|
| Greg's diary |
| May 2013 |
|
|
||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
|
|
|
|||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| Wednesday, 1 May 2013 | Dereel → Ballarat → Dereel | Images for 1 May 2013 |
| Top of page | ||
| next day | ||
| last day |
|
Car repairs, then and now
|
Topic: general, opinion | Link here |
I'm currently reading through my (paper) diary for 1969. It makes interesting reading, but unlike the 1967 diary, I don't think I'll put it all online.
But I did.
In late 1968 I got a brand new Citroën Ami 6 car, and spent most of my time doing things with it. The amount of trouble I had with it makes sad reading. Some of it was my own fault, but it also shows up how terribly bad car service was in the UK at the time.
Today I got a call from Yvonne, who had gone shopping. She had stopped at the General Store in Napoleons, and couldn't start again. Her car, a 15 year old Holden VT Commodore, had had a flaky starter motor for some time, but I hadn't expected it to give out completely. She managed to get it started again, but clearly it was time to replace the motor. Met her at Ballarat Automative and arranged to change the motor; it seems not to be a big job, given a quote of 15 minutes' work, so—at 10:35—we asked if they could change it before we returned home at 13:00. They said they'd try, but first they had to get the motor.
From that point of view, things didn't work well. We got through our shopping much faster, and we were back at the workshop at 12:10—and the car was finished! That was only 95 minutes. What a difference from 1969! And the cars, too, are now so much better. My Ami 6 was only 7 months old (though it had done 35,000 odd kilometres), and I had several problems with it. Yvonne's car is 15 years old, and mine (Mitsubishi Magna TR) is 22 years old, each have done about 270,000 km on the first engine, and we have very few problems with either.
|
ALDI Massage chair
|
Topic: general, opinion | Link here |
While in town, returned the ALDI wireless routers and picked up another special, a “Shiatsu” massage armchair. For once, it came with instructions that were detailed and intelligible—but wrong. In particular, the particularly fiddly attachment of the base to the chair blocked one of the screwholes for the armrest. If it had been done in reverse sequence, there would have been no problem. As it was, just putting bases on the chair and footrest and the armrests on the chair took over half an hour. By that time, I was ready for a back massage.
And how does it work? I don't go in for massages, and I found it uncomfortable, but Yvonne likes it. One thing's clear: the backrest is too short. That's a pity, because apart from that it's also a fully-fledged armchair.
| Thursday, 2 May 2013 | Dereel | |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
|
More laziness
|
Topic: general, photography | Link here |
There's plenty to do, but I can't be bothered. Today I did even less than usual, including not looking at the computational photography course. Yes, I could finish it by Monday, but I'm not in it for the marks or the certificate: I'm in it for the understanding, and to get that I need to look further than the course itself.
|
NFS locking and ports builds
|
Topic: technology | Link here |
Got round to looking at my NFS locking issues today. Simple: by default FreeBSD doesn't start the NFS processes at all, so you have to configure it in /etc/rc.conf. I had that already in my real computers, but not in the ports build box. Problem solved?
Hard to say. The next problem was already there: no xterm. X had built, but for some reason xterm, a dependency, hadn't. Left another ports-try run, which it continued to do for the rest of the day and into the night.
|
Tagine aux petits pois
|
Topic: food and drink, opinion | Link here |
Decades ago, in another life, Yvonne learnt cooking from her then mother-in-law in the South-West of France. She kept notes in a book that she still has. One that she liked was this recipe for tagine aux petits pois:
The shorthand is French, but the dish is North African, of course, and not French. We've come some way with our cooking since those days, and so when we wanted to cook it today, I went looking on the web. The first thing of interest was that it gave me results for tajine aux petits pois, though there are many more hits for my original search string. Found a number of similar recipes, but they all had more vegetables than just peas. They also all used meat cubes, while Yvonne's recipe used mincemeat balls. But we already had the mincemeat, so we made a compromise. And surprise, it tasted very good! In particular, using mincemeat meant that we could add spices to the mix. I have no idea how authentic the recipe is, but possibly my North African friends will tell me.
| Friday, 3 May 2013 | Dereel | Images for 3 May 2013 |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
|
Ports pain, next instalment
|
Topic: technology, opinion | Link here |
On with my month-long attempt to build a clean reference installation of the FreeBSD ports that I use. Today my error was:
So what installed qt version 3? Some out-of-date port? I started building all ports from scratch with a new ports tree just a few days ago. Still, presumably that port would work with newer versions of qt, so decided just to remove the old version. But it wasn't just one port:
Are there really that many ports that depend on old versions? Probably not. So I blew it away anyway and started again:
And on it went. Decided to build qt version 4 before any of them and see how that would work. That meant blowing away all ports and starting again. And then I had Yet Another unexpected problem:
This is ridiculous. I've built gmake (GNU make) dozens of times. Tried it on eureka, and of course it worked. So what was wrong with stable-amd64? Checked the PATH environment variable, which ended in /Photos/Tools. That proved to come from my personal PATH environment variable and wasn't in the root PATH on eureka, so I removed it, and it built. That's clearly a case for a really clean build environment. The build went all day, of course.
|
Preparing for winter
|
Topic: gardening | Link here |
I've been sorely neglecting the garden since the beginning of the year, but gradually it's getting cooler, and I need to get the Hibiscus and the Mandevilla into the greenhouse, so did that. And nothing much else. It's about time I got somebody to help in the garden.
|
Piccola and Zhivago
|
Topic: animals | Link here |
We've had Zhivago for over 6 months now, but the cats have taken their time to get used to him. Lilac still doesn't have much to do with him, but Piccola is gradually coming around:
|
|||||||||||||||||||||||
|
Computational photography continued
|
Topic: photography, technology, opinion | Link here |
Continued with the computational photography course today. Somehow it's tailing off. The first couple of lectures and assignments were interesting, but now it's becoming too superficial. Saw a lecture about panoramas, which had a theoretical discussion of image alignment. Normally this is done by placing control points between individual images, but he didn't mention that at all, instead talking about homography in general terms (showing matrix operations but not really explaining what the individual parameters were), and glossing over things like projections. Not surprisingly, there are no assignments relating to this topic, nor to the next topic, High Dynamic Range Imaging. That's a pity, and it makes the course less interesting for me.
|
Scammers get cleverer
|
Topic: technology, opinion | Link here |
Received a strange email with quadruple spaced lines today. Here's the relevant content:
Clearly this isn't spam. But what is it? There are so many reasons to doubt its veracity. It comes from a gmail account with no obvious relationship to Exchange and Mart, he doesn't explain the relationship, and there's no obvious reason why Exchange and Mart should want this link removed (all publicity is good publicity). In general, though, it highlights a problem of trust which is being exacerbated by sloppy identification techniques even from ostensibly trustworthy sources. Here another example, apparently from ANZ Bank, with whom I have a number of accounts:
That was sent to an email address (represented here as groggysbank@lemis.com) that I have only divulged to ANZ. On the face of it, the message content is kosher, but it comes from a domain d121email.com, and contains links to that domain. It's not until you look at a genuine message that you can see that the one above is a scam. The genuine messages do—almost—come from an anz.com domain:
That's very different, of course. But does it come from anz.com? No, it comes from client-1-11.delivery.net. How can I be sure that it's really from ANZ? The DKIM signature is valid, which is good. But what if there's no DKIM signature, like on the first message?
I was uncertain enough of the true nature of the message that I called up the ANZ hotline, where I was connected to Nick, who told me first that it was a scam because the mail didn't come from an ANZ domain. But then, neither do the genuine ones. He wasn't concerned that it was sent to the secret email address—it seems that scammers can guess them. And then I told him of the content, and he decided that it wasn't a scam. Clearly there's not enough concern about security in the bank.
Sent the message off to the people in ANZ who should be looking at these things. We'll see what they have to say. In the meantime, though, there are enough indications that it is a well-prepared scam: the link to “Open an account” is to http://et.d121.com/EmailTracking.aspx?uid=22992&cid=21578&link=http://www.superannuation.anz.com/%3Fcid=em:391306a, clearly an encapsulation, and the whois information shows that the domain is assigned to a couple of people:
Surprisingly, there is a Digital Logic at that address, but of course the phone numbers are wrong (in fact, they're the numbers of the registrar). I wonder how much they have to do with this matter. This page suggests that they do. So does http://www.d121.com/.
The real issue, though, is trust. You should be able to expect that communications from your bank have only the bank's domain in the headers. And they don't. That makes it easier for scammers.
| Saturday, 4 May 2013 | Dereel | |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
|
Ports: progress
|
Topic: technology, opinion | Link here |
Next port build error today, in X:
That makes no sense at all. More environment variables? I've been working on my .bashrc literally for decades, as the comment at the top states:
It contained a number of environment variables that have passed their use-by date.
Most are harmless, but the first is not. Last year we established that it caused problems, and I removed it. But this time I forgot and it snuck back in. Set to and removed all non-essential environmental variables and tried again. Success!
It spent the rest of the day building qt. But if it will just build normally, I'll be happy.
|
Saturday dinner
|
Topic: food and drink, opinion | Link here |
Chris Bahlo along for dinner tonight—roughly beef kebab with onions and capsicum. On the recommendation of the butcher I used rump steak, not the normal beef cubes, and I grilled them in the oven, directly under the grill element; they took about 15 minutes. And yes, the steak instead of normal beef is a distinct advantage.
|
Another power failure
|
Topic: general | Link here |
Another power failure this evening at 20:03, fortunately only short. But it confused the air conditioners for some reason, and took a bit of power cycling to fix.
| Sunday, 5 May 2013 | Dereel | Images for 5 May 2013 |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
|
House photos: experiments and sloppiness
|
Topic: photography, opinion | Link here |
Normally I take my house photos every Saturday, but it was quite windy yesterday, so I did them today instead. Beautiful autumn weather, with lots of colour:
|
|||||||||||||||||||||||
Decided to try a few more experiments: in particular, the flash animations are quite good, but the fact that you can zoom in on them means that they ultimately get too unsharp. Normally I take them with my Zuiko Digital ED 9-18mm F4.0-5.6 ultra wide angle lens (equivalent to 18-36 mm on a full frame camera) set at 9 mm. To get more detail, clearly I need a longer focal length. Decided to try a series with my Zuiko Digital ED 12-60mm F2.8-4.0 SWD set at 25 mm.
There are a number of issues with this: first, of course, I need many more images. With the 9 mm setting I take 2 rows of 8 images each, and 1 row of 4 at the top. At 25 mm, I needed at least 4 rows of 15 images and one of 8.
But it wasn't done there: what do I do for exposure and focus? At 9 mm I don't have to worry too much about focus:
=== grog@eureka (/dev/pts/3) ~/Photos/20130505 13 -> dof 9 8
At 25 mm it's quite a difference, even at f/22:
The canonical way to address this is with focus stacking, something I haven't tried yet. Like High Dynamic Range Imaging, it requires multiple images, with corresponding problems with ghosting. In the end decided to use autofocus: the field of view of each image is relatively small, so presumably I won't have many individual images which require more than the depth of field that I could get with f/22.
And exposure? I'm continually refining my methods for that. Again, the canonical way to do this is with High Dynamic Range Imaging: take all images at a the same exposure, bracketed 2 EV in either direction. That's what I did for the “normal” panorama, though I've modified it to use automatic exposure, which gains me several EV in dynamic range. But that means 5 images for every view, at least 300 images for the whole panorama. And in general I've found that there are ways to simplify things: take single images with automatic exposure, ensure that no two adjacent images are more than 0.7 EV apart, and let enblend work it out. So that's what I did, and somehow ended up with 102 images, though I later discovered I had missed some out.
How did it work? I don't know yet. That wasn't the only thing that went wrong. I missed an image on no fewer than two other panoramas, and had to re-shoot them, and I also took another few. In the end I had no fewer than 604 images to process, which kept my DxO box running all day long. And when I got round to processing this one, it took Hugin 10 minutes of CPU time even before it started cpfind, and cpfind ran forever before I stopped it to do something else instead. So I'll find out tomorrow.
|
Linkedin: mutual admiration society?
|
Topic: technology, opinion | Link here |
I've commented in the past about strange messages like this one:
The difference in this one is that I know it's genuine: it's from Peter Jeremy, and it makes sense. But this time I went to my Linkedin profile to see what else had accumulated there. It's amazing. I've been endorsed for things I know nothing about, like Solaris, Cloud computing or perl, by people with whom I have had no contact for decades, in at least one case for over 30 years. And I have 26 endorsements for Linux and only 15 for FreeBSD.
Why do people do that? Not primarily to do me a favour, I think. That would make more sense if I had regular contact with them. So that I can endorse them back? Maybe. But in so many cases I have no idea what they're doing now, and somehow it's just not me. Once again I'm left wondering what the point of Linkedin is.
|
Where's Yvonne?
|
Topic: animals, general | Link here |
Out in the hallway today I saw this picture:
|
|||||||||||||||||||||||
They were all waiting for Yvonne, in the bedroom behind the door to the right.
| Monday, 6 May 2013 | Dereel → Ballarat → Dereel | Images for 6 May 2013 |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
|
A day processing a single photo
|
Topic: photography, technology | Link here |
On with the big panorama today. It took me all day, and by the end I still wasn't finished. Here the times it took to align the images:
| Process | Time (minutes) | |||
| hugin | 40:56 | |||
| cpfind | 192:57 | |||
| icpfind | 1:20 | |||
| cpclean | 39:20 | |||
| autooptimiser | 33:30 | First time round | ||
| autooptimiser | 11:08 | After cleaning control points | ||
319 minutes of CPU time just to detect the structure of the panorama! After that, started stitching and got the message:
The panorama you are trying to stitch is 0.6 gigapixels.
Are you sure you want to stitch such a large panorama?
If this is too big, reduce the panorama Canvas Size and the cropped region and stitch from the Stitcher tab. Stitching a panorama this size could take a long time and a large amount of memory.
The message was justified. Running nona, the program that reshapes the individual images, took 1 hour, 43 minutes. And enblend? After a couple of hours, I returned to find:
That's a process address space of 18 GB, 4.7 GB memory resident, and the machine was swapping its heart out. It had only read in about 40 of the 88 photos that it needed, and I needed the machine for other things, so I stopped it and restarted it after I had finished. In sum: all day processing the one panorama, and only half finished.
|
Chris Bahlo: Best in show
|
Topic: general, opinion | Link here |
Off to another graduation today, this time Chris Bahlo's. Now in addition to her agricultural and accounting qualifications, she's now a Bachelor of Professional IT with Distinction.
|
|||||||||||||||||||||||
This time they weren't quite as hurried as last month, but there were no windows at all in the theatre, so I turned up my camera sensitivity to 36° (3200) ISO and managed to get some sharper, if somewhat grainy images:
|
|||||||||||||||||||||||
But wait! There's more! After all the graduands graduated, Chris had to return: she received the Ballarat University Medal for outstanding performance. That wasn't specific to this particular graduation ceremony, for the Engineering and IT departments: it's only given once each year. In many ways, as animal breeders know, it's not “Best in Show” (this school) but “Best of Best” (all schools) Much eulogy about her perfect academic record, her dedication over and above the cause of duty, and such things:
|
|||||||||||||||||||||||
|
|||||||||||||||||||||||
Then Chris went back to return her (rented) academic apparel. That took a while, so I went looking, but didn't find her. She found me. She was right in front of me, but I didn't recognize her:
|
|||||||||||||||||||||||
I don't know if I have ever seen her wearing a dress before.
In passing it's interesting to note how many Asian graduands there were. In the undistinguished and unprofessional Bachelor of IT, it seemed that about 80% of the graduands were Indian or similar. I wonder why that is.
Another detail: last month I complained that the music played during the Academic Procession did not include Gaudeamus igitur. Today it did, admittedly only a small part. And of course we had the Australian National Anthem after the procession. I've never found the words particularly moving, but then it occurred to me that the words to “Gaudeamus igitur” can be translated as “Let us rejoice, for we are young”.
At the time, the Australian anthem started with “Australians all let us rejoice, For we are young and free”. But it seems that we're no longer young (that certainly applies for me, anyway): since 2021, they're “Australians all let us rejoice, For we are one and free”. I wonder what changes will come in the future.
|
Zhivago: attention getter
|
Topic: animals | Link here |
Zhivago didn't want to be left behind when we went to the graduation ceremony, so we took him with us. He wasn't allowed in the ceremony, of course, but we kept him with us until we went in. Once again, he attracted a lot of attention, especially from Asian people, of whom there were many:
|
|||||||||||||||||||||||
|
|||||||||||||||||||||||
| Tuesday, 7 May 2013 | Dereel → Geelong → Dereel → Ballarat → Dereel | |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
|
Periodontics again
|
Topic: general | Link here |
Six months ago I made an appointment for my next regular periodontal checkup—yesterday at 15:30, exactly the time that Chris Bahlo's graduation started. As a result I had to postpone, and the only appointment I could get in the next two months was today at 10:00, which meant leaving for Geelong at 8:45. To make matters worse, there was nothing else I needed to do in Geelong—160 km just for that. On the positive side, everything's in order in my mouth.
|
Photo processing: your computer is too wimpy
|
Topic: photography, technology, opinion | Link here |
Before leaving for Geelong, I checked how my enblend run was going. Again 20 GB of process space, 5.5 GB of memory. It had been running for 14 hours, had used only 80 minutes of CPU time, and had processed about half the photos.
When I got back from Geelong 3½ hours later, it had only used about another 10 minutes of CPU time and loaded another 8 images. It's unclear how long it would have run—my guess is that it wouldn't have finished today—but it was effectively holding up the machine, so once again I stopped it.
|
A new machine?
|
Topic: technology, opinion | Link here |
My photo experiences show that I need more memory at any rate. But my motherboard is nearly 5 years old, and it doesn't take more than 8 GB of memory. Time for a new machine? The current CPU is an AMD Phenom 9550, which PassMark rates at 2493 points. The top of the line processors rate at 14,969, at prices I'm not prepared to pay. But I thought I could find something at about 70% of that rating for under $300. I was almost right; for $280 odd I can get an Intel Core i7-3770K with 9,461 points (63%), and for $180 I can get an AMD FX-8350 (9,144 points or 61%). That's a big difference in price for such a small difference in performance. PassMark conveniently shows bang per buck: 47 points/$ for the AMD, 33.8 for the Intel. Looks like a case for the AMD chip, again.
But there's more to the price than the purchase price. The AMD has a TDP of 125 W, the Intel of 77 W. How much does that difference cost? Currently a kWh costs about $0.28. 1 W is equivalent to 8.77 kWh or $2.45 per year. So a difference of 48 W would amount to $117.60 per year—more than the price difference between the two chips.
The problem here is that the processors aren't running at full power all the time. I need to research further, but it's clear that the Intel is at least be more environmentally friendly, and it could end up being cheaper.
|
Whose NBN?
|
Topic: technology, opinion | Link here |
The discussion about the Australian National Broadband Network is ramping up in preparation for the elections in September. The incumbent Labor government has introduced a very ambitious FTTP solution, originally only with 100 Mb/s maximum speed. But now the opposition parties (“Liberal” (in my mind really conservative) and National) look set to win the next election and replace it with an FTTN solution that is barely acceptable now and most certainly will not be in the future: a guarantee of only 25 Mb/s with the hope of 50 Mb/s some time in the future.
It doesn't help that the NBN project is significantly behind schedule: “This government” clearly doesn't understand networking. Neither, it appears, does the “Coalition”. But the NBN has upped the ante by announcing top speeds of 1 Gb/s down and 400 Mb/s up. Clearly at present that's nonsense, since the backhaul couldn't handle it. But it has certainly shown the limitations, as this site shows.
“This site” washttp://howfastisthenbn.com.au/, but it is long gone.
|
Back to building ports
|
Topic: technology | Link here |
One of the things that I had to put on hold while doing my photo processing was the ports build that has been going on for over a month. Today I was able to continue; and of course several ports had changed, so once again I had the continual hangs waiting for configuration information. I still don't have qt built.
|
Dinner for Chris
|
Topic: general, food and drink, opinion | Link here |
We decided to invite Chris Bahlo to dinner to celebrate her graduation and her medal yesterday. That's not really our thing, since there aren't many restaurants that we consider good enough, but finally decided on Kambei in Main Road. Chose the standard table d'hôte, which they call “Course menu “Kambei””. It was a good choice: there were things on it that I wouldn't have chosen from the menu, such as the edamame (boiled soya beans) as appetizer, or the shiitake and spinach salad. Very enjoyable.
|
Flaky technology
|
Topic: general | Link here |
On the way home from the restaurant, had multiple issues with flaky technology. The GPS component of my navigator stopped running a couple of times, and the alarm system of the car turned itself on and off several times, accompanied by noises and flashing lights. When I got home I disabled the noise, but hopefully the thing isn't going to start annoying me. By contrast a new GPS navigator isn't much of an issue.
| Wednesday, 8 May 2013 | Dereel | |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
|
Ports pain, next installment
|
Topic: technology | Link here |
After fixing my build environment, I thought that my ports would build cleanly. So I wasn't really expecting this when I came into the office:
Checking the environment showed that I still had all the old, mouldy environment variables: su had inherited them from the parent shell. Moral: run su -, not su. That did the trick, at least this time. Later it stopped again in ImageMagick, which proved to be my fault. But the changed ports were a nuisance, so I ran my ports-fetch target, which took quite some time. Then I tried ports-config, which runs the config target in each Makefile. Problem: it does it even for ports which have already been configured, which made it pretty much useless, so I had to go back to the old style of responding to each individual config prompt.
|
Photoshop: Triumph of marketing over technology
|
Topic: technology, photography, opinion | Link here |
So it's official. Adobe will not develop its “Creative Suite” software any more. Instead they have created a “Creative Cloud”. To quote the Creative Suite page:
While Adobe Creative Suite® 6 products will continue to be available for purchase, Adobe has no plans for future releases of Creative Suite or other CS products.
With a little trouble I found the pricing page for “Creative Cloud”. Apart from free trial membership, the very minimum you can pay is $20 US per month. And you have to commit to at least 12 months, so you can't save things up for a few months, buy a month's worth of processing, and then stop again. $240 per year doesn't even buy you much; for the equivalent of CS6 it seems that you have to pay $50 a month.
What do you get for that? From the Adobe page:
- Full versions of our most popular desktop applications — Adobe® Photoshop®, Illustrator®, Acrobat®, and more
- Full access to services to help create mobile-ready content and apps
- 20GB of cloud storage for file sharing and collaboration
At first sight, external processing sounds like a good idea. I've been thinking of alternatives for stitching my behemoth panorama. I could probably get access to an adequately dimensioned system without much difficulty, for example an Intel Xeon X5650 with 24 GB memory.
But just uploading the images would be a massive undertaking: 7 GB of TIFF images up, 1 GB back. That's nearly half my monthly traffic allowance, and if the network behaves the transfers alone would take well over a day. And it would require a system running the software I want to run, not Adobe's. By contrast, a new motherboard like the ones I was thinking about yesterday would set me back less than a year's subscription. So the “cloud” is really not much use to me—in fact, I don't understand who needs it. 20 GB of storage is nothing nowadays: 2 TB disks cost about $100, so it's worth about $1.
Mobile-ready content and apps? Is that serious photographic processing? Again, nothing I would even remotely consider using even if it were free. So what really counts is access to the software. I'm left wondering if the Adobe product managers have lost touch with reality.
Is that worth it? Not to me. I would never have bought Adobe in the first place. But for the customers who are used to it, what will they do? If they have CS6 at the moment, they can carry on using it forever, probably the reason why Adobe offers them a discounted subscription for $30 a month (for how long?). But they'd have to have a good reason to pay that rather than stick with what they have. The alternative might be for a number of people to buy the subscription communally and use it in turns. That probably contravenes Adobe's license conditions, but how could they find out? After all, one of the key features is that you can access the cloud from anywhere.
It'll be interesting to see how it peters out, but I wonder if, like Microsoft's “Windows” 8, it's not a symptom of a fading monopolist trying desperately to hold on to power.
| Thursday, 9 May 2013 | Dereel | Images for 9 May 2013 |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
|
Daily ports breakage
|
Topic: technology | Link here |
This morning's ports breakage:
What's that? I have just installed libebml, and it claims it wasn't installed? In fact, the version that got installed was the version in the Makefile, libebml.so.4. So why did it look for libebml.so.3? The connection isn't clear, but it seems that there was a work directory for an old version of libmatroska in the tree. That shouldn't have been there, since I do a make clean before every port, and it should be recursive. But libeml is an optional dependency, and it looks as if it didn't get removed. Next time I'll do an rm /usr/ports/*/*/work.
Things didn't stop there, of course. Once again the build of gcc failed, this time version 4.8, with the same kind of message:
What causes that? It happens after about 4000 lines of build output, and just tracking it down would be a pain. Tried building it on eureka, and it worked. Why? Is this a difference between a virtual machine and a real one? That shouldn't be the case. Is there something else wrong with the stable-amd64 installation? I've really tried to keep it as generic as possible. Still, hopefully this will be the last time.
|
Old photos resurface
|
Topic: general, photography | Link here |
Talking on IRC with Andy Snow today. He's moving in to a house near Berry in the south of New South Wales. He's currently considering household pressure pumps, and is planning to buy one on the web and install it himself, something that none of us who live in the country think is a good idea. To make my point, I dragged up a reference to my last burst pump and pointed it to him.
And in the process I found a large number of other photos that I had never processed, including evidence that I have, after all, had dinner with Monty Widenius long before I joined MySQL, though not quite what I thought:
|
|||||||||||||||||||||||
|
|
|||||||||
|
Horse clinic starts
|
Topic: animals, general | Link here |
Trisha Wren arrived today to hold a 3 day Connected Riding clinic, starting tomorrow, that will be held mainly at Chris Bahlo's place, though Trisha is staying here with us. A number of other people, both known and less known, arrived by evening, and we had a total of 9 for dinner. Chris has obviously been more successful academically than in her chosen pastime of beating up young men:
|
|||||||||||||||||||||||
The people I didn't already know were both called Paula. One, Paula Coombs, from Singapore, the other from Utrecht. Interestingly 5 of the 9 people spoke German (the others were Nele Kömle and Birgit Kossmann, from somewhere near Münster).
|
|||||||||||||||||||||||
|
|||||||||||||||||||||||
|
|||||||||||||||||||||||
|
|||||||||||||||||||||||
It was also interesting that I have been to just about everywhere where the people come from—Trisha lives in Cambridge, New Zealand, not far from Hamilton, New Zealand, where Jörg Micheel used to live. And Paula Coombs isn't really from Singapore: she lives in Foxground, near Berry, only about 10 km direct line from where Andy Snow is moving. She had a number of comments that should be useful to him.
| Friday, 10 May 2013 | Dereel | Images for 10 May 2013 |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
|
Solar eclipse photography
|
Topic: photography, opinion | Link here |
Another solar eclipse today. Further north it was total, but here it was so minor that you couldn't tell with the naked eye. But it was at a convenient time, and the sun was shining, so I mounted my 300 mm Hanimex lens on a tripod with 2× and 3× teleconverters, which projected a sufficiently large image to show the eclipse without danger.
Of course, after that I just needed to attach a camera and adjust exposure to get a photo. I've been there before, and on that occasion I decided that it wouldn't be worth the trouble. Still, all I needed was a camera, so I attached it and took some photos which, for the most part, confirmed my suspicions. The shortest exposure I could use was EV 27.1, and I needed something closer to EV 30. But postprocessing helped there. What I didn't expect was that the image was so big that it wouldn't fit on the sensor:
|
|||||||||||||||||||||||
|
|||||||||||||||||||||||
Tried stitching it together with Hugin, but there are so few features that Hugin couldn't cope at all. I couldn't even find appropriate control points manually, so I tried just aligning the images on top of each other, not very successfully:
|
|||||||||||||||||||||||
Another serious issue with Hugin is that the fast preview image was invisible: somehow the minute angle of view (0.7°) confused it. When I manually reset the focal length to 10 mm, all was well.
|
French cookery terms
|
Topic: language, food and drink, opinion | Link here |
A couple of days ago I used the term table d'hôte in this diary. For me, this is the opposite of à la carte: a fixed menu instead of a choice of individual dishes. And indeed, that's what it means—in English. But it doesn't have that meaning in French. To start with, the French word hôte doesn't mean what it looks like: it means “guest”, not “host”. So “table d'hôtes” (the usual French spelling) means “guest's table”, and originally meant a fixed meal served at a separate table for people staying the night.
Problem: how do you translate the English use of “table d'hôte” into French? I still don't know. I'm not even sure there is a clear term. The French Wikipedia page doesn't know either: the closest appears to be “menu à prix fixe” (“fixed price menu”).
|
Daily ports breakage
|
Topic: technology | Link here |
After my fixes yesterday, a surprising number of ports compiled without error. The next one to die was chromium (or is that chrome? I still don't know):
What does that mean? Is it because I compiled it against gcc 4.8? It seems that the port is normally compiled against gcc 4.6, so potentially it has bugs in the code which are caught by 4.8, but not by 4.6. In any case, I don't want to know. Used pkg_add -r to load a binary from freebsd.org. That took all day, due in equal part to the flakiness of my network connection (which, fortunately, didn't drop) and the size of the packages, including over 100 MB of gcc 4.6—what does it need that for?
Continued in the evening:
That looks suspiciously like old, mouldy work directories, so as planned removed them all, reclaiming nearly 20 GB of disk:
=== root@eureka (/dev/pts/7) /usr/ports 29 -> df .
=== root@eureka (/dev/pts/7) /usr/ports 30 -> rm -rf */*/work
=== root@eureka (/dev/pts/7) /usr/ports 31 -> df .
So I tried again in the build directory:
That didn't look very promising. And the build failed again in the same place. Looking through the Makefile, it's not clear where the dependency on libz comes from. But I didn't have any time to look at it, and tomorrow is a day of rest for the VM while I process photos, so it'll have to wait until Sunday.
| Saturday, 11 May 2013 | Dereel | Images for 11 May 2013 |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
|
Busy day
|
Topic: general | Link here |
The clinic is in full swing now, and we've been feeding the people at our house. I was involved in the cooking, and between that, baking bread and taking my weekly house photos, I hardly had time to sit down.
|
Creative cloud: good or bad?
|
Topic: technology, photography, opinion | Link here |
Mail from Tom Maynard today:
I know several professional photographers who do not share your view of Adobe Creative Cloud—and, even I dispute some of the statements you made, since I investigated “the Cloud” as an alternative to outright purchase of Creative Suite 6.
OK, that's valid. But what are the reasons? I've been following this thread, in which people object mainly to the same things that I did a couple of days ago, and also—particularly the professionals—to the fact that they have to move their intellectual property offsite.
Tom pointed to this page, which, to his credit, he didn't find very good. Most certainly the counterobjections to apply to me—I'm certainly not a 13-year-old sissy crying about the matter: I don't use Photoshop anyway, and my observations were more seen from my guess at Adobe's perspective. And one claim seems completely wrong (his bold face):
(This is the REAL reason for 90% of the noise!) The haters are mad because they realize they can no longer pirate copies of Photoshop.
I expect that Adobe will have something in place to minimize abuse, but I don't see it being any more effective than what they have now. It will just take a different approach.
The discussion had a useful outcome. Tom conceded most of my points, but there's something in the “cloud” of immense benefit to Adobe and minor benefit to the users: having the software under Adobe's control means that they can continually upgrade it without the need for formal releases, and Adobe will save a lot on helpdesk staff trying to help clueless users install the software or maintain the existing installation. That's in the users' interests too. The question is to what extent updates will be more frequent.
|
X cursor hang: insight
|
Topic: technology | Link here |
Heavy CPU and memory use, such as I cause while processing my weekly photos, frequently triggers this horrible X bug that I've been suffering from for over a year: the X server loops, and the cursor jumps back and forth between two screens. I've taken to not using the mouse when the system is paging heavily, and today I didn't (quite) have the problem.
What I did have, though, was surprising: on the second server, which is a single display spread over 4 monitors, the mouse cursor moved to the wrong screen! Moving left from the second screen from the left should, of course, have taken me to the leftmost screen. Instead it took me to the one to the right. And when I tried to go left from there, it took me to the extreme right screen. I left it there until the load dropped a bit, and then it worked correctly again. I wonder what causes that, and what relation it has with the primary bug. Certainly if one screen thinks left is left, and the one to the left thinks left is right, we have a very similar situation.
| Sunday, 12 May 2013 | Dereel | |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
|
Ports progress
|
Topic: technology, opinion | Link here |
Now that I've processed my weekly photos, I can return to building ports. On Friday I had a strange dependency failure in openjdk. Tried again today in various ways, including an attempt to install version 7 instead of the version 6 that the depending port was asking for. No go: it still complained that libz was missing. In the end gave up and installed the binary package, which worked—and didn't install any libz! I wish I understood why I'm having so much trouble.
|
Updating web browsers, the hard way
|
Topic: technology, opinion | Link here |
The installation of firefox on teevee is ancient (release 6.0), and it doesn't have flash. Clearly I need an upgrade. But how? Under FreeBSD that's done with the Ports Collection. And I'm having enough trouble on a brand-new machine. Just trying it with an old, out-of-date machine seems a Bad Idea.
So I tried upgrading from the binaries. Also, it seems, a Bad Idea. In summary:
The first attempt failed because perl, pkgconf and xcb-util-renderutil were out of date. So I replaced them from binaries. After that I was able to install firefox.
firefox started and complained that sqlite3 was out of date. Installed the new version, and it in turn complained about missing libraries: it wanted older versions than were installed. So I linked the new ones to the old name.
After installation, sqlite3 died with missing functions. This doesn't seem to be related to the previous library issues.
Gave up on that and discovered that opera still ran, and even had flash. But I wanted firefox, so tried to add the old backed-up version. No go: it complained about missing libraries again. So I installed the package with tar instead of pkg_add, which worked as far as I can see. This is all such a mess; high time to get my reference machine finished.
|
X display resolution insights
|
Topic: technology, opinion | Link here |
Since I've had the new 58" TV, some menus, notably in web browsers, have been in minuscule fonts, and I haven't found a knob to tweak to fix it. But it occurred to me that the text is about the same absolute size that it would be on a normal desktop monitor. At 60 cm it would be legible if the display resolution were high enough. At 3 m distance it's illegible.
Further investigation in the nVidia X config options appendix showed that I could override the DPI value for the panel, which X had calculated—probably correctly—at 42 DPI. Increasing the value to 120 seemed to set it about correctly for the distance:
And yes, now the menus and various displays look correct. But clearly there's a misunderstanding here: X is assuming a specific size rather than an angle of view. I don't see that changing.
|
Understanding the Creative Cloud
|
Topic: technology, photography, opinion | Link here |
I thought I had commented enough about “Creative Cloud”, but then I got a message from Michael Hughes:
When you click on what's included, they have:
The world's best desktop applications for photography, video, audio, and design.
So I'm I missing something? It looks like they are still selling you desktop applications. Why are they using “the cloud”?
So I read the page, and I'm sure it didn't say what I saw before. But on the “main page”, I see:
And, as always, your applications live on your desktop, not in a browser and not in the cloud.
This all seems somewhat nebulous to me. I don't really understand. What is the cloud? Are they trying to change things as they go along? It's not clear. Michael's quote comes from this page. And it links to an (off-site) page 5 myths about Creative Cloud. In particular, the files (normally) stay on your machine, and you don't need to be connected to the Internet to use them (but you need to connect once a month for license check purposes).
So: a lot of I said about both pros and cons is Just Plain Wrong. You can't access your applications on the cloud from anywhere you like. You can only use the software on two computers (just like the previous versions). You don't get the benefit of fast remote computers to do your photocrunching for you. Software update is not obviously any easier. Adobe doesn't save on help desk people.
So what is the difference? If I understand it this time, it's just the means of distribution of the software. Instead of physical media, you install across the Internet. Instead of paying once, you pay all the time. “Creative Cloud”? The most creative part appears to be the name.
Or so I thought. Then I found their FAQ page, where the A definitely stands for “Asked”, not “Answered”: to read the answers you need to click on Every Single Question. And the “answers” don't necessarily answer the question:
Why is Adobe discontinuing boxed copies of Creative Suite?
As Adobe continues to focus on delivering world-class innovation through Creative Cloud and digital fulfillment, we will be phasing out shrink-wrapped, boxed versions of Creative Suite. Electronic downloads for Creative Suite products will continue to be available from both Adobe.com, as well as reseller and retail partners.
That's an answer? There are also other questions that seem to completely contradict the statements above. It's all on your computer, right? Where do this question and answer fit?
Where is Creative Cloud hosted?
Creative Cloud is hosted on Amazon Web Services™ (AWS) in the United States, Europe, and Asia. AWS offers a reliable platform for software services used by thousands of businesses worldwide...
Does this refer to image storage or application storage? AWS suggests the former. So the only way I can make sense of these various statements is that the applications are on your local computer, and your images are in the cloud. That doesn't make any sense at all. I'm more confused than ever, and I'm still left wondering what good the whole thing will be for Adobe. But the bottom line: I don't use this software, and I don't intend to, so I don't need to worry about it any more.
| Monday, 13 May 2013 | Dereel | Images for 13 May 2013 |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
|
Horse clinic winds down
|
Topic: general, animals | Link here |
The horse clinic really finished yesterday, but Trisha and a couple of others stayed on until today for some private lessons and plane connections. Finally the last one left—what a relief! Somehow didn't do very much.
| Tuesday, 14 May 2013 | Dereel | |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
|
Ports: your fault
|
Topic: technology, opinion | Link here |
On with the ports pain. Today I had:
So: basically a broken port. But the current version of KDE is version 4. Looking at the ports, I discovered that there are no fewer than 5 digikam ports, three of which start with digikam-kde4. What does /usr/ports/UPDATING have to say?
And that's all. Which is the preferred port? You'd expect the base name, and the UPDATING entry doesn't help. Neither do the pkg-descr files, which are identical. But it looks as if I should install digikam-kde4, so set to doing that. It took all day: once again my network connection is worse than a 19.2 kb/s modem.
Of course, not all ports are maintained by somebody else. I received a message from Ion-Mihai Tetcu, one of my ports mentors, telling me that deinstalling enblend left an empty directory behind. So I fixed it, tested it, and committed it. The result? A message from Alexey Dokuchaev telling me that the removal now tries to remove a non-existent directory. Same here. What about the test? My best bet is that I accidentally deleted another line in the Makefile after the test. But Alexey also pointed out that nothing ever got installed in that directory. Egg on my face. On the positive side, he did send patches, pointing to a number of other things in the process. What an embarrassment!
|
New watch
|
Topic: general, opinion | Link here |
I've been using digital watches for 40 years, since the first LCD watches became available. At the time, I was convinced that they were the way of the future. In my case, it replaced an Omega Speedmaster, which even today commands prices of well over $10,000 on eBay. Mine, unfortunately, was stolen about 12 years ago. But by comparison, digital watches are infinitely cheaper, they have more features, and they're more accurate. I never used my Speedmaster after getting a digital watch.
But the price of purely digital watches was clearly a serious problem for the manufacturers: how can you make a profit on things that are much more reliable and cost only a fraction of the cost of a mechanical watch? In the course of the time they reintroduced analogue watches with electronic time sources. Surprisingly—to me, at any rate—they caught on, and now most purely digital watches are either kiddies' toys or pretty much copies of what I had 40 years ago.
One thing has changed: the battery life. With Lithium batteries Casio now claims battery lives of 10 years. I bought my last watch over 9 years ago, and it's still running on the original battery. But the illumination is getting dim, and a new battery would cost at least $10 and a fair amount of trouble, whereas a new watch on eBay costs only $28. The watch is also looking a little tatty—the “metal” case has lost its metal coating in places, showing the plastic underneath, and a couple of the buttons have been damaged, so it seemed time for a new watch. It arrived today, a Casio DB360. It looks pretty much like what I had 15 years ago, complete with useless “databank” capable of storing text in the watch. I wonder if anybody actually uses that “feature”. And like its predecessor, it no longer has the functionality of letting the countdown timer continue from the beginning after finishing the countdown.
Then, while reading the instruction manual, I discovered that the watches use CR2025 batteries, the same kind that the remote control for my camera uses. Purely by chance I had ordered 5 of them on eBay this afternoon for $1, including postage, so one battery costs only $0.20. I suppose I'll change the battery in my old watch after all.
|
Succumbing to Photoshop
|
Topic: photography, opinion, history | Link here |
Nearly 50 years ago I walked around Kuala Lumpur taking photos of landmarks. They weren't particularly good photos—this was with a simple 35 mm camera with no creature comforts—but the photos have proven to be an interesting historical document, and I've had a number of requests for copies. This was just under a year since the formation of Malaysia (officially on 31 August 1963, in fact, due to Indonesian “confrontation”, on 16 September 1963), so it was just coming up to the first anniversary celebrations. Many new buildings had either been built recently, or, as in the case of the Masjid Negara, were in the process of being built:
|
|||||||||||||||||||||||
But the photos are terrible, at least in part because they were badly scanned. How about a bit of postprocessing? Discovered that DxO Optics “Pro” has a marginal set of film profiles. They have a profile “Generic Kodak Ektachrome™ 100 VS”. I suspect I took the originals with Ektachrome X, so tried that, and it greatly improved the colour rendering. Tried some other things too, but I still wasn't completely happy.
But isn't that the domain of Photoshop? And didn't I download a freely available copy from Adobe a few months back? So installed it on dxo and tried it out, following a tutorial that was delivered with the package. It's confusing to use, to say the least, but the tutorial showed me how to improve my images up to a point. I could probably have done something very similar with xv. Is it worth trying harder?
| Wednesday, 15 May 2013 | Dereel | |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
|
Ports build: resolved?
|
Topic: technology, opinion | Link here |
The good news about ports is that the port build cluster has now recovered from the security incident, though it doesn't seem to have made it to the newsflash page yet. Up-to-date packages are now available, so theoretically I can stop my attempts at building from source. One problem: I need to generate Makefile targets to install only those ports in the list that aren't already installed. So for today I carried on as before, using the packages only when I had trouble with the ports.
The first was in a dependency for enblend, though not my fault this time. The documentation didn't build:
I've seen something like that before, to do with the wrong version of texinfo. I couldn't be bothered to follow up: it's easier to just install the package, so I did that, and then enblend installed correctly. After that I had difficulty with panomatic:
It's not clear what the “distributed files” are, but it proved that the clock on stable-amd64 (build machine) was 3 minutes behind eureka, where the file system was mounted. That's being possibly a little over-pedantic.
| Thursday, 16 May 2013 | Dereel | Images for 16 May 2013 |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
|
Ports: done!
|
Topic: technology | Link here |
I've been working on a complete build of those FreeBSD ports that I use for over 5 weeks. I can't say I haven't made progress, but it was very slow, and all of this should have happened without any problems. Still, it was a little surprising when I read today:
I'm done! What do I do now? Start again, of course, this time with pkgng. But not today. First there was some wound-licking to do. Yesterday's build of gsl wasn't the only one to die trying to run makeinfo—this time it was zsh—so I went to see what was going on. Surprise, surprise! My environment variables again, helped by incompatibilities in texinfo. I have two versions of makeinfo installed, one from the base system, the other (newer, but released under the dreaded GPL version 3 , which, we're told, is far too dangerous to include in the FreeBSD source tree) from the ports. For this sort of reason I have /usr/local/bin before /usr/bin in my PATH environment variable. But it's not right for building ports; there you specify the new version of texinfo by adding this line to the port Makefile:
After that, nip2 failed with more sinister messages:
That was another case for pkg_add -r, not without a certain amount of complaining that suggests that the packages aren't that up to date after all:
=== root@stable-amd64 (/dev/pts/2) /home/Sysconfig/scripts 13 -> pkg_add -r nip2
And in the process I found that my root file system (which includes /usr) was too small, giving rise to amusing error messages like:
=== root@stable-amd64 (/dev/pts/0) /home/Sysconfig/scripts 80 -> pkg_add -r nip2
It seems that 20 GB is really the minimum practical file system size for a combined root and /usr file system, and I only had 10.
But that was all. Now I have an almost up to date system. Started building the latest kernel, and moved on to other things.
|
Flowers in garden, late autumn
|
Topic: gardening | Link here |
It's the middle of the month again, time to take photos of the garden.. The garden didn't do well this summer, and looking back, it's clear there were not one, but two reasons: not only was it hot and dry, but my irrigation made things worse. Some time back I decided to replace sprinklers with drippers, but despite following the instructions and calculations, it's (now) clear that there wasn't nearly enough water. The roses stopped flowering altogether, and only now, when it's moister but much colder, did some start to flower again:
|
|||||||||||||||||||||||
|
|||||||||||||||||||||||
|
|||||||||||||||||||||||
In at least one case, a summer-flowering plant has only now started flowering, the Gentiana scabra that I bought last year:
|
|||||||||||||||||||||||
|
|||||||||||||||||||||||
No wonder the transplanted lime tree died!
And then there are confused spring-flowering plants: Azalea, Acacia and Narcissus:
|
|||||||||||||||||||||||
|
|||||||||||||||||||||||
|
|||||||||||||||||||||||
Peter Jeremy tells me that he also has Narcissus flowering at the moment.
The petunias haven't done well either—I wonder if it's at least partially because some of them are three years old. Next year I should buy fresh plants. But we still have half a flower hanging on:
|
|||||||||||||||||||||||
The Hebes have also suffered a lot, despite above-average attention to water, but they haven't died. Strangely, the Hebes I planted last August, and which I never got round to irrigating, are quite happy and already flowering:
|
|||||||||||||||||||||||
|
|||||||||||||||||||||||
There must be some other source of water round there.
| Friday, 17 May 2013 | Dereel | Images for 17 May 2013 |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
|
Anniversary day
|
Topic: general | Link here |
17 May doesn't sound like a very important day, but for me it marks three changes in my life. From today's calendar entry:
Those three anniversaries roughly sum up my working life. Yes, I started earlier, and finished a little later, but Karstadt and Tandem were my main employers, between them accounting for about half my working life. All the funnier that in Germany, where people always start a new job at the beginning of the month, I should have started two consecutive jobs on 17 May.
|
Back to committing
|
Topic: technology | Link here |
This ports build has seriously held up other work I've had waiting. Today started on the backlog: merging changes in head back into the stable branch. In the Good Old Days with CVS, it was all manual, but now we're using subversion, and there's a merge command. Spent some time learning how to use that.
| Saturday, 18 May 2013 | Dereel | Images for 18 May 2013 |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
|
Winding down the house photos
|
Topic: photography, general | Link here |
My weekly house photos have become enormous. More and more, I'm going over to missing out many photos most weeks. So it was today, with the result that I had everything processed by lunch time. Somehow I didn't do much else all day long. Not as if a bit (or a lot) of garden work wouldn't be a good idea.
|
Enfield rainfall
|
Topic: general | Link here |
It's been raining a bit lately, and it occurred to me that one of the reasons we cooled off on the properties in Enfield was that I was concerned about how wet it would be there in the winter. So off to take a look. Result: no particular problem. I wonder if the evidence of raging torrents that we saw on 5 Inglewood Drive occurred in the really heavy rains that we had a couple of years back.
| Sunday, 19 May 2013 | Dereel | |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
|
Another power failure
|
Topic: general | Link here |
Another power failure this morning at 2:03.
|
Laziness and new houses
|
Topic: general | Link here |
I've been relatively lazy lately, much to the disgust of Peter Jeremy, who complained today that there's not enough to read in my diary. And I always thought it was too verbose. Today I did little. Yvonne found a newly listed house for sale in Tantaus Road, interestingly almost adjoining the property in Snowgum Road that we looked at a month ago. It looked quite good, and since it's only 1.5 km away, we went over to take a look from the outside. To our surprise it belongs to the Leahys, from whom we bought various things last year. The photos are really good; it makes the place look much better. They were in the garden, and though you're not supposed to speak to the sellers directly without the estate agent being present, it seemed rude not to say hello. They're moving to Ballarat for age-related issues. I wonder when we'll end up having to do that.
|
Getting my head around Subversion
|
Topic: technology, opinion | Link here |
I've been using revision control systems, notably RCS, for over 20 years, and for all sorts of things, including this diary:
It's been 5 years since the FreeBSD project moved from CVS to subversion, but I haven't really done any merging until yesterday. That worked fine, but today's didn't. Followed the instructions and discovered that my source file (/usr/src/lib/libc/stdio/printf.3) didn't have any mergedata at all:
=== grog@eureka (/dev/pts/3) /src/FreeBSD/svn/head/lib/libc/stdio 5 -> svn propget svn:mergeinfo -R printf.3
=== grog@eureka (/dev/pts/3) /src/FreeBSD/svn/head/lib/libc/stdio 6 ->
Is this the first time that anything has been merged to this file? Tried the next step:
=== grog@eureka (/dev/pts/3) /src/FreeBSD/svn/head/lib/libc/stdio 6 -> svn merge -c r242738 svn+ssh://svn.freebsd.org/base/head/lib/libc/stdio/printf.3 printf.3
So far, so good. What does the mergeinfo look like now?
=== grog@eureka (/dev/pts/3) /src/FreeBSD/svn/head/lib/libc/stdio 7 -> svn propget svn:mergeinfo -R
That was more than I expected, to say the least. But svn diff shows exactly my change:
=== grog@eureka (/dev/pts/3) /src/FreeBSD/svn/head/lib/libc/stdio 8 -> svn diff printf.3
Where did the other mergeinfo come from? Sent a message to the committers list and got back a very nicely worded reply that basically amounted to RTFM. Further up in the instructions was the admonition:
Never merge directly to a file.
Never, ever merge directly to a file.
Never, ever, ever merge directly to a file.
And I had read that, and even thought the repetition a little silly. So why did I do it anyway? Because that's the way our grandfathers did it with RCS. RCS is file oriented. CVS is too, really. But Subversion is change set oriented, and the mergeinfo is associated with the changeset, not with the file.
|
Pkgng: first impressions
|
Topic: technology, opinion | Link here |
So now my ports are roughly up to date. Can I use Pkgng to keep them up to date? Read the instructions, which are still a little preliminary. The best entry point seems to be the Pkg Primer, though there's also a Handbook section. First you need to run pkg2ng, which takes quite a while, and produced about 4,500 lines of output, not all of it pleasant:
At the end it came up with:
OK, that's simple enough. Remove and reinstall:
=== root@stable-amd64 (/dev/pts/1) /usr/ports 4 -> pkg delete texinfo-5.1.20130419
=== root@stable-amd64 (/dev/pts/1) /usr/ports 5 -> pkg delete texinfo
=== root@stable-amd64 (/dev/pts/1) /usr/ports 6 -> pkg_delete texinfo-5.1.20130419
=== root@stable-amd64 (/dev/pts/1) /usr/ports 7 -> pkg install texinfo
=== root@stable-amd64 (/dev/pts/1) /usr/ports 8 -> pkg install texinfo-5
=== root@stable-amd64 (/dev/pts/1) /usr/ports 9 -> pkg_add -r texinfo
=== root@stable-amd64 (/dev/pts/1) /usr/ports 10 ->
The third line (prompt 6) is subtly different: instead of pkg (Pkgng) it uses pkg_delete (old packages). That might be because it failed to register, of course. But pkg also can't find the remote packages. In other words, pkg doesn't seem to do anything useful—yet. Maybe it's related to the repository, but I didn't find a description of that. This was all in a virtual machine, and I had taken a snapshot before running pkg2ng, so I just restored it. I'll try again some other time.
| Monday, 20 May 2013 | Dereel | |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
|
Computing for non-techies
|
Topic: technology, general, opinion | Link here |
Our neighbour Jenny Bartlett, whom I only met 9 days ago, has been on a shopping spree and bought herself a digital (“personal”) video recorder and a Samsung Galaxy tablet, which she calls an iPad, possibly because that's what the salesperson called it. And she had difficulty installing it, so Yvonne persuaded me (“No, I will not repair your computer”) to go along and help.
The thing's wireless, of course, so apart from the pad she also had a wireless router to connect to her existing installation (laptop connected to IPStar satellite), and her real issue was how to connect things up. That wasn't difficult, apart from the fact that the router required a login to connect. But what do you do with a tablet? This is the same question I asked myself a year ago, and in Jenny's case it's not clear what advantage the tablet would have over her laptop, which, she says, she doesn't use. In addition, I had great difficulty with the touchscreen, which seemed very unresponsive. Chris Bahlo has a Samsung; we should send her by and see what she thinks.
The main issue with the DVR was wiring again, not helped by the lack of an antenna cable. The display (via HDMI) was surprisingly unstable, but it got the EPG and we were able to set it up to record. But what terrible documentation! In general, it's amazing how bad the documentation for everything is. I think there must have been about 20 pages of badly reproduced black and white documentation for the tablet, router and DVR put together, and much of my interaction with the DVR was through trial and error. I wonder how much use Jenny will get out of the devices.
|
More subversion pain
|
Topic: technology, opinion | Link here |
OK, now I have learnt not to talk to files when using subversion—or so I thought. The list goes on (my interpretation of the last one):
Never merge directly to a file.
Never, ever merge directly to a file.
Never, ever, ever merge directly to a file.
...
OK, sometimes you can merge to a file.
Still, the “sometimes” didn't apply here, so I started my next merge as described in the instructions. Something went wrong and I got a conflict where no conflict should have been. Tried moving the directory, checking out again, confirming that the mergeinfo had been reset, and starting again. No changes to the files. Tried again, and this time the mergeinfo had changed. OK, not a problem: that's what propdel is for. But no, I can't do that:
=== grog@eureka (/dev/pts/3) /src/FreeBSD/svn/stable/9/usr.bin/locale 102 -> svn propdel --revprop -r242743 svn:mergeinfo .
So somehow, without trying, I've managed to corrupt the repository metadata, and I can't fix it. This whole setup is too fragile: it shouldn't be possible to make this kind of mistake. I seem to remember having similar problems in Rocksoft years ago, but I don't seem to have noted it in my diary. It looks as if I'll have to blow away the checkout and start again.
|
Watch calibration
|
Topic: general | Link here |
How accurate is my new watch? Hard to say. My old watch gained about 15 to 20 seconds a month. I thought I had set the new one to be 2 seconds behind the old one (because that one was currently about 2 seconds fast), but when I checked today the new one was 2 seconds faster than the old one. Is it gaining a lot, or did I just set it the wrong way round? Today at 15:00 odd I set it exactly to the time on eureka (NTP synchronized). Now I'll have to wait a few days and see how it drifts.
| Tuesday, 21 May 2013 | Dereel | |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
|
Another new house?
|
Topic: general, opinion | Link here |
Round to Tantaus Road this morning to meet with estate agent Phil Hayward and look at the Leahy's house. It's not obvious from outside, but in fact it's quite nice. But it seems that any house I buy between now and my 65th birthday (end of September) will attract in the order of $15,000 to $20,000 “Stamp duty” for the transaction; after my 65th birthday it will drop to roughly 0. So it's not a good time to buy houses.
After that, Phil came around to our place. Once again we had the case of a Really Impressed agent, who agreed with the others about what we could get for the house. I'm left wondering why we really want to move.
|
Network problems solved?
|
Topic: technology | Link here |
I can't say that my network connection has been good at any time this year, but the last 5 days were better than average, and the link stayed up the whole time. That was too good to last: round midday today things got worse again, including a lot of this kind of message, which I don't see very often:
My best bet is that this is an indication of poor reception, but the reported signal strength was better than average. Possibly it's related to my TV reception problems, though I didn't have any of those today.
So it was good to receive a message from Scott Weston: VCAT has rejected Wendy McClelland's objection to the radiation tower: it can go ahead. Hopefully it'll be done soon.
|
Subversion POLA
|
Topic: technology | Link here |
I'm still puzzled by my problems merging fixes with subversion. Started again by checking out a complete source tree on one of the FreeBSD project machines, not helped by the flaky network, and then trying the merge the way it's supposed to be. First, which revisions do I need?
=== grog@freefall (/dev/pts/1) ~/9/usr.bin/locale 5 -> svn mergeinfo --show-revs=eligible svn://svn.freebsd.org/base/head/usr.bin/locale
OK, then the merge command should be:
=== grog@freefall (/dev/pts/1) ~ 21 -> svn merge -r242743:r243202 svn://svn.freebsd.org/base/head/usr.bin/locale 9/usr.bin/locale
Why the conflict? Still, selecting “theirs-conflict” should mean “in case of conflict, take the version in head”. And, sure enough, it accepted the update. Correct? To compare it, I copied the files to my local machine, where I have a copy of head checked out. The results weren't quite what they should be:
=== grog@eureka (/dev/pts/3) /src/FreeBSD/svn/stable/9/usr.bin/locale/freefall2 124 -> diff -wu locale.1 /src/FreeBSD/svn/head/usr.bin/locale/locale.1
Why is that?
=== grog@eureka (/dev/pts/9) /src/FreeBSD/svn/head/usr.bin/locale 10 -> svn blame locale.1
So the missing lines are from revision 242743, the first revision that I had specified. And looking more carefully at the merge command, I see:
=== grog@freefall (/dev/pts/1) ~ 21 -> svn merge -r242743:r243202 svn://svn.freebsd.org/base/head/usr.bin/locale 9/usr.bin/locale
So this is a violation of POLA. -r242743:r243202 excludes revision 242743. It seems that to do it right I should have written -r242742:r243202. To check that I need to blow away the checkout and start again, since even if I remove the modified files, svn's metadata claims that they've been applied. And just checking out the tree takes a couple of hours. Mañana.
| Wednesday, 22 May 2013 | Dereel | Images for 22 May 2013 |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
|
Subversion: solution
|
Topic: technology, opinion | Link here |
So it seems that my problems with subversion were due initially to my misinterpretation of the -r switch: the form -r242743:r243202 excludes revision 242743. Does it work correctly if I write -r242742:r243202? Potentially not, since revision 242742 doesn't relate to these files. Spent another hour checking out the source tree and tried it. Success!
=== grog@freefall (/dev/pts/8) ~ 15 -> svn merge -r242742:r243202 svn://svn.freebsd.org/base/head/usr.bin/locale 9/usr.bin/locale
So far, anyway. Committing was a different matter:
=== grog@freefall (/dev/pts/8) ~/9/usr.bin 16 -> svn ci locale
Huh? Oh, yes. Since the security incident last year, we no longer keep private keys on freefall. That means that I couldn't use svn+ssh to connect to the repository, so I used svn as well. But I couldn't check in that way. Reluctantly decided to check out the complete repository on eureka, my local machine. To my surprise it took hardly any longer than it did on freefall, and though it checked out a total of 1.4 GB of data, it only caused about 300 MB of network traffic. But things still didn't work:
=== grog@eureka (/dev/pts/3) /src/FreeBSD/svn/stable 150 -> svn merge -r242742:r243202 svn://svn.freebsd.org/base/head/usr.bin/locale 9/usr.bin/locale
That's what comes from copying and pasting: The message didn't help, but I still had svn instead of svn+ssh. After changing that, it finally worked.
So: what have I learnt?
Understand that the -r range excludes the first revision.
There are many ways to merge changes. They all seem to work, but they place the metadata in different places, making it difficult to keep the repository tidy. This is part of the background of the admonitions in the instructions
Understand that you can't hack around on a checked-out repository. Once you merge, you're stuck with it, because there's no way to reset the metadata and start again.
I find last two points a great disadvantage of subversion. It's too easy to make mistakes, and almost impossible to recover from them.
|
Installing the latest FreeBSD stable
|
Topic: technology | Link here |
So finally I have a relatively up-to-date installation of FreeBSD and ports on a virtual machine. Now I should be able to copy it to other machines to upgrade them. For some years now I've maintained two root file systems, including /usr: the one I'm using and the one that I can use to upgrade, which I mount on /destdir. The file systems look like this, and you simply boot from the appropriate partition:
So yesterday I copied the new system to the alternate partition on teevee, my TV computer, and tried booting from it. That involves modifying the /boot/loader.conf file on the first partition, to tell it which partition to boot from. For some reason I had this in there:
That's wrong: it should have been disk0p4. So I had to manually correct it, after which the system came up relatively well; only the remote control didn't work, something that always seems to happen. So I reverted to the old system to watch TV last night. Today corrected /boot/loader.conf and tried again. Not what I expected:
|
|||||||||||||||||||||||
|
|||||||||||||||||||||||
What's that? Line 1016 of /usr/src/sys/boot/common/module.c is the very last line, so it's not clear what really happened here. It also happened before I got the loader prompt, so there was just no way to start the machine. Took the disks and put them in monorchid, my test machine, where I was able to boot from another disk and examine things. After some testing, discovered that it was directly the result of specifying the correct partition! If I specified the wrong one and then corrected it, it would work without any trouble.
So what's the problem? loader too old? More investigation needed, but I want to do it without the danger of not being able to watch TV in the evening. Once again mañana.
| Thursday, 23 May 2013 | Dereel | Images for 23 May 2013 |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
|
The joys of a fresh installation
|
Topic: technology, opinion | Link here |
It seems that yesterday's boot problems were related to an older version of /boot/loader that didn't properly understand GPT partitioned disks, so it seemed reasonable to install the newer version from the fresh installation instead. Care! The kernel is in the same directory hierarchy—and somehow I had finger trouble and blew my complete kernel directories of the old installation away!
Spent some time looking for a backup, which proved to be incomplete. I'm not as religious about backing up teevee and cvr2 as I am with other machines, in particular because almost the entire disk contents are video recordings, but clearly I should improve. So the only course is to do what I wanted anyway and ensure that the new system would run correctly. And indeed the loader worked as expected, and the machine came up. Next was to get the remote control working. I've always had difficulty with that, and it was helpful to look back on my experiences last November and two years ago—unfortunately, not helpful enough. I had the correct device, the correct configuration file, and the correct invocation, but nothing happened. The messages in /var/log/lircd were subtly different due to the new version of lircd, so I tried running the old version. In each case, nothing came across. Running ktrace on irw showed that it just wasn't reading anything. Hardware problems perhaps? Remote control batteries empty, or sensor not connected? No, the activity LED worked normally, and the probe messages found that the sensor had been detected:
So what is it? I still don't know. About the only other thing that has changed is the kernel. Do we have USB problems in recent kernels? Looks like this will need some serious debugging. It's a good thing that I can get by without the remote control: I can also use the keyboard.
That wasn't the only problem. Various other configuration things needed fixing. I have a custom version of mplayer, so I moved that across from the old system. But running the invocation script was surprising:
=== grog@teevee (/dev/pts/2) /spool/Docco 28 -> mmp World_at_War-Whirlwind
Syntax error at line 1? That reads:
And looking at line 41 didn't help much either:
Well, not at first. Looking more closely, there was indeed a missing quote at the end: it should read
The RCS control files show that that error had been there for at least a year. It seems that the shell is now finding errors that it somehow missed before.
One of the things that I had expected from the new system was better browser support. Sure enough, Chrome and Opera started up with no trouble beyond the fact that I can no longer find home pages for these browsers. But when I clicked on the firefox menu, nothing happened. Incorrect name? Should I have called it firefox21? Tried running it from the command line: no, it's really firefox, but it didn't behave as expected:
=== grog@teevee (/dev/pts/2) /spool/Docco 44 -> /usr/local/bin/firefox
Doesn't that give you the warm, fuzzy feeling that everything is under control? Decided to recompile it on stable-amd64. More segmentation violations:
So I installed the firefox package from the FTP site. Version 20, now out of date. Not so much an issue in itself, but every time it starts, the damn thing has to tell you so and offer to do what it can't, to install a newer version.
Somehow all this should be easier. Not for the first time I wonder why I stick with FreeBSD. But there's a good answer. To misquote Jawaharlal Nehru:
FreeBSD is good. I say this because other systems are worse.
Other minor issues included not starting the xterms when X starts, and Emacs not reading in its configuration files. Those were both path issues: for some reason, I explicitly requested /usr/local/X11R6/bin/xterm in the startup files, and now it's /usr/local/bin/xterm. In the latter, I had ~/.emacs pointing to (NFS mounted) /dereel/home/grog/.emacs where it should have been /eureka/home/grog/.emacs. But I'm making progress.
|
Somebody to weed the garden
|
Topic: gardening | Link here |
The weeds in the garden are getting so out of control that I'm completely discouraged. I've finally decided to call in somebody to do it for me, and today Bentaz (Ben) and Rowan Brooks came along to take a look. Looks like they can do it for an acceptable price, but there's also Jordan Dickinson, who did the pond for us 18 months ago. But hopefully we'll have the garden looking a lot better soon.
| Friday, 24 May 2013 | Dereel | Images for 24 May 2013 |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
|
More installation refinement
|
Topic: technology, opinion | Link here |
More work on the new teevee today. Most of it was mundane, but I finally solved my lirc problems. The most obvious recognition is that running irw with ktrace is not useful: it just talks to lircd via a socket, and it's lircd that decides what to send. Running lircd with ktrace proved that yes, indeed, it was reading from /dev/uhid0, not surprisingly exactly what I saw two years ago. So it looks like a configuration issue after all. Went looking at the configuration file, and discovered that I hadn't installed the correct version after all: no reference to the dvico keycodes at all. And yet lircd had happily fired up and claimed to be running in dvico mode. How can that work? It seems that the configuration information for individual remote controls are stored subdirectories of /usr/local/share/examples/lirc/remotes, in my case /usr/local/share/examples/lirc/remotes/dvico/lircd.conf.fusionHDTV, and /usr/local/etc/lircd.conf should have had a line like
But it didn't, and lircd didn't complain. I wonder what it thought it was doing.
I could have made that change, but based on prior experience I preferred to try my existing (correct) configuration file first. And sure enough, that worked. Clearly a couple of mistakes on my part. But what kind of software pretends to work when its configuration information is missing?
|
eBay strangenesses
|
Topic: general | Link here |
A notification in the mail today (“You have mail”): a package to pick up from Napoleons post office, my personal signature required. I wish people wouldn't do that: it's a 40 km round trip, which costs money. Off to find that it was a watch I had bought on eBay on Monday—it had arrived from Hong Kong after only 4 days. But this one seems to been sent by a more expensive method—it actually had stamps on it, to a value of $21.10, about $2.80 Australian. By contrast I'm waiting—again—for an order of batteries, this time ordered 11 days ago.
|
Powerline Ethernet speed
|
Topic: technology, opinion | Link here |
I've come to terms with the fact that my TP-Link TL-PA411 powerline Ethernet adapters will only give me a transfer rate of about 7 to 7.5 MB/s in real world applications, despite the claim of 500 Mb/s. Even under ideal conditions I haven't got more than 100 Mb/s out of them. But today, for some reason, the transfer speed dropped from 7 to 5 MB/s. Why? Nothing else had changed. Suggestions were made that it might be due to electrical motors, which is plausible, given that it's now colder and the air conditioners are running almost constantly. To be observed.
| Saturday, 25 May 2013 | Dereel | Images for 25 May 2013 |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
|
The new Hugin
|
Topic: photography, opinion, technology | Link here |
Part of my work last week was polishing the port for Hugin beta, which resulted in my leaving it installed. I still have serious objections to the layout, in particular the bleeding of status text into the top left of the panorama preview:
|
|
||||||||
|
||||||||||||||||||||||
Still, it looks like I'll have to put up with it, so carried on and processed a couple of panoramas. Or so I thought. After stitching the first panorama, my scripts didn't find the result. There's always an issue that I might forget to actually perform the stitch phase, but I save the project files anyway, so I tried again later. Again no panorama. Then I noticed one of the new things about the latest version. Previously the “Assistant” (simple interface) just went ahead and stitched a straightforward panorama. Now I get the choice—maybe:
|
||||||||||||||||||||||
This display is completely new, and it only appears if you choose the “Assistant”. In this case, it doesn't give me the first (previously default) option “Exposure corrected, low dynamic range”. There's still the “Stitcher” tab, which appears to be unchanged, and it represents things very differently, including allowing me to select that option:
|
|
||||||||
Did a bit of playing around and discovered the matching targets in the project Makefile. For a panorama garden-centre.jpeg the project file is called garden-centre.pto, and the Makefile is called garden-centre.pto.mk. It requires GNU make and offers the following targets:
| garden-centre.jpeg.tif | Exposure corrected, low dynamic range | |
| garden-centre.jpeg_fused.tif | Exposure fused from stacks | |
| garden-centre.jpeg_blended_fused.tif | Exposure fused from any arrangement | |
| garden-centre.jpeg_hdr.exr | High dynamic range | |
The exact names are shown close to the top of the Makefile. Here I've omitted some of the lines that seem irrelevant, notably those ending in _PREFIX (part of the names below) and _SHELL (identical to the names below):
Still, maybe there's an advantage in the suggested “Exposure fused from any arrangement”, whatever that means. One of the things I've been doing is taking photos with different exposures, and maybe that's what triggered it. So I tried both approaches for three panoramas. Here in each case the “Exposure fused” on the left and the “Exposure corrected LDR” on the right, both without postprocessing. Move the mouse over each image to change it to the other:
|
|||||||||||||||||||||||
|
|||||||||||||||||||||||
|
|||||||||||||||||||||||
|
|||||||||||||||||||||||
|
|||||||||||||||||||||||
|
|||||||||||||||||||||||
Which is better? It's hard to say. I've had difficulties in the past with the exposure in the right of the second image (garden centre), and here it seems to be better exposed, though maybe the Salvia in the foreground is a little too bright, and the exposure compensation is possibly at the expense of the area to the left of centre. On the other hand, the third image (garden south-east) definitely looks better. More experimentation needed.
And what about the new window with the pretty icons? I don't really understand them. Particularly the “Exposure fused from any arrangement” is confusing. Arguably the specification of height is more understandable, but it's adjustable. How does it adjust? In the “Stitcher” tab you get top and bottom, though not the total height.
One other thing I hadn't seen before: when stitching the verandah panorama in “Exposure fused from any arrangement”, it seems it decided to do it in two steps. The first was to create these two partial images, which it then stitched together in a second step:
|
|||||||||||||||||||||||
|
|||||||||||||||||||||||
I wonder why. Possibly it's intended to make things less memory-intensive, but I've stitched much larger panoramas in a single step before.
|
Powerline Ethernet puzzles
|
Topic: technology, opinion | Link here |
My TP-Link TL-PA411 powerline Ethernet adapters continue to transfer even more slowly than initially. One reason might really be interference from electric motors. When the dishwasher was running, I got only 2.8 MB/s. But the best I could get was still round 5.2 MB/s. But today I tried two transfers at a time, and discovered surprising things:
| Number of | Individual | Aggregate | ||
| transfers | Speed | Speed | ||
| 1 | 5.2 | 5.2 | ||
| 2 | 4.2 | 8.4 | ||
| 3 | 3.0 | 9.0 | ||
Why is that? I've associated that with packet loss and corresponding delays in the past, but wireshark doesn't show any retries. These are half-duplex links, of course, but I can't see why concurrent transfers should get a better aggregate result.
|
Refining cassoulet recipe
|
Topic: food and drink, opinion | Link here |
It's been nearly 7 years since I first put my cassoulet recipe together. Today, to my surprise, Yvonne suggested that I cook one: she says she doesn't like it, and I had wanted to cook something else. Threw some stuff together and discovered that the recipe needs refinement. I put the original recipe together from various French cookbooks, and it took me some time to get the quantities right: the original recipes use so few beans that they could hardly be seen at all, let alone be the dominant part of the dish.
Today things were different again. How long do you cook the beans? The recipes that go into that kind of detail (Bocuse doesn't) suggest at least 3 hours. Today I was using relatively fresh beans, and they were almost overdone after a little after an hour. And once again I had far too much liquid left over at the end. In addition, I've decided that there were far too many onions in the dish. Time for more adjustments.
|
David Yeardley and wife number 5
|
Topic: general, opinion | Link here |
David Yeardley is an enigma. He's pushing 63, and is currently on his fifth wife. After divorcing Chris (now Bahlo), he married first a Chinese woman whom I never met, then Fifi, younger than his own son, and now it's Tuyết. On the face of it, Tuyết is a nice person, but she's been uprooted from Viet Nam to here, she doesn't have a driver's licence, and she doesn't have any company. And David is away for 5 weeks at a time, during which she has basically only Chris as company.
Currently David is back, and for some reason Tuyết is consumed with jealousy regarding Chris. Chris has already been thinking of moving out, against which financial reasons speak; David made himself particularly obnoxious a few months back. But so far her relationship with Tuyết has been OK. My concern has really been how Tuyết will adapt to life here; it's so foreign to her.
But now Tuyết's jealousy seems to be the biggest problem, and they're openly discussing that Chris move out as soon as possible. She's planning to move to the block of land that we looked at a couple of times, but it'll take a while, of course.
Tonight Chris came over for dinner, the first time she's done it alone when David has been here. It seems that despite all the tension over there, he expects her to eat with them. But as things deteriorate, he magnanimously allowed her to come.
When she got back, it seems, Tuyết made a scene, demanded she left, and that David should not allow her to build on the land. David refused her, but I'm amazed. Why should she pick on Chris? Her demands are ridiculous, since Chris is half-owner of the property. They're also very negative for Tuyết, since currently she's dependent on Chris for transport and a number of other things. David is leaving again in about 3 weeks. God knows what will happen then, but it doesn't look good.
| Sunday, 26 May 2013 | Dereel | |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
|
Still more build failures
|
Topic: technology, opinion | Link here |
I'm still trying to find out why clang got a repeatable SIGSEGV while trying to compile firefox. Discussion with Dimitry Andric showed that he had no such problems. It's not hardware, because it happens both in a VM on eureka, and also on teevee. Maybe a transient issue has since been fixed? No harm in building the latest world. But on installation I got:
How did that happen? Did the buildworld fail without me noticing? Tried buildworld again and got:
This time it was a different compiler that crashed, and in a different place. A lot of this looks like hardware issues, but everything else works. Decided to go back to a previous snapshot on stable-amd64 and bring it up to date. That worked for a while, but when building the kernel I got:
This time as! That's three different programs that have died. All of this suggests some kind of hardware issue, especially since it worked when I repeated the build. But that doesn't explain the fact that the first crash was repeatable on different hardware. More Heisenbugs. And time to think again about new hardware.
|
NBN tower interest
|
Topic: general, technology, opinion | Link here |
Now that the radiation tower has been approved, the press is getting interested. Had a call from Tom Cowie of the Ballarat Courier with some questions which he paraphrased in an article later in the evening.
Since then they've hidden the article behind a paywall. Here's the text:
A NATIONAL Broadband Network wireless tower has been cleared for construction in Dereel, after the Victorian Civil and Administrative Tribunal (VCAT) quashed residents’ objections.
The ruling was handed down on May 17 after a hearing last month and means the area – a notorious mobile phone blackspot – will soon have access to high speed broadband.
The planning permit, initially approved by Golden Plains Shire Council, allows for a 40-metre-high monopole and communications facility to be built on Colac-Ballarat Road.
Dereel resident Greg Lehey, a supporter of the tower, said NBN had first been discussed in the area two years ago.
He said it was a step forward for communications in Dereel, but also added he won’t believe it’s happening “until the thing is actually built”.
“There’s no coverage of mobile phones in Dereel, internet coverage here is either wireless or satellite and is flakey. When the tower is finished we will finally have a proper internet connection,” he said.
Mr Lehey, who uses the internet instead of a phone line through Voice Over IP technology, said he thought the tower would be built within months.
He also hoped the decision would pave the way for a mobile phone tower to be built in the area, which many locals have been calling for.
A VCAT summary said the planning permit met the relevant principles in the telecommunications code of practice, including whether the tower had been located to minimise visual impact.
Residents objected to the tower on several other grounds, including potential health risks, lack of community consultation and property values, however, VCAT said these did not fall under the remit of planning.
NBN fixed wireless towers have been controversial in recent times, with a Bunninyong tower knocked back last year by the City of Ballarat due to its height, location, design and appearance.
Local residents in Mt Helen are also fighting a tower that has been proposed in their area, saying it will disturb the amenity of the area.
The NBN rollout is under way in Ballarat, with many properties already able to get access to the network.
| Monday, 27 May 2013 | Dereel | Images for 27 May 2013 |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
|
Still more build pain
|
Topic: technology | Link here |
After rebuilding my kernel yesterday, tried once again to build firefox 21. It took all day, but it worked! So a little later I tried installing the image on teevee:
“Error while running startup cache precompilation”. What does that mean? Poorly specified error message, of course, but what else? And how do you fix it?
One thing that's not typical, of course, is to build the port on one machine and then try to install it on another machine. But it failed on stable-amd64 as well—maybe as the result of damage done by the first installation? The only way to find that out was to start again from scratch and compile another 5 hours overnight. Aren't computers fun?
|
More NBN tower stuff
|
Topic: general, technology, opinion | Link here |
VCAT have released the decision about the radiation tower, which makes interesting reading. I've read similar decisions before, and this one is different: it takes Wendy McClelland's behaviour to task in no uncertain manner:
At a Practice Day Hearing in January 2013 to consider the validity of certain statements of grounds, Deputy President Gibson stated that a number of issues raised in the statements were not planning grounds to be relied upon at the hearing. These issues were:
Deputy President Gibson commented that it is not the Tribunal’s role to enquire into the subject of sickness from high radiation exposure or to consider the basis of Ms McClelland’s disabilities.
In her reasons Deputy President Gibson made comment about the issues listed above and identified the relevant planning considerations as being:
Despite these directions, Ms McClelland maintains her concerns about electro-magnetic radiation, the whole process involved in the assessment of the application, the use of alternative technologies and the NBN programme itself. Although she tabled and spoke to a very lengthy submission of 53 pages, much of the material has no bearing on the matters which I am required to consider under Clause 52.19. For example references to the minimum subdivision size in the Farming Zone and the effect of the Bushfire Management Overlay have no relevance to this application and cannot be used to mount a case against the telecommunications facility. Her commentary about the position of the Dereel Township sign and its recent relocation implied some sinister or underhand motive but as with much of the other material in her submission it has no bearing on the issues I need to consider.
The decision wasn't a carte blanche, though: it did make some minor adjustments to the permit. Some appeared to be minor or reinforcing what was already in the permit, but these two appear to be additional:
Prior to the commencement of works on the site, an amended site plan must be submitted to Council showing access off Dereel - Rokewood Junction Road as opposed to Colac - Ballarat Road to the satisfaction of the responsible authority. Once the plan is submitted, it will be endorsed and form part of the planning permit.
Before the development starts, an amended landscaping plan to the satisfaction of the responsible authority must be submitted to and approved by the responsible authority. When approved, the plans will be endorsed and will then form part of the permit. ... The landscaping must be generally in accordance with the plans submitted with the application but modified to show:
(a) Retention of the existing large trees along the boundaries of the subject site.
(b) An additional row of trees along the Colac – Ballarat Road perimeter of the compound consisting of a minimum of six trees with a height at maturity of 12m.
(c) The establishment of a row of semi-mature large shrubs around the perimeter of the compound to partially screen the development.
Still, hopefully that won't delay things longer than the time frame of 3 months total that I've been told after this decision became known.
Some discussion on IRC about the matter, and I dragged out the plans that I had received last March. Edwin Groothuis suggested I took a series of photos of progress, so off to take a look at the site before they started. Here a view from Rokewood Junction Road, near where the entrance is now required to be:
|
|||||||||||||||||||||||
|
|||||||||||||||||||||||
|
|
|||||||||
The green object in the last detail appears to be something like a plastic water container. It's almost exactly where the tower will be. From the opposite direction on the Colac-Ballarat road it looks like this, again with the watering can to the right:
|
|||||||||||||||||||||||
|
|
|||||||||
Other interesting details about the NBN are that it's now confirmed that fixed wireless will offer speeds of up to 25 Mb/s down and 5 Mb/s up from the middle of this year, easily soon enough for me. In addition, a number of people are commenting on the “Coalition”'s replacement NBN plan, without exception negatively. Sortius goes into much detail about attenuation and the real speeds of FTTN, and Steve Jenkin is suggesting that they might change their tune before the election and offer fibre after all. That's possible, I suppose. It would be much better if they were to bring fibre to Dereel.
|
Hugin stitching methods
|
Topic: photography, technology, opinion | Link here |
I noted a couple of days ago that Hugin has changed its default stitching methods, and that it now offers “Exposure fused from any arrangement” for many of my panoramas, to the exclusion of the default “Exposure corrected, low dynamic range”. It seems to choose that when the exposure of the component images differs by more than a specific amount. In the case of the NBN site, the component images had exposures that differed by 4 EV, so that seemed reasonable, and I stitched the panorama like that. Here for comparison the fused and the corrected panoramas. Move the mouse over each image to change it to the other:
|
|||||||||||||||||||||||
|
|||||||||||||||||||||||
Which looks better? A matter of taste, maybe. The first is more uniform and keeps the details of they sky better. But there's a big downside, as a detail of bottom right shows:
|
|||||||||||||||||||||||
I took the panorama hand-held, and there were a number of matching errors. Normally Hugin does a very good job of removing the problem, but here multiple images are fused, and the discontinuities are very evident. I'll have to compare more carefully next time I take this kind of panorama.
| Tuesday, 28 May 2013 | Dereel | Images for 28 May 2013 |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
|
Day's port pain
|
Topic: technology, opinion | Link here |
Into the office this morning and found that yes, indeed, not only had firefox compiled, but it had also installed. It compiles! Ship it!
The truth was more sobering:
=== root@stable-amd64 (/dev/pts/2) /usr/ports/www/firefox 10 -> firefox
Sigh. That's where I was last week. What do I do now? I'm thoroughly fed up, so I didn't do anything more today. But there are a number of possibilities:
Wait a bit and try again.
Try again with gcc.
Try rebuilding on a different machine, possibly without any other ports at all.
Try debugging the crash.
I'll decide on the course later.
|
More drama at the Yeardleys
|
Topic: general, opinion | Link here |
Yvonne over to the Yeardleys today to help Chris and Gordon (the farrier) with some horses. While she was there, she experienced the next stage of the David and Tuyết story. Lots of screaming inside the house, David comes out in tears (!) and asks Chris (of all people!) to go and have a word with her: Tuyết is somehow consumed with jealousy towards Chris and has asked David to refuse to give her the other plot of land and to sell all the horses. It's not clear how she can expect that when Chris is co-owner of both the horses and the land. Has she not understood that detail, or is she just ignoring it?
There's some reason to believe that she's not quite right in the head. Firstly there's no way she can expect these demands to be met. More importantly, though, David is only here half the time. The rest of the time she was dependent on Chris for just about everything. Now that Chris has moved out (not far, admittedly), and after the way she has treated her, it's not clear what will happen to her when David leaves in 3 weeks. In fact, he's off to Singapore next week. What will happen then? She could go with him, but what about Min Chau? How will she get to school? Although she's had a brand new car for months now, Tuyết still hasn't done more than a couple of driving lessons, and she definitely won't have her driver license by the time David leaves.
And then there are all the things that she accuses Chris of, including breaking up David's last 3 marriages (herself apparently included). How can she come to such a stupid conclusion? Yvonne thinks she might have a brain tumour, and Chris wonders if it's not bipolar disorder. Certainly it doesn't make any sense to me.
The other question is whether there are any long-term prospects. Maybe she doesn't want them. Chris did go and talk to her, and warned of the consequences of continuing like this: “Do you want to kill him?” “Yes!“.
David isn't Yvonne's favourite person, by a long shot, but now she's feeling sorry for him. The next few weeks will be interesting, even if Tuyết comes back to sensibility. I don't see Chris helping her any more, and it's difficult to see her staying here longer than that.
|
Repairing the Mecablitz
|
Topic: photography | Link here |
It's been months since I last looked at my Mecablitz 58 AF-1. It has been nearly finished all that time. All I need to do is solder the connections to the tube and reassemble it. Why have I been waiting so long?
The main reason is that I wanted to get some photos of what I'm doing, and they're fiddly, so I used my standard solution: procrastinate. Today I finally got as far as soldering the connectors to the tube. In the process it's worth noting that despite the care I took while disassembling the unit, I missed one detail. For some obscure reason one wire is connected to the tube via a flange, while the other is soldered directly to the contact on the tube. Here's the end with the red wire (on the right seen from behind), first from behind, showing the flange, and then from the side after soldering it to the end of the tube:
|
|||||||||||||||||||||||
|
|||||||||||||||||||||||
At the other end, the black wire is just soldered to the contact of the tube:
|
|||||||||||||||||||||||
Why?
|
Speeding up DxO
|
Topic: photography, technology, opinion | Link here |
Talking on IRC about photography this afternoon:
GPU acceleration? I hadn't heard of DxO Optics “Pro” offering that. Took a look at the preferences menu, and sure enough, found:
|
|
||||||||
The fact that the selection is greyed out is understandable: this machine only has on-chip graphics. But I have a spare nVidia 9500 GT card floating around. That would be worthwhile—if it fits. The trouble is that dxo (the machine) has a half-height form factor, while the card is full height. So no go in this machine.
Spent some time discussing the matter on IRC. One of the people, known only as ^X, offered me what could be an appropriate card. But is it supported? Took a look at the user manual, available on the web, but not directly. It had exactly two references to GPU acceleration:
1.3.1. System requirements
For the best results with DxO Optics Pro, your computer must conform to the following minimum specifications:
Microsoft® Windows XP 32 bits, Windows Vista® (32, 64 bits), Windows 7 (32, 64 bits). A 64-bit system is required for RAW images of 25 Megapixels and more.
Intel® Pentium® 4, Intel® CoreTM 2 Duo or higher, AMD® AthlonTM 64 X2 or higher.
Support of DirectX 9.0.c or higher for Windows operating systems. 256 MB VRAM is recommended to support GPU acceleration.
Then, further down:
Display and Process
GPU acceleration lets you make use of the graphics card (GPU) to speed up the display. You should disable this control if it causes your display to be unstable.
I had expected a bit more detail than that. Found a supported equipment page, but the equipment to which it referred was only cameras and lenses. Finally, with Google's help, found a system requirements page:
Graphics card: Support DirectX 9.0.c or higher; 256 MB VRAM recommended to support GPU. However, GPU
Warning: GPU acceleration is available only for 64-bit platforms.
And yes, that's the way it reads. But only 64 bit platforms! Wouldn't you think they'd make that more visible? Why isn't it in the user manual? It might also explain why the DxO “support” people recommended moving to 64 bit, but they didn't mention anything about GPU acceleration when they did. Why didn't they? I asked specifically what equipment I would need to run DxO.
Further on there's the information I was really looking for, sort of:
The following graphics cards are fully compatible with OpenCL
Radeon HD 5xxx, 6xxx, 7xxx, GeForce 2xx, 4xx, 5xx, 6xx.
You should choose a powerful card (for example, the GeForce GTX 460, Radeon HD 58xx or higher) in order to enjoy significant improvement. The new Radeon HD 7xxx and GeForce 6xx cards have not been tested yet, but should function very well.
So how does my video card compare? Videocardbenchmark gives it a score of 290, barely 10% of the GTX 460's 2676. Even the new GT 640 card I bought in the summer has a score of only 1309. Is it worth putting in an old card? Certainly nothing to hurry with.
| Wednesday, 29 May 2013 | Dereel | |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
|
Chasing the firefox problem
|
Topic: technology | Link here |
After a day without ports pain, back to investigating the problem today. It seems that some people, but not all, can reproduce it. Callum Gibson can't, and he sent me a package—unfortunately in pkgng format, which I can't install. I could upgrade to pkgng again, but that seemed counterproductive.
Jürgen Lock suggested looking at the core dump. I had deliberately put that alternative at the bottom of the list, but just pointing gdb at it doesn't cost much. To my surprise, the stack trace was short:
Clearly the last valid address was frame 1. So maybe it would be worth debugging. But first I decided to compile with gcc to see if the problem was related to clang,
That took all day again, not helped by a VirtualBox crash which left the VM dead in the water:
Finally it was built:
=== root@stable-amd64 (/dev/pts/2) /usr/ports/www/firefox 27 -> firefox
The stack trace was pretty much the same as before. So it's not clang's fault. What next? Another easy (but not fast) test occurred to me: try it on the i386 architecture. That required bringing stable-i386 up to date, which again took all day.
| Thursday, 30 May 2013 | Dereel | |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
|
Firefox debugging continued
|
Topic: technology, opinion | Link here |
Started with building firefox on the i386 architecture this morning, to the disapproval of people on IRC. The problem has only been reported against the amd64 architecture; what use is it to try on i386? That made sense; so the next attempt was to build under amd64 with clang and symbols.
Building firefox takes hours at the best of times. This was not the best of times:
Signal 9? SIGKILL? Who sent that. It wasn't me. But look at that time! Over 6 hours! By that time it was night. Another day wasted. I'm reminded of what I wrote in “Porting UNIX Software” 20 years ago:
How long does it take?
It is very difficult to gauge the length of time a port will take to complete. If a port takes a long time, it's not usually because of the speed of the machine you use: few packages take more than a few hours to compile on a fast workstation. Even the complete X11R6 windowing system takes only about 4 hours on a 66 MHz Intel 486 PC.
How times have changed! Now the computers are nearly 1000 times as fast, but build times haven't shortened.
One other issue is the load that the VM places on the host machine. It has memory of only 2 GB, and I recently upgraded from 6 to 8 GB of memory in eureka, so it should miss that much memory. But things almost grind to a halt. I don't need to do it like that; I have a real machine lying around, so started bringing that up to date. How long all of this takes!
| Friday, 31 May 2013 | Dereel | |
| Top of page | ||
| previous day |
|
Current network problems
|
Topic: technology, opinion | Link here |
My current 3G network was quite reliable up to about the beginning of the year, but since then the quality has dropped markedly, and Internode just don't want to know. Roll on the radiation tower! Today was particularly bad. Each of these represents a drop in connections, which kills my TCP connections and takes about 5 minutes to recover:
In that connection it's interesting to note a newspaper article in which the CEO of Optus acknowledges that service is not what it should be. A good place to start would be with their coverage maps. Andy Snow has moved even further from civilization than I, though he makes it even worse by a 150 km commute three times a week. And wireless coverage? Even worse than here, it would seem. I tried getting information from Optus' coverage map, but after 15 minutes it had not responded beyond upshifting the address I had entered. Was that my network connection (also Optus' responsibility) or the page? Tried my own address and got an almost immediate response. It seems that I had entered the wrong house number, and this was the page's inimitable way of saying “No such address”. That really makes fun of the statement at the top of the page:
Our commitment to you has always been to deliver a better network experience, today and tomorrow.
|
Future network problems
|
Topic: technology, opinion | Link here |
In the short term, the radiation tower will be a solution for the network connectivity problems. But for the foreseeable future that will offer a maximum speed of 25 Mb/s, and it will max out under 100 Mb/s—surprisingly similar to the “Coalition”'s plans for a castrated NBN. One of their biggest complaints was slow rollout; yesterday NBN produced a report for the Senate Budget Estimates Committee showing that they had greatly improved rollout, in some cases increasing total deployment by over 30% over the total up to the previous month. They started slowly, but they seem to be catching up now. What did the committee talk about? Asbestos and personnel issues.
But that page also had a poll button. “Do you prefer the Coalition's NBN policy?”. Not very trustworthy; you can answer as many times as you want. But it shows 19% for the Coalition and only 81% against. Is that correct? Went looking for surveys, and found, of all things, a Launch of the Coalintion Nationwide Broadband Survey on Malcolm Turnbull's own site, along with statements like:
We want every Australian to have faster broadband sooner and more affordably.
The survey is available at www.fasterbroadband.com.au
10 years later, of course, this link has atrophied, but not without an insulting 403 message:
There has never been a more exciting time to be on my website but you appear to have innovated the wrong URL.
What can I say?
But that's not where it took me; it took me to http://www.malcolmturnbull.com.au/fasterbroadband, for the very good reason that that's what the HTML says:
I've seen that kind of thing before, of course: it's a typical trick of scammers to lure people to a site that differs from the one they think they're going to. I think we can safely assume that this was not Turnbull's intention, especially since http://www.fasterbroadband.com.au/ really does redirect to http://www.malcolmturnbull.com.au/fasterbroadband. But this kind of inaccuracy always puts me on my guard. Why did they do it?
The survey itself is long out of date, so they deleted it and just left an error message where it should have been. Why? Incompetence? The dead link suggests yes, but maybe there's a more obvious reason: the results of the survey weren't what the Coalition wanted. Still, one of the stated aims of a government is to do what the majority of the population need. Why not another survey? Wrote Turnbull a mail message suggesting that he do that. It'll be interesting to see what reply I get, if any.
|
Firefox pain, next installment
|
Topic: technology | Link here |
On with my attempts to build firefox today. Repeated the build that died yesterday with SIGKILL and—it happened again! It wasn't until some time later that Jürgen Lock reminded me that this is what happens when you run out of swap.
In parallel, installed the image on my real hardware, with 3 GB of memory and 3 GB of swap. This time it didn't die with SIGKILL. Instead I got:
What's that? I don't want to know. Started the build yet again with gcc. Another day over and done with.
On the user front, I still didn't have Adobe flash with firefox. Why not? The port was installed, but it didn't perform the final step to include it in the firefox configuration:
=== grog@teevee (/dev/pts/5) ~ 3 -> nspluginwrapper -v -a -i
Interestingly, that also enabled flash for the other browsers I have installed (chrom* and Opera).
|
Finally some garden work
|
Topic: gardening | Link here |
I've done almost nothing in the garden for some time. But now the vine on the verandah has lost all its leaves, and it's high time to remove them, along with some plants that need protection in the winter. Got halfway round to sweeping up the leaves and preparing the pots, but it rained all afternoon, so I decided to put it off until the rain stopped.
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 |