. Steering a boat of this size and with such little power is a slow business, and changes at the wheel can take 20 seconds to take any effect. Nevertheless, we couldn't leave the wheel for more than 10 seconds at a time: the steering drifts, and it's not made any easier by the linkage to the motor. Before leaving I put a rubber band around a spoke which would represent the “straight ahead” position, but by the evening it had rotated 5 times to the port. Certainly navigating a houseboat wasn't the calm, relaxing thing that I expected.
An attempted landing at Caloote was a disaster: the wind blew us around and grounded us on a boat ramp, and though it was so gentle that we didn't even feel the boat ground, the motor wasn't strong enough to dislodge us. A couple of speedboats came along to tow us out, but almost before we got the ropes on, the current dislodged the front of the boat, so in the end only one boat (driven by Walter, surname unknown) was needed.
Pushed on to Mannum, where we had a number of problems. The first was just to recognize the mooring grounds. The shore at Mannum is divided into a number of sections, and though we had a book (“Murray River Pilot”, with not-to-scale charts of the river and details of Mannum, it was impossible to locate the houseboat mooring points until we had nearly berthed: the signs are so small that they can't be read from more than about 20 metres. Finally found a location and tried to moor there, but the wind blew us around again, and we smashed into another houseboat, Jubilee 1—coincidentally, it seems, the one that made so much noise at Avoca Dell last night. Finally gave up trying to moor there and moored at the private point of the Pretoria Hotel, which was reserved for patrons. That's OK: we wanted to have dinner there, and that proved to be enough for them to let us stay there for free.
Then back (100 metres or so) to find out what damage we had done. Not much: we scratched an attached speedboat, and it'll need a polish. Also bent a mooring bracket on the Jubilee, but that can presumably be bent back again. About the only real damage done was that in the excitement Yvonne slipped while tying up the boat and scratched her shins and bruised herself rather nastily.
Dinner in the hotel as planned, not too bad. Spent some time talking to my father and relatively late to bed despite the eventful day.
| Sunday, 10 April 2005 | Mannum -> Murray Bridge -> Echunga | Images for 10 April 2005 |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
As I had feared, mooring outside the pub wasn't the ideal place. Once again we were kept awake until about 0130 by noise. So much for relaxation! We'd have been much better off at home.
Apart from that got a reasonable night's sleep, and up quite early with the intention of getting back to Murray Bridge this afternoon: we've seen what we could, and upstream from Mannum there's nowhere to stop for another 18 km, beyond our weekend's range.
Left early as a result, even before breakfast about 7:20. The weather had changed overnight, and there was no wind, giving me a chance to measure the speed of the boat. Flat out it will do about 9.1 km/h, but normal speeds are more like 6 km/h.
Made fairly steady progress, stopping nowhere until we got back to Long Island Marina, which took us just over 5 hours. The weather deteriorated, and we were happy when we finally got there.
Back home and did little in the afternoon.
| Monday, 11 April 2005 | Echunga | |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
Mondays are becoming a horror, not because of the fact that it's back to work, but because of the kind of work: reading 3 days of backed-up mail messages. Some of those were some ideas that people came up with last Friday afternoon, which required a significant effort to address.
Apart from that, spent some time chasing my ADSL connection, which is no closer to reality than it was a couple of weeks ago. Internode is waiting for the DSLAM to be installed in the exchange, and Telstra showed an unbelievable lack of knowledge: the representative I spoke to didn't know that the lines need to be checked for suitability for ADSL. The net result, anyway, was the same: nothing will happen until next week. I can see myself being on dialup for a while as a result.
| Tuesday, 12 April 2005 | Echunga | |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
Somehow today was full of interruptions, starting when Steve from Meadows Landscape came in with 18 cubic metres of pine mulch. Ben Henderson should have been there to collect it, but his car had broken down, Yvonne was busy elsewhere, so I had to organize the things. It didn't take long, but it started off a trend followed up by long technical discussions on the phone about work issues not directly related to the work I was trying to do (further I/O performance testing). Made a couple of variants on my I/O test programs (to achieve parallelism with processes instead of threads), but didn't have time to start any tests: first I needed to rebuild my hardware.
| Wednesday, 13 April 2005 | Echunga | |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
First thing this morning I planned to start more tests with my process version of the test programs. Things didn't quite work like that: first I had hardware problems (probably a disk not seated correctly in the tray), and when I got that working, the system panicked, repeatedly but not always in the same place, though usually out of GEOM. Looks like another memory allocation error. Tried to take a dump and discovered that I had neither dumpdev nor swap, so did a bit of reconfiguration to get that.
FreeBSD kernel debugging is still broken! That's been nearly a year now. The ddb panic command still doesn't work, so I had to use call doadump. After finally getting a dump, I couldn't find a version of gdb that wanted to look at it. Things are a real mess.
To stop me making any further progress on that front, more long phone calls. It looks like I'll be taking a different tack anyway: it seems that performance is no longer our number 1 concern.
Yana back home today for the first time in over 2 months. She's had an interesting time in Europe and then back here.
| Thursday, 14 April 2005 | Echunga | |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
Another day with little to show in the way of results. Started upgrading quartet to the latest FreeBSD-5 release; at least by the evening userland had built. I suppose it's a sign of the times that even an old machine with only 256 MB of RAM shared between four processors hardly swaps when doing a make world -j8.
More mail messages flying around. Sometimes I think that face-to-face meetings might make more sense. They certainly kept me busy.
To town in the late afternoon to a board meeting of the ICT Council of South Australia. At least there I was able to present a new membership application form.
| Friday, 15 April 2005 | Echunga | |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
My father left for England today; I wonder when we'll see him here again. The flight was in the morning, so Yvonne took him in, experiencing a tremendous traffic jam at the airport in the process: it took her nearly an hour to get from the airport entrance to the checkin. I wonder if they still have these “Relaxing SA” advertisements on the way.
Meeting day today, so got little else done. We now have ADSL 2+ in the office, and Kim Needham has a problem with the interaction between the DSLAM and FreeBSD. It seems that the DSLAM will sometime forget the MAC address of the other end, and no traffic will remind it. On changing addresses (connect a new machine to the modem, for example), it forgets the old address and doesn't respond to traffic from the new address. Strangely, Microsoft and Linux don't have that problem. Things seem to work OK if the ADSL modem is reset. I know that Linux generates a lot of spurious ARP traffic, and Microsoft seems to generate a lot of spurious traffic in general. I wonder if the DSLAM is looking for something in there. To be investigated further.
| Saturday, 16 April 2005 | Echunga | |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
Even by recent measures, today was a very quiet day. Bottled some beer in the morning, and in the afternoon spent some considerable time watching documentaries on TV.
One of them was extraordinary: it seems that some people have made quite a bit of money selling a book claiming that the terrorist attacks of 11 September 2001 never happened. This documentary tried to refute the claims—and failed. Here's a transcript, made by somebody who doesn't seem to be overly critical.
It seems that the original photo of the damage to the Pentagon was not consistent with the claimed hit by a large passenger plane. The documentary spent most of the time trying to denigrate the authors, including pointing to their extremist background. Mostly they were thought to be neo-Nazi, but the US ambassador in Athens claimed that they belonged to “left-wing elements”—that must be newspeak for “Commie”. Not a very convincing defence. About the only thing the documentary didn't address adequately were the facts: why did the photos not show more damage? One “expert” explained how planes can fold up to almost nothing. But that's not the kind of impact that was made on the World Trade Center. And claiming that they're just conspiracy theories isn't good enough: there might be a new one soon claiming that Iraq didn't have weapons of mass destruction before the Americans invaded them a couple of years ago. While I consider it highly unlikely that the claims are correct, the lack of firm refutation makes it more likely that people will believe it. This kind of documentary is fodder for conspiracy theories.
| Sunday, 17 April 2005 | Echunga | |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
Another quiet day. Finally managed to go riding again, though both Darah and I were not in the best of condition. As I say just about every time I go riding: I should do this more often. Still, had a pleasant time. Apart from that, trying to catch up on reading. I'm several weeks behind reading c't, and there's good stuff in there that I should know about.
| Monday, 18 April 2005 | Echunga | |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
My mail gauge shows that I'm getting overworked again: I now have over 2000 unprocessed messages again. One of the reasons for that is the change of emphasis at work, which requires a lot of preparation and yet another meeting. I had asked for the meeting, intended to discuss design approaches for some of the issues facing us. When I got to the meeting, it turned into a design review of one specific potential implementation, one with which I wasn't overly happy. To make matters worse, it lasted all afternoon. What frustration!
| Tuesday, 19 April 2005 | Echunga –> Canberra | |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
Off to linux.conf.au today, but first had lots of things to do, and didn't complete. Things weren't made any better by the fact that I had somehow ended up with the incorrect itinerary for my flight to Canberra—one intended for Sunday, the day before yesterday. By the time I found out, my real flight had left: the first time I have ever missed a flight, and I was suitably pleased. Managed to get another (in fact, better) flight, but I was down $270 as a result.
Got to Canberra in the early evening and found I was staying in the Quest apartments, the same place that Yvonne and I stayed in December 2000. Fortunately this time I was at the other end of the building, away from P.J. O'Reillys, and it wasn't too noisy. I had intended to meet up with Daniel Phillips, but he was at the speakers' dinner, so went off looking for an Indian restaurant I knew from some time ago, in the process going past places that I hadn't been to for five years. How time flies! And isn't it surprising how I always end up in almost exactly the same part of Canberra?
Then to the Wig and Pen, where I was joined by a passing Anthony Rumble and Grant Parnell, both of ELX, and discovered that ELX is offering cheap Hauppauge PVR-250 cards, along with complete kits for building digital video recorders. Had an interesting discussion about the issues; this looks like a good topic for a BoF session.
| Wednesday, 20 April 2005 | Canberra | Images for 20 April 2005 |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
Up early this morning to look for breakfast at the same place where Yvonne and I went on 2 December 2000. Yvonne had been very happy then; I was not today, getting food only barely better than my first time in William Creek:
|
|||||||||||||||||||||||
I complained, and they called the cook out. I wouldn't have been quite so upset if he had apologized and offered to bring a better one. Instead, he told me that many people like it like that. Gave up on that and to the University, where I got a reasonable breakfast for half the price.
After the introduction, had intended to listen to Rusty's tutorial, but my 2000 unread mail messages came back to bite me: one of them had been a list of prerequisites for the tutorial, including a laptop running Linux and qemu. Tried to download that, but my network timed out. On attempting to reconnect, I got:
=== grog@adelaide (/dev/ttyv0) ~ 1 -> ssh wantadilla.lemis.com
@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@
@ WARNING: REMOTE HOST IDENTIFICATION HAS CHANGED! @
@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@
IT IS POSSIBLE THAT SOMEONE IS DOING SOMETHING NASTY!
Someone could be eavesdropping on you right now (man-in-the-middle attack)!
It is also possible that the DSA host key has just been changed.
The fingerprint for the DSA key sent by the remote host is
4e:11:3d:1e:25:08:37:59:f4:0b:bb:a2:9e:6e:b6:42.
Please contact your system administrator.
Add correct host key in /home/grog/.ssh/known_hosts to get rid of this message.
Offending key in /home/grog/.ssh/known_hosts:49
Password authentication is disabled to avoid man-in-the-middle attacks.
Password:
That looked strange: why disable passwords and then ask for one? It didn't accept my passwords either. Then it dawned on me:
=== grog@adelaide (/dev/ttyv0) ~ 2 -> traceroute wantadilla.lemis.com
traceroute to wantadilla.lemis.com.lemis.com (192.168.5.1), 64 hops max, 40 byte packets
1 192.168.5.1 (192.168.5.1) 2.626 ms 1.670 ms 1.574 ms
That's not the address of wantadilla.lemis.com. I'd been had by a trick that I had used myself over 15 years ago. Rapidly out to change all my passwords, and after confirming that the laptop I was using (adelaide, normally called pain and running Microsoft “Windows” XP “Professional”) was too down-rev to install the qemu port, decided to stay outside.
Almost immediately Tridge came along and introduced me to Eben Moglen. There's a lot of talk about the current fiasco with Bitkeeper, about which I was brought up to date in a hurry. That took a while, but it's surprising how I know just about everybody involved in the issue.
After that talking storage with Daniel Phillips, who has been working in a very similar area to me. Compared notes and discovered that he has solved the same problems in a surprisingly similar way (or, in some cases, has planned to do so). He's also working on a file system loopback module which looks like one of the things I was planning to do, so there's a lot of synergy there.
In the evening to the Wig and Pen to have a beer and watch the sun go down—the weather has been very pleasant lately—and was joined by Con Zymaris and a friend, who dragged me along to the AGM of OSIA. Didn't have time to wait for the end of that: had planned to go out with Martin Pool, Benjamin Herrenschmidt and others. Ended up back at the Wig and Pen. I've decided that there's something Evil in the way British-style beer glasses are full when they're about to overflow. The result is a complete lack of attention to the head. The Wig and Pen use these glasses for everything, resulting in an undercarbonated, headless Pilsener. They're currently out of Weißbier, and under the circumstances that's probably just as well. Not too late to bed.
| Thursday, 21 April 2005 | Canberra | Images for 21 April 2005 |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
Over to the university early in the morning, meeting up with Paul McKinney for breakfast. Time flies: we nearly missed Tridge's keynote speech, musings about software development. Somehow they were strangely relevant. At one point, talking about the upcoming release 4 of Samba, he produced a slide containing only:
Threads are evil
This was the first of three such slides: the others were “Processes are ugly” and “State machines drive you mad”. He went on to say that threads are for people who want to ignore that the system has an MMU, that it's like declaring all variables global: it makes things easy to get started, but it's not good software engineering practice. It replaces in software what is available in hardware, and it's always slower than processes.
I don't agree with Tridge here in all points. In particular, I can see many cases where threads must be faster than processes. But it's an interesting topic for consideration.
After that in to hear a talk by Jon Corbet about the 2.6 release of Linux; there seem to have been major changes in the 15 months since the last LCA. Along with the changes in revision control, it looks as if Linux is modifying its direction.
To lunch with Ben Elliston and Warren Toomey, spending most of our time talking toolchain, after which back to the afternoon sessions. First was Paul McKinney with what I think was an expanded version of his talk from last year. Unfortunately he had brought far too much material, and along with practical demonstrations (“An instruction is a sheet of toilet paper; a cross-processor synchronization is two rolls”) ran out of time badly. The demonstration was interesting, though:
|
|||||||||||||||||||||||
Next came Daniel Phillips, who also had problems with his presentation (about cluster RAID). I'm getting the feeling (which I can definitely confirm in my own case) that people no longer have time to do adequate preparation for presentations.
In the next session we had Martin Pool with Yet Another Revision Control System, called not yarcs but bzr-ng. Interesting approach, though it's still very much a work in progress. Strangely, both bzr-ng and Linus' git don't store diffs: they store complete copies of each revision. There's clearly scope for a redundancy elimination file system here.
Next, Warren Toomey talked about Oz-tivo. I was planning to go to that, but Warren confirmed that it was essentially the same talk as at the AUUG conference last year, so went to the terminal room and spent some time catching up on mail instead.
In the evening had the professional delegate's “networking” reception at the CSIRO Discovery Centre. Not the best place for such a reception; there's too much to look at, and the purpose of the reception was to get together and talk. We did that, of course, and had a good time. It's interesting to note how many professional delegates there are this year. Three years ago in Brisbane there was only a handful; this year we could barely move. I've been noticing over the years that the professional aspect of the conference has been on the increase, but the crowd this evening made it particularly obvious.
| Friday, 22 April 2005 | Canberra | Images for 22 April 2005 |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
Keynote this morning was by Andrew Morton, describing the Linux release process from his view. He also mentioned the Bitkeeper issue, of course, but with a difference: he doesn't believe that Linux needs a version control system; instead he wants something that handles multiple origins, and apparently he has something that goes some way towards this end.
Next talk was Peter Chubb, who, not for the first time, has been doing the same kind of work as I am. This time round it looks as if we might end up working together, to the mutual benefit of all.
After lunch, heard from Keith Packard on the future of X. Looks like X.org is finally getting some things moving again, and it looks like we'll be seeing a few changes later this year, notably X11R7. He, too, talked about version control. X still uses CVS, and it doesn't look like that will change in the near future. He doesn't think Subversion is interesting enough, and of course he doesn't have anything to do with Bitkeeper. All in all, an interesting talk, but it reminds me how long it's been since I have paid much attention to X.
Next was Robert Collins on distributed revision control. A complement to Martin Pool's talk yesterday, but it didn't seem to relevant to my particular interests.
Conference dinner in the evening, somewhat different from previous years: Rusty had arranged for entertainment both from himself and from other speakers; interesting to note that many have other talents, including Ted T'so and Bdale Garbee. On the other hand, it mean that we didn't get to talk much, and somehow we got a pretty late night anyway.
What I didn't mention was that Rusty was awarded the inaugural “Rusty Wrench” for (to quote The award is meant to honour people who have rendered outstanding service to the free and open source software communities in Australia), though I took a photo of the event:
Image title: Rusty Wrench Dimensions: 902 x 1399, 240 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 Friday, 22 April 2005, thumbnails All images taken on Friday, 22 April 2005, small Diary entry for Friday, 22 April 2005 Complete exposure details
| Saturday, 23 April 2005 | Canberra | Images for 23 April 2005 |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
Final day of the conference. The late night and party last night weren't a problem, but the drunken idiots shouting outside my bedroom window were. As a result, only barely made it to Eben Moglen's talk, which was excellent rhetoric. I agree on most of his points, though terms like “Citizens of the Free World” sound more political than technical.
Then listened to Anton Blanchard talking about bringing up Linux on a 128-way PowerPC system; issues with SMP contention become paramount, and it's really amazing that the thing worked at all.
Lunch was the traditional pizza, once more far too many for the participants. Spent it talking to Warren Toomey, Anthony Rumble and others about my idea for an end-user-usable multimedia centre. There's interest, but there's also a lot I need to understand as well.
In the afternoon spent time in the hall talking to Daniel Phillips, who asked me to review his solution to the final VM deadlock problem. It's interesting because it relates directly to our intention to loop back into user space. Eben Moglen was there again; somehow he and Daniel don't hit it off, and they ended up shouting at each other. It's not just Eben: I was able to discuss the matter of the upcoming revision 3 of the GPL with him sanely. And it's not Daniel, who's usually pretty quiet. But the combination caused fireworks on Wednesday, and it did it again today.
Finally bought some books from the bookstore stand. One of their best sellers has been a Linux book called The Complete FreeBSD:
|
|||||||||||||||||||||||
But there were other books there too, including one on Linux that might be the answer to my personal problems of not knowing where it diverges from UNIX.
Final session went on for ever. After that, out with all the speakers (a total of about 18 of us) to look for dinner, and discovered that getting a table at a restaurant in Canberra on a Saturday night is not a foregone conclusion. Finally off with Paul McKenney of IBM, Daniel Phillips of Red Hat, and Stephen Hemminger of OSDL and found some passable Indian food. During dinner, the conversation touched on Larry McVoy, who was almost the theme of the conference. I discovered that Stephen and Paul had also discussed Larry's SMP ideas with him, as I (on Paul's instigation) had done in February 2002. I had then felt that I hadn't given Larry a fair hearing, but it seems that Paul and Stephen came to a very similar conclusion.
| Sunday, 24 April 2005 | Canberra –> Echunga | Images for 24 April 2005 |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
Up late this morning, having again been kept awake by noisy drunks. Discovered, as I had feared, that no breakfast was to be found, and finally ended up getting a “kebab” (really a giros or döner kebab wrapped in bread) in a kebab house round the corner when they finally opened at 11 am.
Then tried in vain to hail a taxi, so back to the hotel and got them to call one for me. Spent some time talking to the proprietor, who has taken over the place about 7 months ago. They've obviously spent some effort on the place: though it was OK last time we stayed here, it used to look a little grubby, and they've tidied it up a lot. It seems that the issue with the noise from P.J. O'Reilly's is a long-standing issue, and that they'll probably get something done about it soon.
Off to Tridge's place—the taxi driver didn't know where it was and had to look it up in the street directory—and was rewarded for being one of the first arrivals by being given the task of cleaning and running the barbecue. Fortunately somebody more interested than I took over shortly after I started cooking. Pleasant afternoon, and though there must have been 30 people there, surprisingly few of the local crowd came.
| Monday, 25 April 2005 | Echunga | |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
Spent most of the day not catching up on the last week, but playing with the stuff I brought back from Canberra, in particular with an intention to installing Ubuntu, the latest version of Linux which people at the linux.conf.au have been migrating to in droves. Found a spare 60 GB 2½" drive floating around and spent most of the day making a backup of it. I used Knoppix for that, and noted yet again that there seems to be some problem using it for NFS mounting file systems. Maybe I should have used the Ubuntu live file system instead.
Installing Ubuntu is interesting. It's straightforward enough—a little too straightforward, in fact. Tried for a while to partition my disk for three operating systems (NetBSD, FreeBSD and Linux), then gave up and installed FreeBSD first. After that installed Ubuntu on the partition that I had defined while installing FreeBSD, and noted with amazement that the “auto-allocate” allocated almost exactly the file system and swap what I would have done myself (single root file system and swap. I would have chosen 1 GB of swap, and Ubuntu allocated 790 MB). Installation proceeded without much intervention, with the result that on completion it had overwritten my FreeBSD boot selector with a Linux bootstrap which would boot only Linux. Login put me into GNOME, which I still can't make friends with: it takes forever to find how to start an xterm window, and why do they have to use stupid terminology like “folder” when they mean directory?. It looks clean, though; hopefully I'll be able to set things up properly.
| Tuesday, 26 April 2005 | Echunga | |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
Into town early this morning to talk with Ross about our current redesign proposal, and made better progress than at the corresponding meeting last week. It dragged on, though.
We both went to this month's ADUUG lunch, conveniently just round the corner. It proved to be quite animated; it's strange how each lunch has its own character. Then back with Ross to continue discussions. Despite a lot of progress, it still takes a long time, and we're not (yet) agreed.
Borrowed some books on Mac OS X from Ross. I had been rather looking forward to reading “Mac OS X: The missing manual”, and doubtless it contains a lot of information that justifies its name. But it still seems to have the wrong attitude:
Underneath the gorgeous, shimmering, translucent desktop of Mac OS X is Unix, the industrial-strength, rock-solid OS...“Who cares about the eye candy? And why is the “Unix” hidden?
Another quote is closer to the truth. I looked up directory in the index and found exactly one mention:
Note: Before Apple came up with the user-friendly term folder to represent an electronic holding tank for files, folders were called directories....
using a term like “working folder” within earshot of Unix geeks is likely to get you lynched.
I suppose the latter is closer to being correct. But what's user-friendly about obfuscation? Abusing an existing term to confuse the meaning of another perfectly valid term makes no sense at all. “Holding tank” indeed! I wonder how this analogy explains how the same file can be referred to in multiple directories.
| Wednesday, 27 April 2005 | Echunga | |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
I'm beginning to get the feeling that I can judge my (over)work load based on the number of mail messages in my inbox. If that's the case, I started out badly today, with over 3,300 messages, and spent a goodly time reducing that. Since I'm now also (in the background of my performance testing) playing with gnuplot, decided to set up a cron job to monitor my mail queue and put it on the web. It might be useful for people who wonder why I'm not replying to their mail message (and give them an opportunity to fill in the valleys).
Apart from that, spent some time looking at the impending disaster with my network connectivity: in violation of the user agreement, BorderNET has decided to terminate transmissions on PamAM Sat 8 with only 11 days' notice, and of course Telstra has postponed the DSLAM in the Echunga exchange again without any explanation, so at the end of the month I'll be back on dialup PPP. If that wasn't enough, they seem to have changed the rules for delegating networks, and now I need to get authorization from the administrative contact. That's not a problem in itself: I am the administrative contact. Unfortunately, it's in another life:
=== grog@wantadilla (/dev/ttypc) ~/bin 495 -> whois 192.109.197.0
OrgName: RIPE Network Coordination Centre
OrgID: RIPE
Address: P.O. Box 10096
City: Amsterdam
StateProv:
PostalCode: 1001EB
Country: NL
ReferralServer: whois://whois.ripe.net:43
NetRange: 192.109.0.0 - 192.109.255.255
CIDR: 192.109.0.0/16
NetName: RIPE-ERX-192-109-0-0
NetHandle: NET-192-109-0-0-1
Parent: NET-192-0-0-0-0
NetType: Early Registrations, Transferred to RIPE NCC
Comment: These addresses have been further assigned to users in
Comment: the RIPE NCC region. Contact information can be found in
Comment: the RIPE database at http://www.ripe.net/whois
RegDate: 2005-02-28
Updated: 2005-02-28
# ARIN WHOIS database, last updated 2005-04-27 19:10
‘
inetnum: 192.109.197.0 - 192.109.197.255
netname: LEMIS-LAN
descr: LEMIS Lehey Microcomputer Systems
descr: D-W-6324 Feldatal
descr: Germany
country: DE
admin-c: GL3-RIPE
tech-c: GL3-RIPE
rev-srv: allegro.lemis.de
rev-srv: ns.cls.net
rev-srv: ns.maz.net
mnt-by: AS2871-MNT
changed: ar@deins.Informatik.Uni-Dortmund.DE 19920521
changed: heinen@maz.net 19950723
changed: lillge@maz.net 19951010
changed: ripe-dbm@ripe.net 19990706
changed: ripe-dbm@ripe.net 20000225
changed: ripe@denic.de 20030625
status: ASSIGNED PI
source: RIPE
person: Greg Lehey
address: LEMIS Lehey Microcomputersysteme
address: Schellnhausen 2
address: D-36325 Feldatal
address: Germany
phone: +49 6637 1488
fax-no: +49 6637 1489
e-mail: grog@lemis.de
nic-hdl: GL3-RIPE
mnt-by: DENIC-P
changed: ar@deins.Informatik.Uni-Dortmund.DE 19920521
changed: knocke@nic.de 19941121
source: RIPE
This is the first time I even knew I had that handle. As the change dates show, it's out of date. I got rid of those phone numbers rather more than 11 years ago, and the email address died about 7 years ago. So how do I change the delegation? Things look rough.
| Thursday, 28 April 2005 | Echunga | |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
Still fighting a losing battle with mail. I have a horrible feeling that it'll be a month before I get back on top: I only have a week until I leave for Canada and BSDCan, and I won't be back until 21 May. I can't see myself having much time before then.
On a work front, we're still thinking about restructuring our application. The ideas about how to do it diverge significantly, and I'm reminded of how much better the “Open Source” model works here: we could discuss the matter in a much larger forum. As it is, there are just not enough people involved to ensure a “correct” decision.
Played around with eucla, the laptop on which I had installed Ubuntu on Monday. Yes, the installation was nice, but the rest was not exactly plain sailing. Couldn't find anything in the system administration menus to help me set up multiple boot—one of the basic problems with GUI applications: if it's not in the menu, you can't do it. I know that it's something to do with /boot/grub/grub.conf, but that file is missing, and it's not clear whether grub hasn't been modified. If I had the time, I could experiment, but I don't have the time. I also can't use the FreeBSD boot selector, because I've installed all of Linux in a single BIOS extended partition, and the FreeBSD boot selector can't handle that.
In the end, installed the FreeBSD selector anyway, which means that I can't currently boot Linux. That will change some time in the future, but for the moment I need to get the machine ready for the conference.
| Friday, 29 April 2005 | Echunga | |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
Meeting day today, with a twist: we're still agonizing over our redesign, and I got asked to prepare a document to address the current issues. So missed out on the meeting.
I'm back on dialup again. At 15:30 this afternoon my satellite downlink was re-routed to the PPP connection. Surprisingly, it went relatively smoothly, though it seems to be taking a long time for the routes to propagate, particularly, it seems, the breakin attempts on Microsoft ports. I wonder if they bypass the normal routing mechanisms.
Received what I had expected to be a Hauppauge PVR-250 card from ELX in the mail today. It's an OEM package, and it doesn't actually say what it is, but the accompanying documentation suggested that it was a PVR-150, and I later got confirmation that it is indeed a PVR-150: the supplier had switched without telling. I'm not sure that it'll make any difference, but the price is right. Now I have even more work to do over the weekend.
Later into town to buy a new barbecue. The one we had was 8 years old and was gradually rusting away. The crunch came a few months back when I tried to get a new flame tamer for it, and they wanted $50. The whole unit had only cost $399, and since the condition of the rest was not too good, I decided it would be better to wait for a new one to come on special offer.
Today was the day: not many people buy barbecues in autumn, and so they're reduced. Got a much better replacement in stainless steel rather than wood frame, and with a lid (something that most Australian barbecues don't have) for the same price we paid for the old one 8 years ago.
The lid is interesting. I know the Americans often cook with the lid down, but I've never seen anybody do it in Australia. But it has two non-obvious potential uses, about which I must think:
For making pizza. Original pizza ovens have a fire to one side and the flue to the other, and a marble slab on which to put the pizzas. Cooking is by convection over the top and conduction from the slab to the dough. By comparison, a typical Australian barbecue has a grill on one side and a metal slab on the other, obviously heatable. All I need to do is to arrange for some kind of convection from the grill over the slab and out the side, and I'll have a setup very similar to a pizza oven. I wonder if it's worth the effort.
| Saturday, 30 April 2005 | Echunga | |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
Lots of work to do this weekend: install my new Hauppauge PVR-150, finish installing the latest version of FreeBSD on eucla, one of the laptops I'll be using for my tutorial in Canada, and finish updating the handout for the tutorial. I didn't get anything done.
It turns out that the PVR-150 has another surprise: no remote control. That doesn't worry me, but it's worth noting. It probes like this under FreeBSD:
pci0: <multimedia, video> at device 7.0 (no driver attached)
=== root@teevee (/dev/ttyp0) /home/grog 3 -> pciconf -l -v
none2@pci0:7:0: class=0x040000 card=0x88010070 chip=0x00164444 rev=0x01 hdr=0x00
vendor = 'Conexant Inc (Was: Globespan, ICompression Inc)'
device = 'iTVC16/CX23416 MPEG Codec'
class = multimedia
subclass = video
Coincidentally, John Wehle posted the latest version of his PVR-250/PVR-350 drivers. Extracted them as described, but then ran into problems with the firmware. The instructions said:
3) Place a current version of hcwpvrp2.sys from the
Hauppauge drivers into /sys/dev/cxm. Extract the
the firmware by:
# cd /sys/dev/cxm
# cc -o cxm_extract_fw cxm_extract_fw.c
# ./cxm_extract_fw hcwpvrp2.sys
Unfortunately, this file didn't exist. The .sys files on the CD-ROM were:
-r-xr-xr-x 1 root wheel 26972 Nov 18 12:23 WinTV-NOVA-TMCE_2_102_22323_WHQL_signed/hcw88bar.sys -r-xr-xr-x 1 root wheel 142400 Nov 18 12:36 WinTV-NOVA-TMCE_2_102_22323_WHQL_signed/hcw88bda.sys -r-xr-xr-x 1 root wheel 307267 Nov 18 12:21 WinTV-NOVA-TMCE_2_102_22323_WHQL_signed/hcw88enc.sys -r-xr-xr-x 1 root wheel 8387 Nov 18 12:22 WinTV-NOVA-TMCE_2_102_22323_WHQL_signed/hcw88r9x.sys -r-xr-xr-x 1 root wheel 11719 Nov 22 12:20 WinTV-NOVA-TMCE_2_102_22323_WHQL_signed/hcw88rc5.sys -r-xr-xr-x 1 root wheel 14016 Nov 18 12:22 WinTV-NOVA-TMCE_2_102_22323_WHQL_signed/hcw88ts.sys -r-xr-xr-x 1 root wheel 128577 Nov 18 12:33 WinTV-NOVA-TMCE_2_102_22323_WHQL_signed/hcw88tun.sys -r-xr-xr-x 1 root wheel 579004 Nov 18 12:23 WinTV-NOVA-TMCE_2_102_22323_WHQL_signed/hcw88vid.sys -r-xr-xr-x 1 root wheel 131840 Nov 11 12:43 WinTV-PVR-150500MCE_2.0.18.22316_WHQL_signed/hcwPP2.sys -r-xr-xr-x 1 root wheel 814464 Sep 22 2004 WinTV-PVR-250_retail1_1_18_22266/hcwPVRP2.sys -r-xr-xr-x 1 root wheel 1433280 Oct 21 2004 WinTVPVRUSB2_22_73_22295_WHQL_signed/HCWUSB2.sys
cxm_extract_fw runs fine against WinTV-PVR-250_retail1_1_18_22266/hcwPVRP2.sys, but with WinTV-PVR-150500MCE_2.0.18.22316_WHQL_signed/hcwPP2.sys it says:
=== root@teevee (/dev/ttyp0) /usr/src/sys/dev/cxm 22 -> ./cxm_extract_fw hcwPP2.sys
cxm_extract_fw: decoder image not present
cxm_extract_fw: encoder image not present
I don't know anything about the content of these files, but it looks like I have a problem for the while. Sent a message to the FreeBSD-multimedia mailing list, but didn't get any useful replies.
On a more positive note, revisited the tuner that I had been trying to get to work last month, and after a lot of source code comparison, discovered some ideas about how to proceed to get it to work. What a pain this multimedia stuff is!
Completing the FreeBSD installation on eucla was easier, but I still didn't get it done: I've been working on my system upgrade procedures for a year now, but always when I've been in a hurry, and my scripts are so confusingly named that I sometimes wonder which does what. Started again today and wrote one called installrepotree, long but hopefully descriptive; it installs symlinks to the directories of an existing repo where they're needed in the newly installed system. That worked, but it's still not easy to decide how to perform the checkout, so I left it.
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 |