|
| Greg's diary |
| February 2013 |
|
|
||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
|
|
|
|||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| Friday, 1 February 2013 | Dereel → Ballarat → Dereel | Images for 1 February 2013 |
| Top of page | ||
| next day | ||
| last day |
|
Still more upgrade fun
|
Topic: technology | Link here |
Into the office this morning to find a surprising message:
That was the X server 1. I couldn't find a core dump, but I did find an emacs.core dated only shortly before I came in, so it seems to have happened then. Why? Unless it happens again, I'll never know.
|
Pricing strangenesses
|
Topic: general | Link here |
Into town in the afternoon for a number of things. Most important was a new mouse, which I got at OfficeWorks. Pricing is just plain bizarre: the cheapest mouse cost under $6, and apart from the missing buttons it would have done the job. The most expensive I saw cost $150. I also thought a second (horizontal) scroll wheel, like I had on my old mouse, would be a good idea, but they didn't have any mice at all with a second wheel. After much comparison, ended up with a Logitech m705, which at least has 5 buttons, even if one is well hidden. And it cost me $49!
Not the only strange price. I also needed some hose fittings—I had enough hose, just needed to be cut to length. But a pair of hose fittings cost $6.50. For $5.88 I bought a special offer: 5 m of hose with fittings on each end, and also an el-cheapo sprayer.
Then off to look for cutting paste for cleaning the copper relief I bought last April. Found some at Autobarn. But it cost $30! That's three times the price of the relief. I was able to resist that. I suppose I should see if CJ has some to spare.
Finally to get our toaster oven repaired. We bought it two years ago, and it was very expensive. All the more annoying that it should die with some electronic error (“E 05” in the display). To John Thomas, where they wanted $16 just for an estimate. That suggests that the real repair price would be much higher. I wonder what's really wrong with it; it seems to be something related to the temperature, so it could be as simple as a misplaced heat sensor.
|
Of mice and modems
|
Topic: technology | Link here |
Back home, set up the mouse, which was relatively simple. It has 5 buttons and one scroll wheel, or, as the probe put it:
There are two side buttons for the right thumb (bad luck if you're left-handed), which xev reports as buttons 8 and 9. But when I tried to assign button 2 to button 8, it didn't work. A little investigation shows that the driver thinks they're buttons 4 and 5. I wonder where that confusion came from.
Another reason for moving eureka to the other side of the desk was so that I could put the 3G modem in it, thus eliminating nerd-gw, the Internet gateway machine. That didn't work before because the antenna cable wasn't long enough, and even with a USB extension cable it would have been too far. But it didn't work. The device was recognized, of course, but no devices (notably /dev/cuaU0*) came up. Checked what I did last time, but couldn't find any useful information. After much comparison and some help from Jürgen Lock, loaded the u3g driver, which was in the kernel on nerd-gw, but not on eureka. That gave me some devices, but not the all-important /dev/cuaU0*. Looking at the man pages, it seemed that I needed ucom(4). But that wasn't easy:
=== root@eureka (/dev/pts/9) /etc/ppp 74 -> kldload ucom
=== root@eureka (/dev/pts/9) /etc/ppp 75 -> kldunload ucom
What's that? kldstat -v confirmed that it wasn't there, but why can't I load it? Surprisingly, exactly the same thing happened on nerd-gw. So where are the cuaU* devices coming from?
One other thing that Jürgen suggested proved to be the case: the new display card includes sound hardware:
=== grog@eureka (/dev/pts/3) ~ 1 -> pciconf -lv
So for some reason the system assigned the first device to the graphics card. Time to get out the docco and understand how the audio subsystem works.
| Saturday, 2 February 2013 | Dereel | Images for 2 February 2013 |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
|
More config refinements
|
Topic: technology | Link here |
So my new setup is up and running relatively well. X server 1 is effectively what I want server 0 to be, but for photo processing it makes sense to have one display spread over 2 screens. OK, that's simple enough: that's what TwinView is for, and the config files were conveniently generated with this line for each device:
So I tried setting TwinView to 1. No difference, apart from the fact that the order of the screens changed yet again, and that two monitors came up in low resolution. Looking at the log files, discovered the line:
Why? It didn't bother to say, not even with very verbose logging. Possibly on a three head card I need TripleView or some such, but since the third monitor is not contiguous with the other two, and it also has a different resolution, that didn't seem to be worth trying. I did try the Xinerama settings, and that worked, but what do you want to do with a single “screen” that has heights varying between 1080 and 1440 pixels? So for the time being I'll give up on that one.
The other issue was the resolution. That's nVidia's fault too. It added lines like this in the Screen section:
I've never heard of metamodes before, but it seems that the string “1920x1080 +0+0” didn't match the “1920x1080” of the EDID, so it got ignored. For some reason the first one that matched was 1280x1024, so that's what I got. Removing the line fixed the problem.
Looking at the audio setup showed me no less than 5 devices. Reading the handbook suggests that it hasn't been updated since the days where sound was an optional extra card. But it did help:
Another issue is that modern graphics cards often come with their own sound driver, for use with HDMI and similar. This sound device will sometimes be enumerated before the actual soundcard and the soundcard will subsequently not be used as the default playback device.
Looking at my dmesg output, I had:
That makes it fairly clear that the output I have (rear analogue) is device 3, and sure enough, this fixed it:
|
Eliminating GPRS
|
Topic: technology | Link here |
I've been grumbling for some time that my wireless Internet connection drops back to GPRS from time to time, and is hard to get back to HSPA. I read the manual looking for that capability, without success. And then Andy Snow came up on IRC and said it could be done. It turned out he was using a different kind of modem, but he found the right commands for my Huawei 1762 here and here:
The only important parameter is the first, but they all need to be specified:
14: Only 3G.
2: Search 3G before 2G.
3FFFFFFF: Search all applicable bands.
2: Don't change roaming status.
4: Don't change domain.
But after that I didn't have any more trouble, so I can't try it. Has Optus finally done something about my recent problems?
|
ABC: No iview for you
|
Topic: multimedia, technology, opinion | Link here |
ABC TV reception continues to be abysmal, but there's a new series on TV, The Doctor Blake Mysteries, taking place in Ballarat, so I thought it might be worth downloading yesterday's first episode via ABC's iview service. Surprise, surprise:
Warning: Due to copyright reasons this video program[sic] is available for download by people located in Australia only. If you are not located in Australia, you are not authorised to view this video.
This isn't a warning, it's an error message. What it really means is “You are not located in Australia, so you can't watch the video”. And nobody is allowed to download it; you can only view it with a streaming client. Further discussion shows that there is a downloader available, and the author is in trouble with the ABC as a result. Of course, it's now out of his hands and out of the country, so there's not very much the ABC can do. None of which helps me, since I'm clearly not in Australia.
| Sunday, 3 February 2013 | Dereel | Images for 3 February 2013 |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
|
Shutting down another machine
|
Topic: technology | Link here |
Most of my reorganization is now done. dereel has been demoted to a virtual machine, and my new graphics configuration finally works, though the last attempt at a change to the X configuration file didn't work: once again the position of the monitors change. Here's the layout in the config file:
You don't need to understand much of the configuration file syntax to understand what that means. But I made only these changes to the configuration file:
The intention here was to allow me to switch to lower than full resolution, which is useful for some web pages that assume such a low resolution that they're illegible on normal displays. That didn't work, presumably because of these strange “metamodes”. But presumably the two metamode lines were what told the drivers what the real layout was, not the layout section. So for the time being I'm doing without lower resolutions.
So all that was left was to move the wireless modem to eureka. My last attempt had failed because the devices weren't recognized. In the meantime I built a new kernel with the u3g driver and rebooted—not the easiest thing when you need the help of a virtual machine running on the same machine—and how about that, things Just Worked.
Next the “no fallback to GPRS” feature. Not an immediate success:
Spent some time looking at the documentation and what the modem said. They're not quite the same.
And yes, that's the way it's formatted in the PDF documentation. But what the modem said was:
Spent some time pondering that until Andy Snow suggested that the chat script was interpreting ^S as a control sequence. And looking at chat(8), that's what it does:
But how do you un-escape it? That's not in the man page. \ would seem to be the obvious choice, but that didn't work. UTSL didn't get me very far either, beyond discovering some unpleasant-looking code. So there's nothing like trying it out:
=== grog@eureka (/dev/pts/14) ~ 7 -> chat FOO AT^SYSINFO OK|hexdump -C
=== grog@eureka (/dev/pts/14) ~ 8 -> chat FOO AT\^SYSINFO OK|hexdump -C
=== grog@eureka (/dev/pts/14) ~ 9 -> chat FOO AT\\^SYSINFO OK|hexdump -C
This requires typing (blind) the FOO and OK before you see anything. But clearly the \\ was the way to go, and sure enough, after that it worked. Now to see if the thing still falls back to GPRS or not. Certainly the connection is no better than before, but I'm using 150 W less power, and it's so much quieter in the office. I hadn't realized what a noise nerd-gw made.
|
Repairing the Mecablitz: next obstacle
|
Topic: photography | Link here |
So now I have a new flash tube for my Mecablitz 58 AF-1, and all I have to do is to put it in. But how do I get the old one out? This article tells me to turn the worm screw at the top until the reflector comes out the back. Here's what that looks like:
How do you turn the worm? You can't get at either end. Yet another question for the forum.
| Monday, 4 February 2013 | Dereel | Images for 4 February 2013 |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
|
New system, old bugs
|
Topic: technology | Link here |
Things are up and running happily on my reshuffled hardware, but the past isn't completely gone. Today I had Yet Another case of the X hang bug that should have been fixed months ago. And again it happened while I was doing other things with the USB subsystem, this time reading in photos from SD cards. As if to confirm my suspicion that all is not well with FreeBSD USB subsystem, the second SD card didn't register. The reader I have reports multiple devices, normally something like this (the first card):
In this case, only da5 was actually present. After reading the card, I replaced it with the other one. This time I got:
And that was all. Where was da5? I removed the reader and intended to replace it, but X hung. Looking at the log file, I found:
Had I not waited long enough? It had been a minute since I inserted the reader, and 30 seconds since any activity took place. But no, nothing to be done, and neither keyboard nor mouse (both USB) worked any more. Nothing to do but reboot. But of course, that didn't help either:
|
|||||||||||||||||||||||
And there it hung. Big red button time. Unfortunately I didn't have dumps enabled—something to add to the next kernel. Maybe I should use another machine to read my SD cards.
And in the evening, another thing came to bite me. I've retired not only the physical machine dereel, but also its /home file system. I was just about to wipe the partition when Yvonne came to me and told me that the expenses database was missing all data since 29 June 2012, the last entries she made before I migrated to eureka. Somehow in the course of the migration I lost the correct database. I ran these queries on dereel, because newer versions of PHP are too leet to run my old scripts. And the database, nominally /var/db/mysql/household/, was really on /home/var/db/mysql/household/—the file system I had removed and was about to erase.
A bit of panic looking for the correct tables, which weren't easy to find. I ended up in a maze of twisty little symlinks, all the same, and it took me some time to get past the ones pointing at eureka and finally find the tables. But it certainly confirmed the correctness of my intention to back up the file system before erasing it.
|
Mecablitz: one more step
|
Topic: photography | Link here |
An answer to my questions on how to proceed with dismantling the Mecablitz 58 AF-1: turn the worm with your finger. I obviously don't have the same fingers as the responder: I couldn't get at it with my fingers. Tried with other objects, finally using a small screwdriver, and how about that, it really did work. But somehow this all sounds like an “you can do it this way too”; surely Metz has some more professional way of doing it.
So finally I had the reflector carrier free—almost. Here's the view from the back and from the right-hand side of the housing:
|
|||||||||||||||||||||||
|
|||||||||||||||||||||||
The wire looped on the right-hand side goes to the little circuit board on the back, which is fixed to the carrier. And there seem to be some catches that I couldn't see, either on the photo or with the naked eye. Another question on the forum, this time getting an answer questioning my ability to perform the supposedly simple tasks. All well and good, but one wrong step and something could be damaged beyond repair. Yes, of course the wire on the side has to be unhooked, but there are no clips, just more wires to unhook. Tried that, and sure enough, I had the carrier out. But by that time it was evening, so it's a case for another day.
One remaining question: how to discharge the capacitor? It's inside the main housing, and I don't really feel like taking that apart. But the voltage should be present on the head as well, and I couldn't measure anything there. Can I take it as an indication that the capacitor is already fully discharged? If not, at least it seems that there's no dangerous voltage anywhere in the head, so there should be no problem.
| Tuesday, 5 February 2013 | Dereel | Images for 5 February 2013 |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
|
Tidying up the office
|
Topic: general | Link here |
I've more or less tidied up parts of my office now, but I have had lots of papers lying around in Yvonne's office for months now, waiting to be filed. I dreaded doing something about it, but Yvonne has forced my hand by getting a new desk from Chris Bahlo, so I need to move all the papers. Today Yvonne and Chris went to Olivaylle to visit Conchita de Moya, who is passing through, so I had time to do it while she was gone. And how about that, it wasn't even that difficult.
|
Cleaning drains
|
Topic: general | Link here |
CJ Ellis along this morning, mainly to repair some fence damage caused by kangaroos, but he also brought along a cable to clear a blockage in one of the drains:
|
|||||||||||||||||||||||
We established that the blockage was right at the beginning, so rather than push it down the drain, decided to suck it out with a pump. After the usual messing around trying to get the right fittings, finally had it in position, turned it on and... blew a circuit breaker. The pump had seized up. CJ took it back home with him to get the thing moving again.
|
Network problems: worse than ever
|
Topic: technology, opinion | Link here |
My network problems continue to be catastrophic. Called up Internode Support and asked what was happening, and got a call back from James, clearly somebody who knows what he's talking about. It seems they've repeatedly supplied the information to Optus, who identified the rogue cell as one of the cells on the Rokewood tower, but they say that it's functioning normally. No explanation of the poor response times, which have now reached times reminiscent of RFC 1149:
|
|
||||||||
|
|
||||||||
That's a worst-case response time of 234.833 seconds, nearly 4 minutes! And I'm not the only person having the problem. Why does Optus deny there's a fault? But then, maybe it's part of their strategy. You have to wonder when you see images like this one, found on their home page:
|
||||||||||||||||||||||
Still, I now know that the rogue cell is in Rokewood, a bearing of about 200°, and that in the past I've had good connections with the tower in Linton, a bearing of about 310°, and also with the Willowvale tower, about 230°. My antenna is pointing about 230°, so maybe moving it further west might stop the Rokewood tower annoying me.
But why all this cell-hopping in the first place? Is it related to the temperature? It happens mainly when it's warm, and in the afternoon. I recall having the same problem at this time of year two years ago, and James reminded me that I had a similar problem this time last year. But last year it wasn't nearly as bad, and last year Optus stated at the time that the problems were due to congestion. Roll on the Radiation Tower.
|
VCAT sets a date
|
Topic: technology, general | Link here |
On the topic of the Radiation Tower, VCAT has finally set a date for the hearing: 24 April 2013, 10:00 to 13:00. I hope that the short duration of the hearing will mean that the result is a foregone conclusion. Then we could finally be rid of this horribly flaky wireless connection.
|
Recipes: don't trust Groggy
|
Topic: food and drink, opinion | Link here |
Cooked lasagne al forno today, for the first time in quite a while. My recipe isn't that old, but the quantities looked like they were plucked out of thin air: 150 g ragù bolognese and 100 g salsa di pomodori. In the end I used 1 kg ragù and 300 g salsa. Where did I get those quantities? I grumble about incorrect quantities when I find them in other recipes; it's rather disappointing to find them in my own recipes.
| Wednesday, 6 February 2013 | Dereel | |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
|
Another quiet day
|
Topic: general | Link here |
Some days I seem to do lots of things, and on others there's almost nothing to report. Today was one of the latter. For no apparent reason I didn't do anything.
|
Network improvement?
|
Topic: technology, opinion | Link here |
Yesterday's network performance was by far the worst I've experienced since getting the UMTS link. I had multiple timeouts, not surprising with RTTs of up to 4 minutes. My long discussion with James at Internode support was interesting, but basically ended with “We can't get Optus to do anything about it”. So I wasn't expecting anything to change in a hurry, especially given my hypothesis that the problem might be related to the hot weather: today was the hottest day in the last couple of weeks, with a top temperature of 39°.
And of course I had my problems. One disconnect, a firmware reset at 9:40. And from then on things seemed just to work. Why? Is this one of these cases where they do fix something without telling me? In either case, let's just hope that it stays this way until we get the Radiation Tower.
| Thursday, 7 February 2013 | Dereel | |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
|
Another slow day
|
Topic: general, animals | Link here |
Once again I didn't do much today, not even take another look at the Mecablitz 58 AF-1. which I should be able to get up and running without too much difficulty. I blame the continuing hot weather. So, it seems, do the cats—Piccola is normally nowhere to be seen during the day, but today she and Lilac spent most of their time in their basket in the laundry.
|
Network improvement: coincidence
|
Topic: technology | Link here |
Yesterday's improvement in the network quality didn't last. Today wasn't quite as bad as two days ago, but it's still unacceptable. Still, yesterday's experience suggests that I can rule out heat as a contributing factor. Today I ended up in a situation where the modem was just losing and reconnecting to the same cell all the time:
By the time I noticed, this had been going on for 30 minutes, and it continued after I disconnected and reconnected the modem. The modem is now in a powered hub, and I just disconnected the cable, not the power. It wasn't until I power cycled the modem that it stopped, and after that communications were “normal”.
So: is it the modem? That doesn't explain why it's happening to other people, nor my throughput issues with cell ID 8fc8e4a. My guess is that the modem firmware became confused by something that had happened and lost the plot. Normally a cell gets lost when the RSSI drops to 0 or below; here the (5) is the RSSI, so there's no issue there.
| Friday, 8 February 2013 | Dereel | Images for 8 February 2013 |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
|
Yet another lazy day
|
Topic: general, photography | Link here |
Somehow I'm not doing anything any more. Well, nothing in the garden, and nothing interesting. I have managed to get the papers in Yvonne's office tidied away, though ever more show up. And I did some more looking at the Mecablitz 58 AF-1. How do I connect the trigger? That part of the tube is completely missing—I couldn't even find most of the glass, nor the trigger. The new tube I have has a wire, but I can't see a connection on the head, just a loop which seems to contact the tube directly. I suppose I should finally bite the bullet.
|
Mystery flower
|
Topic: gardening | Link here |
Where does this lily come from? It's in the north garden to the west of the arch, and there are two more to the south of the east verandah:
|
|||||||||||||||||||||||
I can't find any photos of it from this time last year, nor a mention in my diary, and it's pink, a colour that Yvonne doesn't like. In addition, it only flowers for a couple of weeks—here's one south of the verandah just starting to flower, only 11 days ago, and already they're starting to fade.
|
|||||||||||||||||||||||
| Saturday, 9 February 2013 | Dereel | Images for 9 February 2013 |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
|
No wind any more
|
Topic: gardening, technology | Link here |
My weather station showed no wind at all for nearly 24 hours. Looking outside shows why:
|
|||||||||||||||||||||||
|
|||||||||||||||||||||||
|
Darah with swollen eye
|
Topic: animals | Link here |
Yvonne in this evening to say that Darah had some problem with her left eye. Fortunately Chris Bahlo was there at the time, and she promptly investigated it and poured water over it to cool it down:
|
|||||||||||||||||||||||
|
|||||||||||||||||||||||
It's not clear what caused it, but the eyeball appears not to be affected. I saw (and killed) my first March fly of the season this morning, so maybe one bit her in an unfortunate position.
|
Cooking quantities and times
|
Topic: food and drink, opinion | Link here |
Baked another batch of bread rolls today. Over the course of time I've been trying to refine the quantities. The last batch wasn't bad—67% water:flour ratio, 2.4% yeast to flour ratio. But can I use less yeast and more water? Of course I can, and that's what I did today: 68% water and 2.0% yeast.
What a catastrophe! It was so much worse than the other times that I'm not at all sure that it was that difference that caused it. In any case, the dough was really sticky, where previous batches were relatively easy to handle. And the bread didn't rise nearly as much, and I ended up with a lot of flat buns that looked more like Nan.
And then in the evening Chris Bahlo came over, for the first time in over a month. Cooked a 760 g Girello roast, which according to my cooking time page should have taken about 38 minutes to reach 53°, but in fact it took 45 minutes. On the other hand, it appeared overcooked, so maybe I have issues with thermometer latency.
|
Yet Another X Hang
|
Topic: technology | Link here |
I've got to admit it: since building my last kernel, the X hang problem is back. Many times when I connect cameras or disks to the USB bus, I end up with this cursor jumping problem. It happened twice today. How can I fix it?
| Sunday, 10 February 2013 | Dereel | Images for 10 February 2013 |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
|
Finally: aligning images
|
Topic: photography, technology | Link here |
In March 2011 I tried in vain to align 3 images with Hugin so that I could compare them with mouseover image manipulation (run the mouse cursor over the images below to see what this means). The main problem was that one of the images was taken at a different focal length, so they didn't line up:
|
|||||||||||||||||||||||
|
|||||||||||||||||||||||
I asked on the Hugin mailing list, but didn't get the answers I wanted. Today I finally worked it out and wrote a description of how to do it. At least the results are better now:
|
|||||||||||||||||||||||
|
|||||||||||||||||||||||
While on the subject, also revisited the Hugin port, which is still greatly out of date. The biggest issue is tclap (or maybe TCLAP), a 5600 line replacement for getopt(), for which it seems I'm going to have to create a new port. Apart from that, the Hugin build went well.
|
Wrong season for gardening
|
Topic: gardening, opinion | Link here |
I've been neglecting the garden, and it shows. In particular, the vine on the east verandah is growing in all directions, mainly 4 m above ground level. Spent some time today with a tree trimmer trying to cut it back. Not a success. Apart from the flies, the trimmer jammed Every Single Time, and in the process managed to skin my knuckles several times. And for some reason it's really difficult to cut the things anyway. Maybe I should just let them grow and cut them all back in the winter.
|
Traceroute to Star Wars
|
Topic: technology | Link here |
Somebody pointed me at this today:
=== grog@w3 (/dev/ttyp1) ~ 1 -> traceroute 216.81.59.173
| Monday, 11 February 2013 | Dereel | Images for 11 February 2013 |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
|
Rearranging Yvonne's office
|
Topic: general | Link here |
Yvonne has decided to change the furniture in her office. A few days ago she bought a filing cabinet, and recently Chris Bahlo has given her a desk that she no longer needs. So today was the big day to replace the furniture. It's not easy, because the office is tiny and well packed:
|
|||||||||||||||||||||||
In particular, the bookshelf at the left narrowed the entrance to the point that we couldn't get the first table out. It had to be removed, not made any easier by the fact that we could no longer move the table back in either:
|
|||||||||||||||||||||||
That wasn't the only thing. I couldn't get that same table out onto the verandah without first removing the door handle:
|
|||||||||||||||||||||||
Finally we got all the stuff out:
|
|||||||||||||||||||||||
We got the new stuff in relatively easily, but it took all afternoon, and we weren't finished. The “after” photo will come tomorrow.
|
Documentation: The newer, the better
|
Topic: photography, technology, opinion | Link here |
After fixing the Hugin alignment page yesterday, I sent out a message to the Hugin mailing list asking for review. No replies apart from a thank-you, but when I went to look, I discovered that a “Thomas” with no further identification had changed it—to reflect the current development version! So now, again, it doesn't work. Why do people do that? Admittedly, he did have one great simplification, but in general this just confuses people.
|
More ports
|
Topic: technology | Link here |
Finished my port of tclap today, and finally committed it. Also updated the Hugin port to the latest release—by no means too early—but that's Vadim Dimov's baby, so I'll have to send it to him. At least things are looking tidier now.
|
Mail falling into black hole
|
Topic: technology | Link here |
I was expecting a lot of mail today, but somehow nothing much arrived—so little that I started to check. First in /var/log/maillog:
That's not very much. Why wasn't postfix logging? Tried sending myself a message locally. No log messages. Nothing in my inbox. Nothing in ~/Mail/backup, where I store everything that arrives, even spam.
Panic time. Was this another problem with procmail? Took a look in the procmail log. Everything looked normal: mail arrived, processed, stored in /var/mail/grog. So why wasn't it there? Looking at the file with Emacs, it was there!
Stopped my mutt and restarted it. The messages were there! Somehow mutt had hung itself when re-reading the file. Part of the problem identified, but why no log messages? newsyslog didn't help, but restarting syslogd did.
It took me a while to understand what was really going on with logging: I had both dereel and eureka logging to the same directory. It seems that their settings in /etc/newsyslog.conf were different, so dereel rolled /var/log/maillog over every midnight, while eureka was happy to continue logging to the same file, which dereel had just deleted. Moral of the tale: don't share log directories across machines. But I still have no idea why two different instances of mutt didn't find the newly delivered messages in two different places.
| Tuesday, 12 February 2013 | Dereel | Images for 12 February 2013 |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
|
Completing the office reorganization
|
Topic: general | Link here |
Yvonne continued with her office this morning, and I reinstalled the computer—it's amazing how many useless cables collect behind a computer. The result certainly looks better:
|
|||||||||||||||||||||||
|
|||||||||||||||||||||||
And yes, the field of view is roughly the same. The bookshelf on the left has now been moved to the back wall.
|
Documenting Hugin, continued
|
Topic: photography, technology, opinion | Link here |
A reply to my message in the Hugin mailing list this morning, from Thomas Modes, explaining a few things, but still a little astonishing:
The wiki page are also used as help files and shipped with Hugin. If the new version would be shipped without the updated pages, it would also confusing. And we can't update all pages at once. So we started with the update of the wiki pages. The first beta release will follow in the next weeks. So there is a short time, where wiki and release are out of date. .... So we take this short time of not sync wiki and release as acceptable.
It's difficult to know what to say to that attitude. I wonder how many users consider it acceptable.
|
Aligning panoramas
|
Topic: photography, technology, opinion | Link here |
The photos of Yvonne's office aren't ideal: the “before” and “after” images don't line up. And the instructions in my alignment page don't help. If the focal length of the images are close to the same, the “view” alignment doesn't work—this is the background to one of the things that Thomas Modes changed, but it still needs clarification. This is going to keep me busy for a while.
In fact, the photos have since been replaced with better-aligned ones. See below for a discussion
|
Documentary quote of the day
|
Topic: multimedia, opinion | Link here |
Watching a documentary about Apollo 13 on TV today. Sometimes I wonder why I watch these things. The cause of the accident was apparently a tank filled with “highly flammable liquid oxygen”. By contrast, giving the liftoff time as “13:13 pm” seems harmless, even though it was really 2:13 pm local time. Clearly that makes it easy to find a 13:13 somewhere in the world.
| Wednesday, 13 February 2013 | Dereel → Ballarat → Dereel | Images for 13 February 2013 |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
|
Aligning panoramas, continued
|
Topic: photography, technology | Link here |
More playing around with Hugin today trying to align the “before” and “after” images of Yvonne's office. Made some progress, but clearly it's getting to be time to investigate what all the different optimizations really do at a technical level. The following are the first cut and the second cut for comparison. Run the mouse cursor over the images to see the “after” image.
|
|||||||||||||||||||||||
|
|||||||||||||||||||||||
The real issue appears to be to find enough control points in the right places for nona to remap them correctly. Here the picture on the right and the curtain rails are OK, but on the left I couldn't find enough control points to completely align the two. It's also worth noting how much smaller the maximum crop has become. I can see months of learning ahead.
|
To Ballarat again
|
Topic: general | Link here |
Into town again today to the dentist: a small piece of an inlay had broken off one of my lower front teeth. Surprise, surprise: it wasn't an inlay at all, but a piece of scale. That's particularly surprising because I had my teeth de-scaled only 3 months ago. Still, considering that it's been years since I was last at the dentist's, it's nice to know that there was nothing for him to do.
Nearly 4 years ago, for some mad reason, I picked up some encyclopaedias on Freecycle. I've almost never used them, and they're just in the way. Finally somebody asked for encyclopaedias on Freecycle, so I brought them into town and gave them to her—much more than she had expected, of course, but it seems her son will be suitably happy to have them. That's much more satisfying than throwing them away.
|
Shutdown stupidity
|
Topic: technology, opinion | Link here |
Watching TV in the evening, and then for some reason wanted to look at something on eureka. Did that, then shut the machine down. teevee? No, eureka! What a pain. And the more I reboot eureka, the more idiotic this historical reliance on dereel, now a virtual machine, becomes. It took me over half an hour to bring the machine up again. There's no reason why the executables and libraries need to be on an NFS mount from dereel—tomorrow I'll move them to a local file system.
| Thursday, 14 February 2013 | Dereel | |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
|
Carola Schlanhof time
|
Topic: general | Link here |
It's February in an odd-numbered year: time for Carola Schlanhof to show up, and she did so today. Her previous visits started on 5 February 2009 and 4 February 2011, so she's a little late this year. She and Yvonne did horsey things all day, though Carola had arrived at 02:15 and had little sleep.
|
The new enblend
|
Topic: photography, technology, opinion | Link here |
My updated enblend 4.1 port is now ready for committing. I had done some testing with it on stable-amd64, but it seemed reasonable to try it on eureka with some more complicated panoramas. A good thing I did:
|
|||||||||||||||||||||||
|
|||||||||||||||||||||||
Spent hours investigating, but it seems that this is a bug in enblend 4.1 which occurs with 360° panoramas. I wasn't able to stitch a single 360° panorama correctly. Entered a bug report and wrote a description page. I wonder if it occurs on other systems than FreeBSD. I tried on Microsoft and Apple, but couldn't get the correct version: there isn't one yet for Apple, apparently, and the binaries available for Microsoft are 32 bit (only) enblend and 64 bit (only) Hugin. It wasn't until later that I realized that I didn't need Hugin to do the test, but by that time I had wasted the whole afternoon on the problem, so I couldn't be bothered.
| Friday, 15 February 2013 | Dereel → Ballarat → Dereel | Images for 15 February 2013 |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
|
More stuff in town
|
Topic: general | Link here |
Into Ballarat again today for the results of last week's blood test and to talk to Peter O'Connell about investments—I have what I thought was far too much in cash and not enough in investments, especially since the stock market indexes seem to be on the rise again.
Timing wasn't perfect: I had the doctor (Majid Najafi Zeini) at 10:30 and Peter at 11:00, and of course Majid had to be running behind schedule. I was just about to leave and reschedule when he finally appeared, and I didn't get out until 11:02. How I hate being late! I don't suppose it worried anybody else.
Also to Bunnings to buy a single nipple for my pond irrigation. These things take far longer than they should.
|
Enblend bug confirmed
|
Topic: photography, technology | Link here |
Reply to my message about the enblend bug today. It appears to be a known problem, and the respondent (Christoph Spiel?) even provided a link to a patch. This all begs the question why it hasn't been made more public. I'd consider that one a show-stopper. It seems to have less to do with 360° panoramas than with images with more than one seam line. I suppose I should try the development version.
|
Too hot for flowers
|
Topic: gardening, photography | Link here |
Today was the monthly Flower photo day. There wasn't much to see: the hot, dry weather has seen to that, and things look much worse than this time last year. Here some comparisons: the ever-flowering Alyogyne huegelii has usually recovered from the summer heat by this time of year, but I could only find two flowers, compared to lots of them this time last year (first two images):
|
|||||||||||||||||||||||
|
|||||||||||||||||||||||
|
|||||||||||||||||||||||
And the roses that tried to come again have stopped flowering again. On the other hand, there are a few pleasant surprises:
|
|||||||||||||||||||||||
|
|||||||||||||||||||||||
|
|||||||||||||||||||||||
The first is the Clematis I mutilated a couple of months ago. It's not still blooming: that was the spring flush, which finished just under a month ago, and what we have now appears to be the autumn flush. And the Gladiolus isn't just late: it's one of the corms I planted in spring, and it shouldn't really be flowering at all this year. And the third photo is of last year's Chile poblano, which is bearing what look like quite nice fruit, long before this year's plants have much to show for themselves.
The Hebes are also recovering from their clogged drippers:
|
|||||||||||||||||||||||
|
|||||||||||||||||||||||
The fact that the flowers are so small suggests a slow recovery.
| Saturday, 16 February 2013 | Dereel | Images for 16 February 2013 |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
|
More photo processing strangenesses
|
Topic: photography, technology, opinion | Link here |
House photo day again today. Despite improvements in my technique, it took all day, at least partially because I had to start late. But once again I've run into some strangenesses in processing that may be due to the newer versions of the software. Here an example done with the old (2011.4.0) and new (2012.1.0) versions of Hugin:
|
|||||||||||||||||||||||
|
|||||||||||||||||||||||
As in most cases, in this image I violated the rule that you should take all component images with the same exposure, but not by very much: the ends were exposed with 12.6 EV, and the middle with 14.3 EV. In other successful panoramas I have had up to 6 EV difference, so that shouldn't be the problem. Unfortunately, this panorama also needed some attention to the control points, so it's difficult to know exactly where the problem arose.
|
Fisheye lens projections
|
Topic: photography, opinion | Link here |
More discussion about fisheye lenses on the German Olympus forum recently, about the lenses I was looking at a couple of months ago. The Samyang lens is sold in Europe under the name Walimex, so the lenses they're talking about are the same ones I was looking at in November last year. At the time I noted that focal length and field of view seemed to be unrelated. In fact, further research shows that it's all a matter of projection. The Samyang approximates to a stereographic projection, as described here. Most fisheye lenses have an azimuthal equidistant projection, which is what allows them to have angles in excess of 180°. There's more information here, and I've also found this page, but not read it yet.
| Sunday, 17 February 2013 | Dereel | Images for 17 February 2013 |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
|
Exploring fisheye projections
|
Topic: photography | Link here |
The recognition that the Samyang fisheye lens has a different projection from most fisheyes is interesting, but what does it mean? I have yet to see the same view taken by both kinds of lens.
But there's a solution: Hugin. It can produce panoramas with both stereographic projection (like the Samyang) and azimuthal equidistant projection (“fisheye”) like most fisheye lenses. So did a bit of playing around with that.
The results were surprising: it was really difficult to get the same crop for each kind of projection. Here are the results, first stereographic, then azimuthal equidistant:
|
|||||||||||||||||||||||
|
|||||||||||||||||||||||
What does this say, beyond the possibility that the choice of source image is possibly not the best? I don't know. But the edges look less extreme in the stereographic image. It's not clear why the aspect ratio has changed: I went to some trouble to ensure that the edges of the crop were in the same place. I wonder how accurate Hugin's projections are.
|
Enblend fixed
|
Topic: photography, technology | Link here |
How about that, a new version of enblend, 4.1.1. No explanation, but the bug is gone, so presumably that's a result of my report—not such a bad response time after all. Updated the port and asked Jürgen Lock to commit it for me. He went to a lot more trouble, in the process discovering a number of issues not directly related to the upgrade, notably documentation. But finally it's there. Now to move on to the Hugin port.
|
Austrian and German
|
Topic: language, general, opinion | Link here |
Carola Schlanhof is one of the most Austrian people I know. We've known her for years, but we still have great difficulty understanding her German. So it was nice of her to bring a dictionary with her: Austrian-German.
But it seems that things work the other way round too. She doesn't understand German very well. This evening we put on a German film for her, Zweiohrküken—but she didn't understand it! So I dragged out an old disk with some episodes of Stockinger, which at least she understood. It's amazing that a professional in Austria should have difficulty with “German” German.
|
Another power failure
|
Topic: general | Link here |
Another power failure this evening, at 22:58. It was unusual in its duration: 5 seconds, too long for a recloser event, and too short for anything else.
| Monday, 18 February 2013 | Dereel | Images for 18 February 2013 |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
|
Fun with travel
|
Topic: general | Link here |
Spent a considerable amount of time today helping Carola with her travel plans. She says that she is a Qantas Frequent Flyer, but she doesn't have a card, and the only information she has were emails that disappeared when her (not backed up) computer “crashed”. Qantas doesn't know either. How could that happen?
So she tried paying for her upcoming flight to Devonport by credit card. That didn't work either! After I finally persuaded them to ring MasterCard, they discovered that they were performing maintenance work (round midnight on Sunday in Austria), and that as a result she couldn't use the card. I thought that sort of thing went out decades ago. Come “morning” all was well.
|
Welcome Dominguero
|
Topic: animals | Link here |
Yet Another Horse arrived today: Dominguero, a horse we sold to Natalie Eggenberger in Byron Bay some years ago:
|
|||||||||||||||||||||||
|
|||||||||||||||||||||||
He's suffering from Queensland itch, so she sent him down to recover.
| Tuesday, 19 February 2013 | Dereel → Shelford → Geelong → Apollo Bay → Maits Rest → Cape Otway → Triplet Falls → Dereel | Images for 19 February 2013 |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
|
More Great Ocean Road
|
Topic: general | Link here |
Off to look at various eastern parts of the Great Ocean Road today. Carola told us recently that years ago she was attracted to Geelong, though she had never been there, so decided to go there and take a quick look.
On the way is Shelford, where Jeanette Lees lives. We visited her in September last year when we were still looking for a Borzoi. As it happened, the litter we saw was sired by Zhivago, and since we were taking him with us, we dropped in to say hello. All the pups we saw there are still there, and this may be the first time that Zhivago had seen them:
|
|||||||||||||||||||||||
In particular, “Spider” was still there. He is a particularly promising dog who had been “sold” when we were last there, but whose purchaser (in Spain!) had backed down because of the transport costs. He looks amazingly like Zhivago, though he still needs a lot of filling out:
|
|||||||||||||||||||||||
|
|||||||||||||||||||||||
|
|||||||||||||||||||||||
Jeanette doesn't want to sell him, but I can see Yvonne trying anyway.
On to Geelong, where we did some shopping for Asian groceries, and then took a walk along the waterfront.
|
|||||||||||||||||||||||
|
|||||||||||||||||||||||
Then on to Apollo Bay. I had forgotten how far it was—96 km—and how slow traffic on the Great Ocean Road is. It took us 2 hours to get there. Stopped for lunch at a place we had been before:
|
|||||||||||||||||||||||
Last time I complained about the cost, but didn't mention that the fish and chips had been particularly good. This time the cost was even considerably higher—$46.60 for the three of us—and it wasn't as good. It was amusing to see how many people were impressed by Zhivago: at least 4 different groups of Chinese and Japanese tourists wanted to take photos of him, mainly with them, and others also stopped to take a look.
Then on to Maits Rest, where I had planned to take some panoramas:
|
|||||||||||||||||||||||
Carola found a hollow tree trunk and asked me to take a photo upwards. Not the easiest, and with a 5 exposure bracket with exposure times between 4 and 60 seconds:
|
|||||||||||||||||||||||
From there to Cape Otway, which I had thought would be interesting. Maybe it is, but they wanted $18.50 per person admission, and there wasn't even any description of what we could expect for that. In general, people didn't seem to be very interested in dealing with us, perhaps because it was already 16:30. So we left again, on the way stopping for an amazing number of Koalas:
|
|||||||||||||||||||||||
|
|||||||||||||||||||||||
|
|||||||||||||||||||||||
From there to Triplet Falls. Unfortunately I had my GPS navigator set to include unmade roads, and it took us along about 20 km of one of the worst unmade roads I've seen in a long while, so bad that Google Maps doesn't even want to plan a route along it. My GPS jumped out of its holder twice, something that I haven't had happen until today. Admittedly it was less than half the distance. Finally we were there, and we set off along the 1.8 km track at 17:34, Carola and Yvonne proclaiming that we'd get there and back much faster than the 1 hour that they claimed it would take.
But we didn't know what was in store. The 1.8 km were along the hypotenuse. There must be a difference of level of at least 50 m (it felt like 100), and most of the way was steps. Even worse for Zhivago was that about 100 to 200 m were wire mesh:
|
|||||||||||||||||||||||
That's usually enough to stop any dog, but Zhivago hardly hesitated—until we reached mesh steps, which were obviously too painful for him, and Yvonne had to lay down her clothes for him to walk on:
|
|||||||||||||||||||||||
And then there was a tree that had fallen down across the path:
|
|||||||||||||||||||||||
|
|||||||||||||||||||||||
Zhivago managed to jump that with no difficulty. Finally it was done, and we made it to the falls. Was it worth it?
|
|||||||||||||||||||||||
That's the top of three platforms, and gives the best view of the waterfall. You've got to pick it out, slightly to the right of centre:
|
|||||||||||||||||||||||
Got back to the car at 18:29—55 minutes in total. OK, there were the photos to take and the dog to help, but it was a lot longer than anybody expected. Didn't get back home until 20:03. 10½ hours and 360 km on the road. Hopefully the photos will be something.
| Wednesday, 20 February 2013 | Dereel | |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
|
A whole day of panorama processing
|
Topic: photography | Link here |
Spent literally the whole day processing yesterdays 429 exposures, and I didn't even get finished. Some of the panoramas don't look at all bad, especially in the 360° animated versions. But they're not perfect: the one of Maits Rest is a little light, the wind has caused ghosting in the HDR images, and despite lots of effort in masking and setting control points, the panorama from the bottom of Triplet Falls still shows obvious discontinuities:
|
|||||||||||||||||||||||
Why? I'm pretty sure that I had everything adjusted correctly, but of course the rail was very close, only about 40 cm from the lens. I can see more work ahead.
The surprising one, though, was a “simple” panorama of the beach at Apollo Bay:
|
|||||||||||||||||||||||
Hugin thought that it was a very good fit, but clearly panomatic, the control point detector, got completely confused. More manual searching for control points ahead.
In fact, this was about the best I could do. See the entry for tomorrow for details.
| Thursday, 21 February 2013 | Dereel | |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
|
Quiet day
|
Topic: general, gardening | Link here |
Today we were supposed to do more driving, to visit Amy and Thea Royal in Yambuk and Nele Kömle in Garvoc, another nearly 400 km. It's mainly horsey stuff, so I bowed out, and Yvonne and Carola went by themselves. I had intended to do some work in the garden, and started, only to discover that the temperatures were over 30° again—a clear indication to me that my laziness is directly related to the temperatures. Instead, I spent much of the day watching TV documentaries, something that I haven't had much chance to do lately.
|
Beach panoramas: impossible
|
Topic: photography, opinion | Link here |
The other thing that I had wanted to do was to finish the last panorama that I took on Tuesday. And then I was Enlightened: it's almost impossible to make a good panorama of a beach with waves. The one I did yesterday was about the best I could do. Getting the waves right would be almost impossible:
|
|||||||||||||||||||||||
| Friday, 22 February 2013 | Dereel | Images for 22 February 2013 |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
|
Printing a PDF
|
Topic: technology, opinion | Link here |
Carola is leaving for Tasmania on Saturday, and she finally has her flight bookings complete. So she asked me to print out the documents for her today.
Nothing difficult about that. They're PDF documents. All I need is to convert them to PostScript and print them. Arguably a print filter should do that for me. There's an issue with acroread, but I have pdf2ps, part of ghostview, so used that instead. But the printout wasn't what I expected:
What does that mean? At an abstract level, that the printer didn't like the PostScript I sent it. OK, I have other methods. Tried xpdf, which didn't work either. Tried an existing PostScript document (from “The Complete FreeBSD”). That worked at any rate, so it wasn't just a dead printer. Peter Jeremy suggested pdftops (clearly not the same as pdf2ps), but that's part of CUPS, and that way madness lies.
In the end I gave up. I have dxo, this new Microsoft box. Surely it can print PDF. And I'm even gradually learning to drive the thing. Sure enough, it recognized the network printer and even offered a driver. Well, the printer is a Brother HL-2700CN, but the driver was for a HL-2700N. Surely there can't be that much difference? But there was; the “test page” just produced a grey smudge at the top. Off to Google for the driver, and found it fairly easily, along with some of the most complicated installation instructions I have found on Microsoft, including requiring you to create a directory, unzip the driver, find the directory and go down a language-specific path to find the setup program. But I found it, installed it, and it worked. Finally I could print out the tickets, after only about ¾ hour.
But that was with Microsoft, and somehow I felt dirty. Why doesn't acroread work on FreeBSD? Why, doesn't acroread work on FreeBSD? There's the issue of an unimplemented Linux system call:
But today I tried it running as root, and it worked. Tried again as myself, and it crashed, even when I setuid'd it. More googling showed nothing obvious, but then Callum Gibson told me “It works for me”. And indeed it did. He got the inotify_init message, but his acroread didn't crash. Why not? No idea, except that the inotify_init message was a red herring. But the googling suggested that this problem only occurs with acroread version 9, so installed version 8, and it worked.
In the meantime, though, there are other possibilities. Jürgen Lock suggested okular, so I installed that.
=== grog@eureka (/dev/pts/6) ~ 48 -> okular
Huh? What's it called, then? Looked in the package list, /usr/ports/graphics/okular/pkg-plist, and found:
=== grog@eureka (/dev/pts/6) ~ 50 -> less /usr/ports/graphics/okular/pkg-plist
So where is it? /usr/local/kde4/bin/okular, of course. One more reason not to like KDE. But yes, it works too. Problem solved.
Somehow things all seem to be so difficult. 20 years ago we put up with installation problems—anything was acceptable just to be able to run Unix on your PC—but times have changed, and if Microsoft appears easier than FreeBSD, then something's wrong.
|
Hacking Hugin
|
Topic: technology, photography, opinion | Link here |
I've been trying out the new version of Hugin. I can't say I particularly like it. It now comes with user levels (“Interfaces”), “Simple”, “Advanced” and “Expert”. The default interface is the “Simple” one, of course, and it shows a combination of the old “Assistant” and Fast Panorama Preview windows. It also bleeds text, which I find ugly.
|
||||||||||||||||||||||
|
|
||||||||
Once you select the “Expert” interface things don't look too different from before. Except for the “Identify” function in the Fast Panorama Preview. This shows the locations of the individual images and any masking, which is a very useful for more complicated panoramas.
|
||||||||||||||||||||||
|
|
||||||||
But it's gone!
Well, no, it has just been castrated. It is only active in the “Preview” tab, as shown, and only after pressing the “Identify” button on the left. And it's automatically deactivated when you change to another tab. To quote Thomas Modes:
So why not make it configurable? There's a typical answer to that question: “Send patches”. So I started looking at the code, which I suppose isn't that badly written for C++ code. But it is C++ code, and I drifted away from C++ decades ago, and I've never regretted it.
Somehow I got sidetracked. A thing that's even more irritating than this identify function is that in normal operation Hugin has two windows, and both really need to be full screen. OK, that's not a problem here, since I have 4 monitors. But how do you tell Hugin where to put the images? When you start it, it honours the DISPLAY environment variable, but there's no way to tell it to put the second window (fast panorama preview) on another display.
Went looking at how to do that. The displays are handled by wxWidgets, which has relatively copious documentation. But how do you find a description of something that might not even be implemented? I didn't, but Callum Gibson did. Within 10 minutes he came up with wxSetDisplayName (), which appears to be exactly what I want. Set off to hack it. But how? I don't do WxwidGetS, and every detail needed to be researched. The parameter is just the display name, which for the sake of testing I hard-coded as :0.1. But:
But the documentation said it's defined in wx/utils.h, and I included that explicitly. Looking at the header, though, I saw:
So the name is wxSetDisplay, not wxSetDisplayName. Documentation and header file out of sync. Tried again. Same problem. It seems that the preprocessor variable __X__ wasn't set. Set that, getting lots of warnings on the way. But it came further. It didn't like the literal parameter:
OK, no automatic promotion from char * to wxString &. Reading the documentation suggested that this should work:
But it didn't. Reading other parts of the code suggested that this should work:
wxT is a macro that widens them to Unicode. And yes, indeed, that did it. On to the link phase:
OK, which library is it in? There are only 72 to choose from, and I couldn't find the correct invocation of nm to show the contents. How about searching the source code?
=== grog@eureka (/dev/pts/8) /usr/ports/x11-toolkits/wxgtk28 61 -> find . | xargs grep wxSetDisp
So the only place that it's mentioned is in the header file. The function doesn't exist at all! But it's been in util.h for 10 years or so, it seems. Further investigation shows that it does exist in release 2.9.4, but I don't have that installed, and who knows what cans of worms that might cause? But I have the code for the function, so potentially I can incorporate it in my (badly) hacked version of Hugin. My impression of wxwidGets has not improved.
|
Flowering artichokes
|
Topic: gardening | Link here |
My neglect of the garden is showing in all sorts of places:
|
|||||||||||||||||||||||
That's one of the artichokes that got burnt in the recent heat wave.
The tomatoes are also bearing still-unripe fruit, and my tying to stakes has been less than successful. Somehow all this is too frustrating.
| Saturday, 23 February 2013 | Dereel | Images for 23 February 2013 |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
|
Shutdown hang
|
Topic: technology | Link here |
Yvonne woke me this morning to tell me that her machine hadn't shut down—something about “not ready”. At least that's better than “an error occurred”. But I saw something I have never seen before:
|
|||||||||||||||||||||||
|
|||||||||||||||||||||||
What causes that? Why did it want to suspend? It's supposed to be shutting down. Peter Jeremy investigated and came to the conclusion that it did make sense, but of course there's no way to know what really happened, and it didn't happen again. Presumably some hardware glitch.
|
Goodbye, Carola
|
Topic: general | Link here |
Carola Schlanhof left to go to Tasmania today. It's fun having her here, but it's also very tiring.
|
More fiddling with Hugin
|
Topic: technology, photography | Link here |
So it seems that I will need to install wxWidgets 2.9 to have a chance of setting multiple displays, and even then it's not clear that it will work. But maybe there's an easier way: Hugin honours the DISPLAY environment variable, so how about setting that inside the program, before the window is created? Did that—how easy C is in comparison to C++—but it had no effect. Presumably the widgets, or possibly GTK+, have looked at the variable on startup and hidden it somewhere difficult to find. That was to be expected, but I wonder how much sense it really makes? The idea of Object-oriented programming was to make things simpler, not more difficult.
Today was also house photo day, so I had my first opportunity to try the new version of Hugin in earnest. It didn't do well: the stitcher died without any message. And when I saved my projects, I discovered that they had spaces in the file names! The current version of Hugin gives a default project file name something like 00-19.pto, where 00 and 19 are the first and last image numbers. But now that's been changed to 00 - 19.pto. Vomit! And as a result, my scripts hadn't saved them.
So back to the old version. Today was anything but an ideal day for the photos: cloudless sky and bright sunshine. And it's becoming clear that I need to reconsider how I take these photos. In principle using automatic exposure for the component images works well—if the difference in brightness isn't too big. When it is, things can look less good. In these two consecutive images, the exposure differs by 1.7 EV, and the sky looks completely different:
|
|||||||||||||||||||||||
|
|||||||||||||||||||||||
Surprisingly, enblend does quite a good job of merging the images:
|
|||||||||||||||||||||||
Still, I need to pay more attention to the skies, which in general didn't look too good this week.
One other thing that I discovered while processing the abortive Apollo Bay beach panorama: I accidentally took the first image at 17 mm focal length, and the others at 12 mm focal length. I had intended to discard the first one, but the control point detectors fitted it correctly, so I tried stitching it anyway to see what would happen—and it worked! This is a more extreme example of the image alignment problems I described earlier this month. When I first tried it I had no success at all: the control point detectors couldn't work it out. What has changed? I wish I knew. But today I tried another one: the “verandah east” panorama has an issue with the exposure of the Hebes to the left of centre, which are in the shade. So I took the entire panorama with a focal length of 9 mm, and then just the shadow part of the Hebes at 18 mm. And Hugin had no difficulty aligning them:
|
||||||||||||||||||||||
Clearly there's much room for experimentation here.
And yet another discovery: I have been using the panomatic control point detector, because it seemed to work better than CPFind, but today it failed on two panoramas. Tried CPFind, and it had no difficulties. There's something very strange about these control point detectors.
| Sunday, 24 February 2013 | Dereel | |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
|
Hacking test
|
Topic: technology, opinion | Link here |
Talking with Peter Jeremy on IRC this morning, and he bemoaned that fact that test(1) doesn't have comparison operators between files based on creation time or access time. All that's currently available is a comparison between modification times:
=== grog@eureka (/dev/pts/9) /var/tmp 29 -> touch foo
=== grog@eureka (/dev/pts/9) /var/tmp 30 -> touch bar
=== grog@eureka (/dev/pts/9) /var/tmp 31 -> test foo -nt bar; echo $?
=== grog@eureka (/dev/pts/9) /var/tmp 32 -> test bar -nt foo; echo $?
=== grog@eureka (/dev/pts/9) /var/tmp 33 ->
-nt means “newer than”; there's also an -ot. Wouldn't it be nice to have, say, comparison operators -cnt and -cot? I had often thought that myself, and it's not difficult, so set to doing it. There's only one source file, test.c, and it has an struct array t_op with entries like:
That's a diff; the lines starting with + are my additions. Each entry contains the operator name, a numerical value, and the kind of operator. Further on there's a big case statement checking for the value:
Clearly I needed to write the functions, but that's a straightforward copy of the functions for modification time:
And that's just about all, at least to get it running. Compiled and tried it out:
=== grog@eureka (/dev/pts/19) /src/FreeBSD/svn/head/bin/test 11 -> test test.c -cnt test.o
=== grog@eureka (/dev/pts/19) /src/FreeBSD/svn/head/bin/test 12 -> ./test test.c -cnt test.o
The first one was the old test executable in /bin, so that was to be expected. But what went wrong the second time? What does that message mean? Looking at the code, it's at the very end of main:
nargc is a copy of argc, which is apparently kept to keep track of unprocessed arguments. And here it found some. Why? gdb is your friend:
=== grog@eureka (/dev/pts/19) /src/FreeBSD/svn/head/bin/test 20 -> gdb test
=== gdb -> b 238
=== gdb -> r test.c -cnt test.o
=== gdb -> p nargc
=== gdb -> r test.c -nt test.o
=== gdb -> p nargc
So this appears to be its way of reporting syntax errors. Looking further showed that the function newercf didn't get called. Why not? The code is pretty straightforward. Finding the operators is a linear search through the ops array:
OK, iterating through that I found:
=== gdb -> disp op->op_text
=== gdb -> c
=== gdb -> p s
OK, that's clear. We have the parameter s set to -cnt, and op->op_text is also -cnt, so strcmp should return 0, right? Up in there to have a look:
=== gdb -> b strcmp
=== gdb -> c
=== gdb -> i reg
=== gdb -> x/10x $rsp
=== gdb -> fin
Round about here I was reminded that I haven't learnt any assembly language properly since the Z-80, over 30 years ago. My knowledge of the ia32 instruction set and registers is incomplete, but I know almost nothing about amd64. Still, some things should be clear: many of the register names have genealogies going back to the Intel 8008, and almost certainly rax is the accumulator and rsp is the stack pointer. And looking on the stack, I can see the addresses of the parameters. In particular, the first two words are the return address, 0x00000000004019f3 in full. Let's see what strcmp returns:
=== gdb -> b *0x004019f3
=== gdb -> c
=== gdb -> i reg
1! Not equal! How can that be?
=== gdb -> p s
=== gdb -> p op->op_text
As expected, they're the same. But strcmp says they're different. Character representation problems? Let's look at them in hex:
=== gdb -> x/10xb s
=== gdb -> x/10xb op->op_text
Well, they're the same, up to the end. But instead of a 0x00 at the end of op->op_text, there's a 0x15! What's that doing there, and why isn't it being displayed? Everything's clear once you understand the issue. struct op is defined as:
In other words, the first element of the struct isn't a string, it's an array of characters, only 4, including the terminating \\000. Adding operands with 4 characters overflowed into the terminator. And gdb only showed as many characters as there were in the element, so it wasn't clear that the string wasn't terminated. The fix is simple:
I never cease to be amazed how easy it is to get off on the wrong track. It even happened while reading this diary entry. I went back to test.c and found:
That's what I had expected, and it would have avoided the problem. But what's it doing there? I didn't put it there, and it's not what I saw yesterday. More research shows that this is the version in 9-STABLE, and it has been changed since then:
Is that worthwhile? The string requires a pointer for each of the 45 operators, a total of 360 bytes that can be saved by this change. On the other hand, making it a character array wastes 40 bytes, and now I have expanded the size, it wastes another 41 bytes. Total savings: 279 bytes. The size of the executable is:
=== grog@eureka (/dev/pts/20) /src/FreeBSD/svn/head/bin/test 4 -> size test
The page size is 4 kB, so the text occupies 3 pages. Add 279 bytes to that, and the text occupies 3 pages. Net gain: nothing. Net pain: described above.
| Monday, 25 February 2013 | Dereel | Images for 25 February 2013 |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
|
Hacking test: So nice, so nice, we do it twice
|
Topic: technology, opinion | Link here |
Hacking test(1) was easy, apart from the issue I had by not reading the struct header. Peter Jeremy thought so too, so he came up with his version, which went further than mine: I compared the modification timestamps, creation timestamps or access timestamps of two files. Peter made it more general: compare any two timestamps, including the birthtime stamp introduced with UFS 2, a total of 16 possible comparisons instead of the 3 that I had envisaged. What use are they? Who knows? You can't guess what people might like to do with the tool.
That required even longer strings than I had envisaged. I had added things like -cnt (creation newer than) analagous to -nt. Peter did it the other way round: -ntcc (comparing both creation timestamps). And he did it without making the individual struct entries any longer. Here his operator structure:
My comments from yesterday still stand: why bother? He considered that having the table entries on 8 byte boundaries could add some efficiency, but it's all relative when it's a linear search, and usually performed only once or twice per invocation. And just the effort of launching the process would be orders of magnitude more than the run time of the algorithm. And all in all, that time is:
=== grog@eureka (/dev/pts/24) ~ 7 -> time test ! 2 \< 1
That's a total of 4 times through the table, and the CPU time is still not measureable.
|
Strange exposure problems
|
Topic: photography, opinion | Link here |
A couple of weeks ago I saw an article by Hans-Peter Brandhueber in the German Olympus Forum (here in Google translation) about some photos that were spectacularly underexposed. They were taken in the snow, and the general consensus was that the snow had influenced the exposure. I wasn't convinced: some of the photos were far too underexposed for that to have been the case.
Then Hans-Peter put up the images with EXIF data, and Yvonne Steinmann did some corrective work on them. Her results were interesting: she had to increase the brightness by between 0.95 and 3.45 EV—a difference of 2.5 EV. In particular, it applied to these two photos (shown here without correction):
|
|||||||||||||||||||||||
|
|||||||||||||||||||||||
But they're almost exactly the same motive! And it's clear that the first one is underexposed even in the highlights. How can the exposure meter be responsible for that?
But there's more. Looking at the EXIF data, the first image was exposed at EV 17.3, the second at 16.6. That's only 0.7 EV difference. Yvonne found a difference of 2.5 EV. Where did the other 1.8 come from? Looked more carefully and found a strong correlation between shutter speed and underexposure:
| Photo | Shutter speed | Adjustment (EV) | ||
| 30295787 | 1/4000 | 3.45 | ||
| 30295873 | 1/4000 | 3.0 | ||
| 30296015 | 1/4000 | |||
| 30295938 | 1/3200 | 1.4 | ||
| 30295961 | 1/3200 | 0.95 | ||
| 30295985 | 1/3200 | 1.7 | ||
| 30295823 | 1/2500 | 0.95 | ||
| 30296277 | 1/1600 | |||
| 30296168 | 1/160 | |||
Some adjustment values are missing: Yvonne didn't do all the images. But two of the missing ones are interesting: the third exposure at 1/4000 s, and the exposure at 1/160 s. Here all three at 1/4000s:
|
|||||||||||||||||||||||
|
|||||||||||||||||||||||
|
|||||||||||||||||||||||
By comparison, the one taken at 1/160 s looked roughly OK:
|
|||||||||||||||||||||||
So it looks very much as if there's something wrong with the shutter. 1/4000 s is the fastest shutter speed on this camera, and even minor timing differences can have a profound influence on the effective shutter speed.
|
Measuring exposure accuracy
|
Topic: photography, opinion | Link here |
Looking at Hans-Peter Brandhueber's photos made me wonder: how well does my camera do at high shutter speeds. There's an easy way to test that: mount the camera on a tripod, set the camera to aperture priority metering, set the aperture wide open, then tweak the ISO rating until the camera chooses the fastest shutter speed. Take a photo, stop down a stop, and repeat. Compare.
Tried that out, and the results were much more uniform than I expected. Here the results for 1/8000 s at f/2.8 and 1/125 s at f/22), with mouseover alternation:
|
|||||||||||||||||||||||
|
|||||||||||||||||||||||
I processed these images with DxO Optics “Pro” with no corrections, which made a big difference in the processing time. Normally I get times that, for no recognizable reason, vary between 25 s and 60 s per image. Without any correction I processed 7 images in 1:56 minutes, or about 17 s per image.
There's as good as no difference in the brightness. Allowing DxO to choose the optimal exposure shows a very slight tendency. In all cases there was a decrease in exposure, most noticeably at higher speeds:
| Exposure (/s) | Compensation (EV) | |
| 8000 | -0.47 | |
| 4000 | -0.41 | |
| 2000 | -0.46 | |
| 1000 | -0.46 | |
| 500 | -0.44 | |
| 250 | -0.39 | |
| 125 | -0.25 | |
But that's so minimal that it's not even clear whether it's the shutter or the diaphragm. If it were the shutter, you'd expect 1/8000 s to be more markedly different. So I'm at least happy that my camera doesn't appear to have any such problem. Hans-Peter Brandhueber has an Olympus E-M5, which has a significantly different construction from my E-30. Is there some issue with this model in general? Or with Hans-Peter's camera? Or is it only an issue when it's really cold?
|
Aligning images: still no silver bullet
|
Topic: photography, technology, opinion | Link here |
One of the things that I wanted to do with the exposure comparison was to compare the histograms as well. To do that I had to take screen shots of the histograms and then align them somehow—and what better tool than Hugin?
I don't know, but it seems I need one. I followed my instructions, but the images didn't get aligned properly without selecting “View” optimization, and when I did, it couldn't cope. Maybe part of the story is the concept of focal length, which is completely missing here. More head-scratching needed
|
Power surge
|
Topic: general, opinion | Link here |
Power surge in the afternoon. No non-electronic devices had problems, but it caused the UPS in the lounge room to shut down Yet Again while I was watching TV, thus turning off the projector that must never be turned off abruptly. And that's a 1000 VA UPS delivering about 250 W. Why do UPSs turn off because of a surge? It also wedged the Ethernet switch in Yvonne's office. Grrr.
| Tuesday, 26 February 2013 | Dereel | |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
|
Powercor pays up
|
Topic: general | Link here |
Letter from Powercor in the mail today: a cheque for $100 because I had had more than 20 hours of “sustained” (over one minute in duration) power failures in some unspecified calendar year. That fits last year, anyway. But according to the sheet they enclosed, there's also money for “momentary” (less than one minute) interruptions: $25 for 24 in a calendar year. And in 2011 we had 28 of them. I need to check what else they have forgotten.
|
Differing fields of view?
|
Topic: photography, technology, opinion | Link here |
Another message in the German Olympus Forum today: differing fields of view for Panasonic and Olympus cameras. The latter is claimed to have a slightly wider field of view. My guess was that the submitter was processing Olympus raw images with a program like ufraw, which leaves the edge in place. These images have mouseover alternation:
|
|||||||||||||||||||||||
|
|||||||||||||||||||||||
They also show (second image) why I don't use ufraw any more.
|
Hottest February on record?
|
Topic: general, opinion | Link here |
Watching Al Jazeera TV today, a message scrolled across the bottom of the screen: Melbourne has already had its hottest February on record, with 15 days over 30°. Afterwards went looking for confirmation. I found none. The ABC had an article that suggested that the record of 14 days had been equalled, but not broken. Still, that was yesterday. So I looked on Al Jazeera's web site. What a catastrophe! No mention on the front page, and the searches all ended up nowhere. A search for melbourne heat wave gave me three results, all 2 years old, including this article about the Fukushima nuclear plant. But then there was a link 15 day heat wave worst, but that took me to yet another results page with no result less than a year old. I'm left with the impression that they have forgotten to update their search engine for the past 12 months.
Still, it has been particularly hot here, and my own results show:
17 days with maxima over 30°, and a highest temperature of 39.2°! By comparison, the highest temperature in Melbourne was only 37.2°. And yes, I believe these temperatures are correct. No wonder I haven't wanted to do anything in the garden. Last year the maximum temperature was significantly higher (41.8°), but we only had 12 days over 30°.
Later, while writing this entry, I found updates: Weather News - Melbourne breaks heat wave record. Broken link. And other links showing only 14 days of temperatures over 30°. But there's always the Bureau of Meteorology's monthly (“daily”) tables, and they still only show 14 days over 30°.
|
Comparing histograms: not worth the trouble
|
Topic: photography, opinion | Link here |
Finally got round to getting my histograms trimmed for comparison, with no better tool than xv, manually trimming them to the same crop. I don't know if it was worth it. Here a couple (1/8000s and 1/125s) with mouseover alternation:
|
|
||||||||
|
|
||||||||
Yes, there's a difference there, but I can't see what it's trying to tell me.
| Wednesday, 27 February 2013 | Dereel | |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
|
Powercor: the sting in the tail
|
Topic: general | Link here |
Yesterday we got $100 refund from Powercor because of poor quality of service. This morning, it seems, they're trying to work up to the next cheque. No less than four short dropouts, followed by a longer one of 2 hours, 24 minutes. That's a total of 2:39 hours of “sustained” outages this year, and 8 “momentary” outages. Extrapolate that for the entire year and we'd have 16 “sustained” interruptions ($150) and 48 “momentary” outages ($35). I'd still rather have no outages.
|
More laziness
|
Topic: general, opinion | Link here |
Somehow didn't get much done today. It's no longer as hot; in fact, the temperatures have dropped markedly, and they continued doing so for most of the day:
So why am I still so lazy? Because I can? I've currently been reading my diary for late 1968, and it seems that I was just as lazy then. So maybe this is the Real Me.
| Thursday, 28 February 2013 | Dereel | |
| Top of page | ||
| previous day |
|
Summer winds down
|
Topic: general, gardening | Link here |
The arguably hottest February on my record went out with a whimper, not a bang. After 17 days this month with top temperatures over 30°, today barely managed 20°. And that gave me some time to do some work in the garden, which is sorely in need of it. If only I could keep up my momentum...
|
Reinstating my ports commit bit
|
Topic: technology | Link here |
I've been updating a number of ports recently, and since I have handed in my ports commit bit, I had planned to get other people to commit them for me. But in each case they approved the commit and I ended up committing it myself. That's possible, but clearly not what's intended. And then Beech Rintoul sent a message asking for a new maintainer for the ImageMagick port, a tool I use quite a lot, so it seemed reasonable to take over maintainership. Spent much of the day getting it and a dependency, webp up to date. So it was clearly time to apply to get my commit bit back. Why do I do these things?
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 |
