|
| Greg's diary |
| January 2012 |
|
|
||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
|
|
|
|||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
This view of the diary is limited to these topics: technology. There may be lack of continuity in the text, and some days may be completely missing. In case of doubt, please enable the complete display.
| Monday, 2 January 2012 | Dereel | Images for 2 January 2012 |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
|
Reprocessing old photos
|
Topic: photography, opinion, technology | Link here |
After getting the Olympus E-510 in August 2007, It took me quite some time to get good results from the postprocessing. Some of it was a learning experience, some was lack of suitable software. I've learnt a lot since then, and I've found ways to improve the quality of the images, but can DxO Optics "Pro" do better? Trawled through my old diary entries and got as far as three years ago, when I had serious issues with exposure and gradation. It proves that, for some reason, some of the images really were underexposed, but DxO brought much better results. Here the comparison of processing the same image with ufraw 0.13, ufraw 0.14.1, dcraw and DxO. Mouseover shows the next image for comparison.
|
|||||||||||||||||||||||
|
|||||||||||||||||||||||
|
|||||||||||||||||||||||
Also reprocessed the panoramas. I couldn't completely fix the verandah panorama, because I took it without parallax correction, but the results look a lot better.
|
|||||||||||||||||||||||
|
|||||||||||||||||||||||
Another couple of comparisons are the backlit flower baskets:
|
|||||||||||||||||||||||
|
|||||||||||||||||||||||
That one's not as clear-cut.
How well could I have done without DxO, just with better tuned ufraw? I don't know, and I don't intend to look yet. First there are other images to compare.
| Tuesday, 3 January 2012 | Dereel | Images for 3 January 2012 |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
|
More photo processing comparisons
|
Topic: photography, technology | Link here |
Spent much of the day with further reprocessing old photos. The results were quite useful. On 4 April 2009 I had another case where a panorama changed from “useless” to “acceptable”:
|
|||||||||||||||||||||||
|
|||||||||||||||||||||||
If it hadn't been for the reflections of the sun in the lens, it wouldn't have been at all bad, though it's clear that it challenges the limits of the dynamics.
Then I looked at some comparisons I had done on 12 April 2009, where I had had extreme problems with contrast. At the time I hadn't realized that some of the camera functions, such as “vivid” and “soft” images, only applied to the JPEG output. In addition, I didn't find a camera JPEG image for all the comparison photos, so today's comparisons aren't all they could be.
A few days later I received a comparison photo for the first sequence from Michael <no surname> in South Australia, processed with Photomatix. In the following comparison I have the original from the camera, the best I could do to improve it with ufraw, the results from Photomatix and the output from DxO Optics "Pro". As usual, moving the mouse over the image gives the next in sequence.
|
|||||||||||||||||||||||
|
|||||||||||||||||||||||
|
|
|||||||||
|
|||||||||||||||||||||||
I think it's fair to say that DxO produces by far the best image, and probably about as good an image as I could hope for from this particularly difficult scene. The shadows are more dynamic than Photomatix, and the highlights are much more natural. Conceivably Photomatix could do better—Michael said that he didn't try too hard, which is understandable—but then DxO might do better too, and there's only so much work I want to put into these images.
The other image wasn't really a good comparison 3 years ago, but here it is with DxO, which once again does a good job:
|
|||||||||||||||||||||||
|
|||||||||||||||||||||||
A couple of months later I tried some different scenes, this time comparing images from the Olympus E-510 and the E-30. No hard contrast, just overall image gradation. Today I was more interested in the postprocessing, and tried the following: JPEG original from the E-30, DxO with no correction, DxO with default correction, DxO with “realistic” HDR-style tone mapping, and DxO with “artistic” HDR-style tone mapping.
|
|||||||||||||||||||||||
|
|||||||||||||||||||||||
|
|||||||||||||||||||||||
|
|||||||||||||||||||||||
|
|||||||||||||||||||||||
The tone mapping doesn't seem to make much difference, presumably because the dynamic range was limited, but the “artistic” version makes it gaudy enough. Better than default? Here's the comparison:
|
|||||||||||||||||||||||
A matter of taste, I suppose.
| Wednesday, 4 January 2012 | Dereel | |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
|
Migrating to amd64, next try
|
Topic: technology | Link here |
So far my experience has shown that DxO Optics "Pro" is quite a useful program. The only problem, apart from the fact that it needs Microsoft to run, is that it uses so much memory that I can't run it on VirtualBox on my 32 bit machine. I started trying to upgrade to amd64 (64 bits, without the memory limitations) nearly 8 months ago, and gave it up two weeks later because of X issues. I then tried again in August, but the pain of upgrading was just too much.
By today I had forgotten some of the pain, and so started trying to upgrade again. I've had the machine running as a test machine all this time, so all I really needed to do is to upgrade the kernel and the ports. The kernel upgrade was trivial. The ports, once again, were not.
First,
So? Why does it stop there and wait (maybe all night) for me to take notice of it for the hundredth time? The same happens for postfix.
Then there was one I hadn't seen before:
OK, so what's the replacement? php53? No, they've done almost the right thing and removed part of the version number from the preferred version. It's called php5, though it should be called php of course. But I had to go looking for that. People tell me I should be reading (/usr/ports/)UPDATING, but why? Why can't that information be output when the build fails? In any case, I couldn't find any mention of php52 there (in a file nearly 100 pages long). And after that, of course, I could wait for the messages about java16 and postfix.
The next one was a superb shoot-in-foot incident:
Why that? It says it has upgraded it, so there was no reason to keep the old version. But there was no new porrtmaster installed. Why not? I couldn't even find the backup I had asked for, so I had to reinstall it first.
And that's as far as I got in a whole day. This process should run in the background without my intervention and give me a summary of anything that fails. And failure should be a rare event, not like every time I've tried to upgrade ports in the last few years. No wonder Kirk McKusick has given up on straight FreeBSD.
| Thursday, 5 January 2012 | Dereel | Images for 5 January 2012 |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
|
Ports build, continued
|
Topic: technology | Link here |
Continued with my ports build today. Came into the office and found an old acquaintance:
In summary: I still don't know how to build perl from the Ports Collection. Maybe if I blew everything away and started again, it would be better. But that's not sure, and it could be even more pain. As I did last time, I just downloaded the package, once I found it:
=== root@defake (/dev/pts/0) /usr/ports 90 -> pkg_add -r perl5
=== root@defake (/dev/pts/0) /usr/ports 91 -> pkg_add -r perl
=== root@defake (/dev/pts/0) /usr/ports 92 ->
But that's perl-5.12. There's a 5.14 out there. That must be newer. Why have I been given an old version of perl? See if I care.
Then there was docbook. I seem to have 7 different versions installed. It reinstalled 6 of them:
Why? And why?
| Friday, 6 January 2012 | Dereel | Images for 6 January 2012 |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
|
More ports pain
|
Topic: technology | Link here |
Continued with my attempts to update the ports tree today. More pain. First it died in libass, whatever that may be:
What's that? Took a look in the build directory, and sure, ./configure is there. But the directory name is libass-0.9.13, not libass-0.10.0. Why that? Did a make clean, built, and it installed.
It went on, though, although I removed the C option. Later I got a crash after this message:
And of course COMPAT_FREEBSD7 was set in the kernel config. Ran things again, and the message didn't occur again. Instead, firefox died with:
That's ridiculous. I had just (with much pain) installed version 5.12.4. Did some old version get left behind? I should have checked config.log, but instead I checked with ktrace and found the same thing:
Well, it's right there. There is no perl5.10.1. But the error message is misleading: it suggests that there is an old version of perl there. And where did it get that from, anyway? It didn't try anything else. This one is apparently (marginally) my fault: I ran portmaster with the C option, which tells it not to run a make clean before building. Of course, the build process should have noticed that the config.status was out of date, but that one is understandable. Moral: don't use the C option.
So finally firefox version 9 was installed, so portmaster went off and started building firefox version 5. Why? And how? We don't have a version 5 in the Ports Collection any more, but there's a /var/db/pkg/firefox-5.0,1 alongside /var/db/pkg/firefox-9.0.1,1. Did a pkg_delete on firefox-5.0,1, and of course tore firefox-9.0.1,1 apart too:
=== root@defake (/dev/pts/0) /usr/ports 118 -> pkg_delete /var/db/pkg/firefox-5.0,1
=== root@defake (/dev/pts/0) /usr/ports 120 -> pkg_delete -f /var/db/pkg/firefox-5.0,1
=== root@defake (/dev/pts/0) /usr/ports 121 -> wh firefox
So I had to reinstall it. That was fast, but ineffective:
=== root@defake (/dev/pts/1) /usr/ports/www/firefox 40 -> make install
=== root@defake (/dev/pts/1) /usr/ports/www/firefox 41 -> firefox
The problem here is that the Ports Collection notes in the build directory whether the package has been installed or not. To get it to reinstall, you first need to remove a secret file:
=== root@defake (/dev/pts/1) /usr/ports/www/firefox 42 -> rm work/.install_done.firefox._usr_local
=== root@defake (/dev/pts/1) /usr/ports/www/firefox 43 -> make install
Edwin Groothuis tells me there's another target, reinstall, that does this for you.
Looking at what was going on, this seems to be a bug in portmaster. It finds a directory /var/db/pkg/firefox-5.0,1 and establishes that it's out of date, and that the directory in the Ports Collection is /usr/ports/www/firefox, which currently builds irefox-9.0.1,1. But it doesn't check whether that is installed or not. Maybe it would have rebuilt it and reinstalled it. If it does a make clean it won't see that it's already installed any earlier.
Still, finally I had firefox upgraded. Easy, isn't it? Enough pain for a day.
| Saturday, 7 January 2012 | Dereel | Images for 7 January 2012 |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
|
More photo experiments
|
Topic: photography, technology, opinion | Link here |
Another windy day for the house photos, but managed to get most of them anyway. I've been taking these panoramas for over three years now, but I continually run into new experiences:
A few weeks back I changed the use of flash in the verandah photos to use TTL flash for the first image, which meant taking the first image with flash, turning off the flash unit, taking the other two, turning the flash unit on again and moving on to the next position. And from time to time the camera would hang when the flash was turned off. It happened again today. After turning it on again, the camera still hung—but only for a while. Then there was some memory activity (the LED flashed for a while) and it continued.
This happened twice, and each time it was after the flash fired at full intensity instead of normally—something that I need to work on. So there appears to be some connection there.
One of the panoramas didn't close properly: Hugin found control points for all except one overlap. It was a 360° panorama, so that was “enough”, and it told me I had a good match: average error 1.3 pixels, maximum error 3.8 pixels. But when I looked at the preview, things were anything but good. The trees on the left don't line up:
|
||||||||||||||||||||||
Adding control points at that overlap didn't help; the preview still showed this enormous discontinuity. Strangely, though, the result looked OK: something in the final stitching must have improved the situation.
This seems to be another case where history makes a difference with Hugin. In this case, it had determined that the effective focal length of the lens was 9.11 mm, when in fact it was closer to 8.9 mm, and I think this might have made the difference.
It was a pretty grey day today, and the images looked pretty washed out. Tried some new settings with DxO Optics "Pro", and they seem to help. Here the normal and then the enhanced view, as the mouseover shows:
|
|||||||||||||||||||||||
|
|||||||||||||||||||||||
That was done with the “Artistic” variation of the “single shot HDR” “preset”. In general it looks too gaudy, but it's probably what I need here.
Did more exposure comparisons with automatic and manual exposure settings for the components of the garden-se photos. Today there was no sun, so the exposure differences weren't as extreme as they might have been.
|
|||||||||||||||||||||||
|
|||||||||||||||||||||||
The second image was taken with automatic exposure, offset by -0.7 EV by mistake. But the results look almost identical. Clearly this kind of lighting isn't as much of a problem as the strong contrasts in sunshine.
The photo from the south-east corner of the house also looks like it could do with automatic exposure. In particular, the right-most component is extremely underexposed with uniform manual exposure:
|
|||||||||||||||||||||||
|
|||||||||||||||||||||||
Probably another one to try with automatic exposure, but the “HDR” processing and Hugin generated a reasonable result, though the shadow detail on the right leaves something to be desired:
|
|||||||||||||||||||||||
It looks as if it will be a long while before I get all this sorted out to my satisfaction.
|
Port build pain continued
|
Topic: technology, opinion | Link here |
Didn't have much time for my ports build today, but that didn't mean things went smoothly. While building nmap, got this message:
That wasn't a portmaster issue: it happened when I tried to build it in the normal manner. On three different machines. With two different versions of nmap. And on checking, discovered that there is no mention of nmap.h anywhere in that directory. The file itself was in the parent directory.
Discussed it on IRC, and Jürgen Lock suggested that I try it without any preset environment variables (su - root -c /bin/sh). And yes, that worked. So it's an environment variable, but which? Or is it? Went back to build the port again, this time with my standard bash shell and environment, and it worked too. Just more pain.
While watching the build information fly by—I have collected 336212 lines, or about 5,500 pages of it so far—discovered a number of these:
So far there have been 9 of them. Why? And what is this port?
Things continued as far as:
Then I gave up for the day. But the Ports Collection wasn't done with me. Tried to back up the photos I had been processing and got:
=== root@defake (/dev/pts/1) /usr/ports 98 -> syncphotos
Where did rsync go? Was it in the process of building it? No, it seems that it thought it had built it, but it was gone. Looking through the log files, I find:
Some success! It seems that I shouldn't have even tried to use portmaster. But in general, the amount of trouble I've had here is an order of magnitude worse than any other system I have used, and it's not the first time. If I could find something better that didn't want to tell me how to think, I'd move to it immediately.
| Sunday, 8 January 2012 | Dereel | |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
|
eSATA docking station
|
Topic: technology | Link here |
It's been two weeks since I received my hard disk docking station. The first experience wasn't good, and I didn't have a SATA disk handy to put in it; the only one was my 2 TB backup disk, and after seeing what happened with an SD card, I didn't want to risk that. Today finally got round to taking a spare SATA disk out of the external housing and putting it in the docking station, in the process discovering confirmation of what I had suspected, that the eSATA interface was only for the SATA drive:
|
|
||||||||
The SATA disk fitted better than the attempts with the PATA disk, and it powered up. And that was all. The system didn't recognize the disk, so I rebooted—I've had trouble before with another low-quality eSATA external housing—but it still wasn't recognized. What's wrong with it? Do I care?
|
Upgrading ports, less pain
|
Topic: technology | Link here |
On with the ports upgrade today, without running into many problems. It wasn't until I rebooted defake that I discovered that I could no longer communicate with it: xterm had gone away. Why? That's the second program that just disappeared, after rsync.
rsync wasn't to be outdone, though. It went away again! I don't understand that, since it happened when I thought I hadn't done anything. But I can't be bothered crawling through the equivalent of 5000 pages of log information to try to piece together what went wrong.
| Monday, 9 January 2012 | Dereel | Images for 9 January 2012 |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
|
Portmaster: give up
|
Topic: technology, opinion | Link here |
Portmaster is finished!
That must be about the twelfth reinstallation of poppler, but it's finished. Well, it reinstalled poppler many times, but it removed xterm, and it's still gone. Why? What else is gone?
It's been 5 days since I started this update. The intention had been to do it faster than a complete rebuild of the ports. But it doesn't seem to have been. What went wrong? Jürgen Lock suggests that the presence of the FORCE_PKG_REGISTER environment variable or the -C option (don't run make clean) that I had at the beginning might be to blame. But I don't care. It has just taken too long. Went back to reinstall the ports from scratch, once again going through the pain of hundreds of default configurations. And then Jürgen told me about the -P option for portmaster, which downloads the precompiled packages. Maybe I should have tried that. I probably will next time.
|
Recharging NiZn batteries again
|
Topic: technology, photography | Link here |
It's been two months since I've paid much attention to the Nickel-Zinc batteries I bought in October. I haven't been demanding so much from the flash unit, but today it finally told me they were discharged. And indeed they were: the voltages were down to 1.549 V, 1.486V, 1.437 V and 1.381 V, the last one that I had already marked as not being quite up to the same level as the others. It's difficult to be exact, because it was clear that they were still recovering from the flash, and the voltages were gradually rising. Still, a long way below the 1.6 V in the specifications. After charging, the voltages were 1.832 V, 1.832 V, 1.829 V and 1.826 V (this last the marked battery).
| Tuesday, 10 January 2012 | Dereel | Images for 10 January 2012 |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
|
Building ports from scratch again
|
Topic: technology | Link here |
As planned, continued building the ports on defake with my old, incomplete build method. In principle it should work, but for some reason today the build died at the start because I had two rules to build X. How did that happen? Checked out the last version of the rules file again, and discovered it was already in there. My best bet is that previously the duplicate rules didn't do any harm, while now they caused the build to fail.
What did do harm was checking out the rules file. It seems that in so doing I overwrote some changes I had made since the last checkin, and I was missing rules for a number of ports. In the course of the day, had to update the rules file a number of times, but nothing else went wrong.
In the process, found the reason for my FORCE_PKG_REGISTER environment variable:
No mention there of any side effects.
| Wednesday, 11 January 2012 | Dereel | Images for 11 January 2012 |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
|
Ports build log
|
Topic: technology | Link here |
The ports build continued relatively smoothly today. multimedia/dvdauthor failed:
Typical reasons for that are out of date ports, so synchronized my repository, ran cvs up -Pd, and tried again. No improvement. Further investigation revealed:
=== grog@defake (/dev/pts/1) /usr/ports/multimedia/dvdauthor/files 2 -> l
=== grog@defake (/dev/pts/1) /usr/ports/multimedia/dvdauthor/files 3 -> cvs update -Pd
What's that doing there? The whole idea of the d option is to remove things. This is clearly because it was modified. But how? Did I do that?
=== grog@defake (/dev/pts/1) /usr/ports/multimedia/dvdauthor/files 4 -> diff -wu patch-src__subgen-image.c.orig patch-src__subgen-image.c
=== grog@defake (/dev/pts/1) /usr/ports/multimedia/dvdauthor/files 5 -> cat CVS/Entries
Nope, wasn't me. And it was a long time ago. What went wrong there? Anyway, removing the files solved the problem. And, apart from further entries missing in my portrules file, that was the only problem. Certainly easier than portmaster so far.
| Thursday, 12 January 2012 | Dereel | Images for 12 January 2012 |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
|
Port build: take a break
|
Topic: technology | Link here |
In this morning to discover that my portrules file had been overwritten again with the last checked-in version. Do I have some hidden automatic checkout somewhere? I don't know where it would be. Fortunately I had a backup of the last version, so I was able to continue, but somehow this is all just too much pain, so the next time it stopped, I left it until tomorrow.
|
Network errors again
|
Topic: technology | Link here |
My Internet connection has been surprisingly reliable in the last few months. If I had known about it 4 years ago, I would have set it up then, but the threshold of installing new hardware for an uncertain outcome, combined with the terrible service I got from Telstra ensured that I didn't try it until I had confirmation from others that it was a viable alternative.
I've been keeping a statistics page for the link, in line with older ones for satellite and ADSL. But it doesn't really give much useful information any more. Many short dropouts, including restarting the PPP process, go unnoticed, and others are due to issues beyond the ISP's control, such as this horrible Huawei 1762 modem hanging (solution: pop the modem from the USB slot, replace, wait for one failed connect).
Apart from that, things have been good for months. In the previous 3 months I had only 8 dropouts, some doubtless for other reasons. Today, though, things were worse:
Popped the modem, replaced it, but the system didn't reconnect: the authentication failed. Several times. Finally it worked, but the throughput was terrible, with the old 2 minute ping times. Called up Internode support and spoke to Jesse, who went through the usual stuff, and promised to call back when he had something. He didn't, but the network came good after a couple of hours. Hopefully this is only an isolated incident.
| Friday, 13 January 2012 | Dereel | Images for 13 January 2012 |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
|
Ports progress
|
Topic: technology | Link here |
On with the ports very much in the background. With the help of locate found an old copy of the portsrules file that greatly helped fill in the puzzles. Nevertheless found one port that didn't build:
This time it wasn't a CVS consistency issue. Clearly there's something wrong with the port. But it builds in the build cluster, so what's the issue? I can't be bothered. I wasn't even sure what flphoto is (it's a basic image display program, which I seem never even to have tried out), so I installed it from the package. But why is this all so painful?
|
FreeBSD dmr edition
|
Topic: technology, opinion | Link here |
FreeBSD 9.0 was released today (or yesterday in the USA; it wasn't intended to be on Friday the 13th). And for the first time I recall, it had a dedication:
The FreeBSD Project dedicates the FreeBSD 9.0-RELEASE to the memory of Dennis M. Ritchie, one of the founding fathers of the UNIX™ operating system. It is on the foundation laid by the work of visionaries like Dennis that software like the FreeBSD operating system came to be. The fact that his work of so many years ago continues to influence new design decisions to this very day speaks for the brilliant engineer that he was.
May he rest in peace.
Amen.
|
System crash
|
Topic: technology | Link here |
While watching TV, teevee hung. NFS problems. It turned out that dereel had crashed, apparently because of still more USB disk problems. If only I could get this eSATA stuff working well!
| Saturday, 14 January 2012 | Dereel | Images for 14 January 2012 |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
|
Unexpected results
|
Topic: technology | Link here |
I often see messages like this one:
But this one was different. It was from Wikipedia. It wasn't repeatable.
|
Problems syncing
|
Topic: technology, photography | Link here |
That wasn't the only problem I ran into. While syncing my photos to the external web site, I got this:
A repeat was successful. What's that? A network problem?
| Sunday, 15 January 2012 | Dereel | Images for 15 January 2012 |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
|
Ports build: done!
|
Topic: technology | Link here |
I have finally finished my ports upgrade. Building from scratch worked well, with ultimately only one port (flphoto) requiring to be fetched as a binary package. And now FreeBSD 9.0 is out. Time to start over again?
In the meantime tried upgrading the build machine. Found a couple of 1 GB SIMMs lying around, so put them in the machine, and also an nVidia display card that I had been given recently. I had been told it was PCI, and I didn't really check, but on examination it's a 16 lane PCIe board. And it only has a DVI connector. And I don't have anything I can connect that to. I'll have to mess around yet more to get it up and running. About the only thing I could confirm is that Hugin crashes when I try to start it with a display set to dereel. But I think that's a known bug: it doesn't understand X properly, and only works on local servers.
| Monday, 16 January 2012 | Dereel | Images for 16 January 2012 |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
|
linux.conf.au in Ballarat
|
Topic: technology, opinion | Link here |
Today was the first day of linux.conf.au, in Ballarat. I was involved in the organization at the beginning, but I dropped out, and I didn't go to the conference. Why not? Somehow it's difficult to accept, but my conference days are over. And times are changing. I think what really broke it for me was the discussion of the Code of Conduct. As I said last year, the necessity for something like that turns me off. And now I hear it's printed in the conference handbook. The programme isn't. What does that say?
It seems I'm not the only one. A number of the regular visitors from the past are not going, including Rasmus Lerdorf, who once spent half his life going from one conference to another. We're all getting older—my first LCA was 11 years ago—but I don't think that's all. Times are changing, and this is maybe one indication.
| Tuesday, 17 January 2012 | Dereel | Images for 17 January 2012 |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
|
Building FreeBSD 9.0
|
Topic: technology | Link here |
So now I have finally built all my ports. But FreeBSD 9.0 has just come out, so it seems a good idea to install it before cutting over to 64 bits.
How do you check out a new branch from the FreeBSD Subversion repository? I had it written down somewhere, a “brain dump” by Peter Wemm. But where did I put it? I couldn't find it in my HOWTOs (though it's there now). Went looking on the FreeBSD web site. Couldn't find anything there either. Should it be in the handbook? That's for end users, who don't have access to the Subversion repository. Should it be in the Developers' handbook? Arguably, but that's mainly technical stuff, and it's not there either.
Finally Edwin Groothuis told me: http://wiki.freebsd.org/SubversionPrimer. How did he find that? Google. Somehow that's an indication that something's wrong with the navigation on site.
So I checked out and started building. Then I ran into an old friend:
Despite a lot of searching, I hadn't been able to find the cause of that problem. I had had a number of repeats, but then it went away, and I haven't had it at all in the year or so that I've been trying to upgrade to 64 bits. Why has it shown up again now? We had already discussed the problem of environment pollution in regard to ports, so once again I tried to build the system with a clean environment. Success.
So it's pretty clear that it's an environment variable (or just barely possibly a function). But which? Started removing things at random (well, alphabetically), but it's a slow business, and I didn't get finished. But somewhere in the middle of the alphabet I got past the problem, so it must be one of the 20 or so environment variables that I removed on that occasion.
Also tried playing with the new nVidia display card, but didn't get very far: the system doesn't recognize it. Is that the kernel, the card or the BIOS? The BIOS settings are a little funny—by default only one display is enabled, and there's one on the motherboard—so possibly it's that, but since it was a freebie, it's possible that the card itself is defective. More experiments necessary.
|
More network problems
|
Topic: technology | Link here |
My network problems still don't seem to be over. Today had a number of dropouts, and even when I was reconnected, I only got a GPRS connection, which is so slow that you couldn't tell the difference from being disconnected. But the RSSI was showing 16, which is 16 to 18 dB better than I normally have. Popped the modem and it reconnected with normal signal strength and HSPA. It looks like I had been connected to a tower with only GPRS (there's one to the east somewhere). Is this the sticky result of a failure on the correct tower, or is it the modem itself? How can I tell?
| Wednesday, 18 January 2012 | Dereel | Images for 18 January 2012 |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
|
The build failures: caught
|
Topic: technology, opinion | Link here |
Continued this morning looking for the cause of yesterday's build failure. It's obvious once it's been found:
That's something I put in my environment about 3 years ago to enable me to have only one Makefile in the parent directory of my daily photos. And it seems that this one Makefile didn't reset it, so it went looking for files in the wrong directory. Remove it and all is well.
Well, almost all:
=== grog@dereel (/dev/pts/4) ~/Photos/20080906 8 -> unset MAKEFLAGS
=== grog@dereel (/dev/pts/4) ~/Photos/20080906 9 -> make web
So this kludge was a little too general. How do I Do It Right? There are many ways to fix things for the photos:
Run make -I .. web
Run make -f ../Makefile web
Create a link to the parent Makefile in the current directory.
Create an alias mp equating to make -f ../Makefile.
None of them seem particularly elegant, though I'm tending to the last. I'm left with a feeling that I've forgotten some make magic that would do it better, but I can't find anything obvious in the documentation.
|
Shut down Wikipedia... to the idiots
|
Topic: technology, opinion | Link here |
The protests against SOPA are coming to a head: a number of prominent sites, in particular English language Wikipedia, went offline at 16:00 local time today (05:00 UTC):
|
|
||||||||
Or did they? The whole thing is implemented in JavaScript. Disable JavaScript and at least normal lookups are as good as they ever were, though I could imagine that editing pages would pose a problem. But why did they do it this way? Surely they knew this. My best bet is that they're leaving a back door for people “in the know” to get in anyway, in the assumption that people stupid enough to think that SOPA and PIPA will work will also not be able to work out how to access the site. Or maybe they're trying to show that you can circumvent just about anything.
And the effect of the protest? I fear it won't work. The people who don't understand the web won't miss it as much as those who do. And it's the kind of protest you can't do too often. I fear that, one way or another, we're in for a rough time until law-makers learn what the Internet is.
| Friday, 20 January 2012 | Dereel | Images for 20 January 2012 |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
|
Environment pollution
|
Topic: technology, opinion | Link here |
I'm still playing around with my amd64 system, though soon I should start to cut over. One of the issues I've had for some time is that newer versions of Emacs don't respect the settings that I have been using for years, and use fonts that produce a window far larger than the screen. Of course I should look for the reason, and it's certainly part of what I need to do before I can consider my installation complete, but for the time being I've just been manually resizing the window.
And today that didn't work! It wouldn't let me make it any smaller, though I could have enlarged it. But why did that happen? People on IRC suspect my environment, and possibly that's the case; again, when I ran it with a shell with no environment, it worked. But it still uses this GTK thing, which I don't really like, and Peter Jeremy provided me with his defaults, which don't use GTK and seem to work better, and which didn't display the problem, so I'll go with that.
The whole thing makes you wonder about the value of environment variables. I still haven't worked out a good way to handle the MAKEFLAGS issue. I should probably investigate the Makefiles and fix it; in the meantime, the idea of an alias has proved to be workable, but very inconvenient. I keep typing make instead of mp. If I can't fix the build, it would be easier to remove the MAKEFLAGS before building.
| Saturday, 21 January 2012 | Dereel | Images for 21 January 2012 |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
|
Still more disk problems
|
Topic: technology | Link here |
After my photo processing today, started a backup. Or I tried:
=== root@defake (/dev/pts/0) /home/grog 2 -> mount /dev/ada0p1 /photobackups
That's mount's inimitable way of saying “look at /var/log/messages to see what's up”. As I feared, it said:
This is an eSATA drive. Did I forget to umount it before disconnecting it? Started fsck, which failed with unreadable sectors starting at 128. It also vomited all over the kernel message buffer and the log file:
As the messages said, there was no repeating: the device entry had been removed. So I powered down the disk, rebooted defake and tried again. Same thing. Were the sectors really not readable?
=== root@defake (/dev/pts/0) /home/grog 5 -> dd if=/dev/ada0p1 of=/dev/null bs=128k count=100 conv=noerror
=== root@defake (/dev/pts/0) /home/grog 6 ->
So there appear to be two hardware faults on the disk, and after the second it wasn't able to continue. So much for the increased reliability of eSATA. And this disk had 1.2 TB of data on it. It's all recoverable, but the time! Copying the data over 100 Mb/s Ethernet will take about 36 hours. That's for tomorrow, though: it was time for dinner.
| Sunday, 22 January 2012 | Dereel | Images for 22 January 2012 |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
|
Disk recovery
|
Topic: technology | Link here |
Connected yesterday's defective disk to lagoon, Yvonne's computer today, with a USB connection. Yes, it required fsck. No, no other problems. So there's clearly something wrong with the eSATA connection to defake. Is it the newest version of FreeBSD? Or just the fact that fsck was required? The backtraces indicate memory allocation failures, though it looks more like this was a consequence of hardware problems, not the cause of the reported problems. Probably I should look at the driver in more detail. But not today.
| Wednesday, 25 January 2012 | Dereel | Images for 25 January 2012 |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
|
Strange mount problems
|
Topic: technology, opinion | Link here |
Yvonne had trouble booting her machine this morning. On asking, she told me “Could not mount /Photos”. That's the photo disk on my machine, mounted on hers via NFS. And yes, I couldn't get it to mount. In log file I found:
The last two messages repeated every two minutes. But what's that? I haven't seen that before, and the problem persisted even when I restarted mountd. I had to first umount /Photos, and then remount it, and then things worked. It's clearly something to do with the power failure—the first message is from the reboot following restoration of power—but that's about all I'm sure about. What happened at 21:00? That's when my nightly cleanup scripts run, but that doesn't help, since they don't back up that file system. Maybe it was background fsck. Wouldn't it be nice for it to say when it is finished?
|
CFA: We encourage unsafe networking
|
Topic: general, technology, opinion | Link here |
Finally got round to sending in the comments about the CFA web site that I described yesterday, and tried to send it off. What did I get?
|
|
||||||||
I need Facebook to submit comments to a government agency? Are they out of their minds?
Well, not completely. I was given a choice: ““Get Satisfaction”, Google Mail or Twitter. What happened to real email? And why does a government agency do this sort of thing? Considered sending it from my gmail account (yes, I do have one, because some broken sites insist; it just forwards to my real email). And, not surprisingly, I got a message:
|
|
||||||||
I most certainly do not approve; in fact, I disapprove strongly. What business does a government agency have to ask me to share my personal details with a non-government agency I have never heard of? And why is there no warning?
Further investigation shows that the previous page had given me the option to send the email via other means, including normal email. But that's very much the exception, and I would guess that fewer than 0.1% of all correspondents use it. I'm horrified. I suppose I should put in a formal complaint, but that would only be sensible if it was handled by somebody who understood the issues, and it seems amply clear that there are very few such people around in the government (or even banks, for that matter—they of all institutions should know better).
|
More photo processing with DxO
|
Topic: photography, technology, opinion | Link here |
Over the last few days, I've been reprocessing old photos with DxO Optics "Pro", sometimes with spectacular results. But there's a problem: most of the photos I took in the first year with my Olympus E-510 were taken only with JPEG format, and some of the DxO “presets” are intended for raw images only.
Or are they? I tried the “one-shot HDR” preset, supposedly only available for raw images, and found that it works with JPEG as well. Probably not as well as with raw images, but the photos of the barbecue four years ago are amazingly much better:
|
|||||||||||||||||||||||
|
|||||||||||||||||||||||
And that was the series of photos that eventually got me going the way of HDR images. For the time being, I think I can accept that I don't need that any more.
| Thursday, 26 January 2012 | Dereel | Images for 26 January 2012 |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
|
Web sources of garden information
|
Topic: gardening, opinion, technology | Link here |
So we've decided to transplant the citrus trees. How do you do that? My guess is to wait until winter, and then dig out as wide a section of the roots as possible. But what's the web for? https://www.google.com/search?q=transplant%2bgrapefruit%2btree%2b&oq=transplant%2bgrapefruit%2btree should tell me. And indeed, it comes up with http://www.ehow.com/how_5547756_transplant-grapefruit-trees.html:
How to Transplant Grapefruit Trees...
3 Obtain a young grapefruit tree in a container. Grapefruit trees are readily available from many nurseries.
What does that have to do with transplantation? Nothing. But somehow it seems typical of ehow.com. Still, there are other hits, nearly all from forums. The very first one gives no advice at all, just degenerates into a fight between inhabitants of Arizona and California. The others aren't much better.
The issue here seems to be that answers on forums are treated as being as reliable as other answer. In all probability there's a good document there written by somebody who knows what he's doing, but I couldn't find one.
| Top of page | Previous month | Greg's home page | Today's diary entry | Next month | Greg's photos | Copyright information |
