|
| Greg's diary |
| July 2011 |
|
|
|||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
|
|
|
||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
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.
| Friday, 1 July 2011 | Dereel | Images for 1 July 2011 |
| Top of page | ||
| next day | ||
| last day |
|
Australia's Urban Broadband Network: not National
|
Topic: technology, opinion | Link here |
One of the projects that the new government brought in when they came to power was the “National Broadband Network”. It was a welcome move: prior to that, a couple of rival companies, first and foremost Telstra, who owns all the copper, had stifled competition and thus progress. By putting the national infrastructure in the hands of a company who did not compete in the retail market, and which would ensure coverage for all of Australia, we could only have advantages.
But that didn't happen. Paradoxically, Australia is one of the most urbanized countries in the world. According to Wikipedia, 89% of the population lives in cities, and over 50% of the population live in the 3 cities Sydney, Melbourne and Brisbane. And it's the urban population, most of whom can already get ADSL or better, who will benefit. The first rollouts of 100 Mb/s fibre have already taken place, mainly to people who don't really care. I can't find the details on the NBN site, but I recall that fibre would be installed to only about 93% of the population—easy when 89% lives in cities. The remainder, making up well over 90% of the land area, don't benefit.
What options are available? In rough (decreasing) order of desirability, they are:
ADSL (or cable). Even in outback communities such as Leigh Creek, you can get good ADSL coverage at the same cost as in the city. It's really, as the government like to call it, “metro-comparable”. All you need is a telephone exchange and an ISP who cares enough to install the hardware.
WiMAX, a dedicated wireless method for accessing the Internet. It's pretty much as fast as ADSL, and only a little more expensive. Unfortunately, the coverage is very limited.
3G networking via the mobile phone network. Available in most areas with mobile phone coverage. Not as reliable as the previous, and considerably more expensive.
Anywhere in Australia you can get satellite, by far the worst option. Satellite communications are bad because of latency, reliability and traffic cost. I've ranted yet again about this issue.
The previous Government introduced a so-called Australian Broadband Guarantee for people who couldn't get ADSL. It recognized the relative quality of WiMAX and satellite, and would only sponsor satellite if WiMAX wasn't available. But that wasn't good enough for the “National” Broadband Network. They have now fixed that by introducing the Interim Satellite Service, which is effectively the old broadband guarantee stripped of everything except satellite. Quoting them, their criteria for “metro comparable” are:
Access to the internet at a peak data speed of at least 512/128 kbps and 3GB per month usage allowance (with no restrictions within these limits on downloads or uploads or usage time);
Costing the consumer, over three years of no more than $2500 including equipment, installation, connection, account establishment, travel costs and ongoing provision of the service; and
The retailer offering the broadband service can install the service within a reasonable period of time.
No mention of traffic. No mention of latency. No mention of reliability. But they've now killed WiMAX stone dead. Or at least, that's the way it looks from that page. Elsewhere they do mention “wireless”, though they don't say what kind. But it seems to me that they're throwing out existing infrastructure to introduce ultra-fast networking in some place and the same old connections for the rest.
One of the reasons I feel so strongly about this, of course, is because I'm directly affected. ADSL or better isn't difficult to install—in the case of Dereel it's a question of installing a local DSLAM, hardly an expensive investment. If the Broadband Network Company wants to earn the term “National”, they should at least introduce something comparable to the Universal Service Obligation for telephones. At the very least, ADSL or better should be available for anybody with a fixed-line telephone. That's not as difficult as some of the things they're doing.
| Sunday, 3 July 2011 | Dereel | Images for 3 July 2011 |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
|
Importing Trapster to iGO 8 GPS navigators
|
Topic: technology | Link here |
Last week Peter Dilley had found many more speed cameras with his BlackBerry than I had with my GPS navigator, and I've been doing some investigation since then. It seems that the most reliable and up-to-date source of data is Trapster. But that's really for Garmin and TomTom navigators, which use different formats.
I've spent a lot of time trying to find out details of the formats that my navigator uses. It's made by Nav N Go, and for a long time I thought it was called iGO 8, but it seems this is somewhat generic term, and my particular version is called iGO My way. It's not clear how much the difference is—maybe it's all in the eyes of the web site or the marketing department—but it doesn't help when you're looking for details.
What I've established is: the map and associated data are stored in a subdirectory content (also, since this is a Microsoft “Windows” CE device, CONTENT), which has the following subdirectories, most of which appear not to be necessary for normal navigation:
Most of the files in these directories have names relating to the country to which they apply. Most of the files are in a binary format that file(1) doesn't recognize.
building contains images for displaying buildings in 3-D mode. This one is interesting because it contains special information about petrol stations for some countries only:
I have no idea what the difference is between files with the extension .3dc and .3dl.
I haven't worked out what dem is good for. On my navigator, I only have a file for Australia.
lang contains language information, I think for the navigation texts. They're in ZIP format, for some countries there is more than one file, and many of the file names contain spaces:
map contains the map data.
poi contains descriptions of Points of Interest.
scheme contains what looks like data for various display styles, again in ZIP format.
skin presumably contains more display style information, though the file names (only two of them) don't give much away beyond the beta status of one of them:
speedcam is the one I want, the speed camera information. I'll describe it in more detail below.
voice contains the voice information. It seems that the TTS files are just relatively small add-ons for the main voice file. For example, the following two describe voice information without and with TTS:
speedcam is a little different from the rest. On my old navigator, I had:
Those are three different binary file formats, and I couldn't find any reliable description of what they are. Some people on the web claim that they (which?) are sqlite files, but sqlite3 doesn't want to know:
My new navigator has slightly different files:
Despite the older date, Australia.spb has identical content. But what's speedcam.txt? It's about the only file I've found that isn't in binary format:
It's clearly information about additional speed traps, and I'm pretty sure that I've established that it's used in addition to the others. And yes, I can find information about this file on the web. It's clearly CSV, and the fields are:
X is the longitude of the camera location.
Y is the latitude of the camera location.
TYPE describes the type of the camera:
SPEED is the speed limit at the camera.
DIRTYPE is an indication of the directions in which the cameras operate. According to the link, we have:
That's not the whole truth, though. I can add 4-way speed cameras too, though I haven't investigated what value gets put in the file. In the files I've seen, this field is always 0.
DIRECTION is a separate direction indication, in degrees from north. This field also always seems to be 0.
Comparing this to Trapster, it seems that Garmin also has a (different) CSV format. But at least I can convert it—almost. The only values that correspond directly with iGO are the geographical coordinates. Garmin doesn't have any values for the last three fields, and it reports different types of traps, each in its own file. But I was able to download the files and convert them with a pair of scripts: one converts the Garmin format to something the navigator will understand, and the other merges the results with the existing speedcam.txt file and removes duplicates. With a bit of testing, it now works. Interestingly, I still don't find as many traps as Peter did. But there are plenty more on side roads, and I think his BlackBerry was rather generous in its interpretation of the speed camera location.
| Monday, 4 July 2011 | Dereel | Images for 4 July 2011 |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
|
More Internet infections
|
Topic: technology | Link here |
Just before leaving for the Dilleys, got a call from Hazel at “E Protection”, a company in North Sydney (aren't they all?), telling me that my computer had been infected by the Internet. I asked for the phone number, but she said only registered users were allowed to have the phone number. Unfortunately I didn't have time to string her along any further, and just hung up. Somehow I feel guilty.
|
Boat people and the web
|
Topic: general, opinion, technology | Link here |
Australia has been a leading country for immigration since the second world war, and to a certain extent immigrants are welcome and often actively solicited. But the bureaucrats have to make the decision, and as in other countries, one thing that greatly lessens your chance of being accepted is to arrive without a visa. I've been there, done that.
But what if you have no choice? The US aggression in the Middle East has created millions of refugees, and in Africa, without US help, ethnic violence has been just as bad. So lots of people arrive in Australia by boat with the “help” of people smugglers. They've had persecution, torture, rape and death behind them. So what do we do with them? We lock them up, spare no expense to keep them there for years, and then maybe send them back. I'm not the only person who thinks that this is inhuman.
In particular in the case of refugees from Iraq, the Australian government has been a cause of this disaster. The least they could do would be to treat them humanely when they arrive. But both sides of politics continually talk about “stopping the boat people”, as if it were the most important issue in Australian politics, and with no clear idea how they're going to do it or what's going to happen to the boat people. I had assumed that most people in Australia would share my disgust with this approach, but it seems that xenophobia is alive and well.
Recently SBS Television did a three-part film
where 6 volunteers did the boat people trip in reverse, all the way back to Iraq and the
Democratic Republic of Congo. Given that they had volunteered to spend nearly a month of
their lives in the process, it's amazing how little they understood of the issues, and
consequently how surprisedshocked they were to find out the conditions that the
refugees had to bear. Hopefully this series will make a difference to more than the 6
people who were immediately involved.
What does this have to do with computers? Very little indeed. Even the text advertising the web link was wrong:
|
|
||||||||
The correct URL is http://sbs.com.au/goback, but they've spelt it in capitals. I thought this might be catering to a bug in Microsoft “Internet Explorer”, so that's what I tried. But no, I just found another bug: the heading “Cannot find server”, when clearly this is a 404 document from the server. And, of course, it's too difficult for the web masters (aren't they all?) at SBS to adapt the 404 document to redirect to the correct page. Clearly the web masters and the programme content masters don't talk to each other.
| Tuesday, 5 July 2011 | Dereel | Images for 5 July 2011 |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
|
Transferring the fbbg zone
|
Topic: technology | Link here |
I've been dragging my heels on the fbbg.org.au web site, at least partially because I don't understand the web-based interfaces. Got hold of the registry key and put myself in the whois data as tech contact, but still couldn't find a way to get hold of the zone file (I was sure there was only one). Yes, I can change the name servers, but I'd have to guess at the content. In itself, that's not difficult, since there wasn't much more there than the web site, but it's untidy. Spent a lot of time looking round the TransACT web interface, which showed me nothing, and spoke to people who know their way round TransACT, all to no avail.
Finally found a number (13 30 61) and called up, first having a bet with some friends that they wouldn't know what a zone file is. I was on hold for 10 minutes, during which time I was informed about a dozen times that there was an outage in Mildura, but at least the hold music was bearable (Mozart horn concerto). Finally connected to a very faint-sounding Tim, who did know what a zone file was, and quickly connected me to Max, who apparently knew that Jenny Burrell was the tech contact, even without looking. I wonder how many sites they host. He told me that they don't normally give the zone files to customers (!), but that he would send it to me.
And indeed I did get the zone file about an hour later, sent from Nathan Fletcher, who had gone to the trouble of sending it not just to the email address registered in whois, but also my real email address. I wonder where he got that, and why he bothered. The zone file contained one RR that I didn't expect, and one that looks Just Plain Wrong:
Presumably the second RR has been mutilated by an overly anxious MUA. The mail RR probably doesn't get used, but if it does, it could have been very difficult to debug if I hadn't included it.
|
More reception problems
|
Topic: multimedia, technology | Link here |
On the whole TV reception hasn't been bad lately, but last night was an exception which might teach me something. On cardid 2 (the PCI tuner) I recorded a programme on SBS 2 from 19:32 to 20:37, and a programme on 7TWO from 20:37 to 00:30. Concurrently on cardid 1 (the USB tuner) I recorded a programme on SBS 1 from 20:27 to 22:00. The recordings on SBS 1 and SBS 2 recoded without any error messages, though I found massive error reports for the SBS 2 recording in the log file—I guess. The log message format doesn't state the process ID, so in general I can't be 100% sure, though in this case it was the first recording of the day, so it's pretty clear that it must have been that.
But the recording itself looked OK, in crass distinction from the recording on 7TWO, which reported more errors than I have seen in a very long time. Why? I've had a number of hypotheses, but this particular one rules some of them out:
Output of the first tuner interfering with the second tuner. This case clearly disproves that. Although the system numbers the cards 1 and 2, from the point of view of the antenna, the first card is card 2, and the output from card 2 goes into card 1. The bad recording on card 2 happened at the same time as the second recording on card 1, so clearly that can't be the case.
Antenna problems, including wiring. At least in this case, this scenario disproves that too.
Problems with the transmitter. I've seen a surprising number of problems with this particular transport stream, which includes PRIME and 7MATE. We've seen general reception problems lately, so for the time being I'm putting my money on this one.
That doesn't mean that all reception problems are the same, of course. But it does suggest that I might gradually consider replacing the third tuner, since I only removed it because of concerns with hypothesis 2.
Looking at these results, there's a fourth hypothesis I hadn't considered: the immediate change of programme at 20:37:
Could it be that this sudden change has something to do with it? On the face of it, no, but I'll keep my eyes open.
| Wednesday, 6 July 2011 | Dereel | |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
|
Migrating the friends' web site
|
Topic: technology, gardening | Link here |
So now I have all I need for migrating the web site of the Friends of the Ballarat Botanical Gardens, so did so. Well, at least we now have two web servers running in parallel, though www.fbbg.org.au is still the old site. It's obvious that the friends are not particularly technical, and the last thing I want to do is to have things go wrong somewhere. But—so far—everything works as it should.
About the only issue is going to be migrating from ftp to scp for updating the site. In itself, that's not a problem, of course, especially since Jenny Burrell, the only person who actually does the updates, uses an Apple. But probably there are some funny graphical ftp clients which people can use, and potentially there's no corresponding scp client. But that's down the track.
|
Recode or transcode?
|
Topic: technology, opinion | Link here |
Peter Jeremy had a query about the term “recode”, which I use for describing the transformation of my MPEG transport streams to program streams. He preferred the term transcode.
Which is right? Both terms are in use, but if you believe Wikipedia, transcoding is a repacking of the data into a different format, while recoding maintains the format. But the examples they give are not very convincing. I need to think about this one, but “recode” sounds more logical to me than “transcode”.
|
Wondo Gonseff lives!
|
Topic: technology, general | Link here |
Decades ago I had a colleague called Wondo Gonseff. He was a good bloke, but he was somehow ashamed of his name and preferred to call himself by an anagram, Geoff Snowdon. I was reminded of him by yet another question about why I call myself groggy, and went googling for him.
He's back! He has come out of the closet and is now on LinkedIn. Hopefully I can reestablish contact.
| Thursday, 7 July 2011 | Dereel | Images for 7 July 2011 |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
|
Backup disk failure and eSATA
|
Topic: technology | Link here |
Yvonne reported boot problems on lagoon, her machine, today. That proved to be a failed mount of the NFS file system /dump: my backup disk (on dereel) has died.
That's nothing that serious; disks die. But it occurred to me that keeping two backups on the same disk is sub-optimal. I really should find a way to alternate between two different disks, like I do with photos, so that I'll have at least one backup if a single disk fails.
Took the drive out, in the process rejumpering the eSATA adapter that I bought last month for external connections. That showed some difference: lots of disk activity, to judge by the disk lights, but no reaction from the operating system. Discussed with Peter Jeremy, who has the same kind of adapter in the same kind of system: that's not what he sees. More playing around needed. Maybe I should try it in a test machine.
Put the dead disk in a test machine, and it came back to life again, long enough for me to back up some of the files across the network. Then it died again, and the next time I tried it didn't stay alive long enough to finish the boot. Possibly a thermal issue. I should get another disk on the local machine and try backing up there, preferably with something like rsync that will take over where I have left off. Mañana.
| Friday, 8 July 2011 | Dereel | Images for 8 July 2011 |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
|
eBay: The power of negative feedback
|
Topic: general, technology, opinion | Link here |
It's been two weeks since I picked up the replacement GPS navigator in Melbourne. At the time, Maurice promised to get me new maps “soon”. I wasn't overly surprised that I hadn't heard back from then, but we needed closure, so I sent them a message, including:
Please supply up-to-date maps and file by CoB on Friday, 15 July. Otherwise I will leave the following negative feedback:
Software and maps seriously out of date, missing files, promised updates missing
That worked. I got a call back from a girl with a very faint voice who effectively confirmed the content of a mail message they had just sent: they've found maps dated January this year, and also up-to-date POI information that allegedly won't crash the application the way it did last month. They've put it on a 4 GB micro SD card and it should go out on Monday. We'll see, but it's amazing what the threat of negative feedback will do.
| Saturday, 9 July 2011 | Dereel | Images for 9 July 2011 |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
|
More TV reception pain
|
Topic: multimedia, technology | Link here |
My TV reception is going through a bad phase again, but it's clearly related to some stations, notably 7TWO. Last night I had the following recordings:
| Start | End | Daisy chain | Number of | |||||||
| time | time | Channel | Tuner | position | recoding errors | |||||
| 20:27 | 22:00 | 2203 (SBS HD) | 1 | 2 | ||||||
| 21:17 | 22:49 | 2002 (ABC1) | 2 | 1 | ||||||
| 00:57 | 03:20 | 2062 (7TWO) | 1 | 2 | 42 | |||||
The recoding errors aren't the only kind I get; there are plenty of messages in the log files like:
The trouble here is that there's nothing to indicate which programme it is if more than one is being recorded. In addition, they seem to be relatively mild, and streams that have produced lots of these messages may recode perfectly well and show no problems on playback. They clearly indicate some kind of subliminal problem, but it's difficult to correlate, so I won't pay too much attention to them.
I didn't record the number of recoding errors for the first two recordings, but there weren't many. But why so many on the third one? I've decided to keep this kind of record of all recordings for the while, with a little more information.
Today was another problem: three concurrent recordings, which meant putting my third tuner in the machine. Then did some test recordings, which came out less well than I would have liked:
| Programme | Start | End | Daisy chain | File | Number of | |||||||||
| name | time | time | Channel | Tuner | position | size (GB) | recoding errors | |||||||
| Tomato | 15:43 | 16:20 | 2062 (7TWO) | 1 | 3 | 1.2 | 36 | |||||||
| Big Sleep | 15:43 | 16:20 | 2080 (GEM) | 2 | 1 | 2.7 | 1 | |||||||
| Among the ruins | 15:43 | 17:00 | 2203 (SBS HD) | 3 | 2 | 5.3 | 1 | |||||||
| Medicine men | 16:26 | 17:30 | 2002 (ABC1) | 1 | 3 | 2.3 | 1 | |||||||
Clearly there's still more pain to endure. I suppose the next step would be to investigate the tuner configuration again. But why should that help? In the meantime, I'll take out the third tuner and see if that makes any difference.
| Sunday, 10 July 2011 | Dereel | Images for 10 July 2011 |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
|
More reception problems
|
Topic: multimedia, technology | Link here |
As planned, backed up my mythconverg database and re-scanned the channels. There seems to have been no change in the reception parameters, but some of the channel names have changed, in some cases subtly: for example, “7MATE ON PRIME” has become “7mate Ballarat”, and “7TWO ON PRIME” has become “7TWO Ballarat”, enough to confuse MythTV. One of the programmes I had wanted to record was “Electric Dreams” on 7TWO, but for some reason MythTV insisted on trying to record it from 7mate. Trying to record manually only added to the problems. The log file shows it trying to record once only, interrupted by the power failure:
But the list of recorded programmes shows a total of four “Electric dreams”, referring to two files. It seems that this is a database problem, and that I didn't really record from two different tuners to the same files. In particular, the restart at 16:04 was immediately after the power was restored, and it's only there once.
The reception itself is no better. Here the results for today, including the interruptions due to the power failures:
| Programme | Start | End | Daisy chain | File | Number of | |||||||||||
| name | time | time | Channel | Tuner | position | size (GB) | recoding errors | |||||||||
| Food Safari | 17:30 | 18:05 | 2203 (SBS HD) | 1 | 3 | 2.5 | 0 | |||||||||
| Gardening Australia | 18:28 | 19:00 | 2002 (ABC1) | 1 | 3 | 1.1 | 6, failed at 93% | |||||||||
| Manhattan | 20:27 | 22:34 | 2022 (ABC2) | 1 | 3 | 3.4 | 26 | |||||||||
| Pirates | 20:27 | 23:50 | 2006 (PRIME) | 2 | 1 | 7.7 | 33, failed at 9% | |||||||||
| Ferpect Crime | 22:32 | 01:00 | 2032 (SBS 2) | 3 | 2 | 3.2 | 0 | |||||||||
| Gabriel | 22:57 | 01:50 | 2088 (GO) | 1 | 3 | 6.4 | 0 | |||||||||
| Minder | 23:53 | 01:30 | 2062 (7TWO) | 2 | 1 | 3.3 | 36 | |||||||||
| Went the day well? | 06:27 | 09:00 | 2080 (GEM) | 1 | 3 | 12 | 0 | |||||||||
| Electric dreams | 13:57 | 15:26 | 2063 (7mate) | 1 | 3 | 5.6 | 28, failed at 40% | |||||||||
| Electric dreams (2) | 16:04 | 17:00 | 2063 (7mate) | 1 | 3 | 5.6 | 17, failed at 37% | |||||||||
| Clash of the Titans | 14:20 | 15:26 | 2080 (GEM) | 2 | 1 | 5 | 0 | |||||||||
| PSA | 19:27 | 21:00 | 2063 (7mate) | 1 | 3 | 6.9 | 13 | |||||||||
| Nin's Brother | 21:27 | 22:30 | 2203 (SBS HD) | 1 | 3 | 0.7 | Failed at start | |||||||||
So, once again, I'm at a loss. But some of the details are interesting: for example the “Gardening Australia” recording on Saturday evening seemed perfect until shortly before the end. In all probability, the damage is after the end of the programme. And in general, the damage seems to be localized, which suggests interference.
| Monday, 11 July 2011 | Dereel | Images for 11 July 2011 |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
|
Powercor eats computers
|
Topic: technology, general | Link here |
First thing this morning, as I walked past the lounge room, I saw an unusual sight: the disk access light of teevee.lemis.com, the computer that runs the TV projector, was shining brightly: heavy disk access. Never mind that it shouldn't have been on at all—I had not been able to turn it off after yesterday's power failure—why was it accessing disk at all, let alone so heavily?
Some investigation showed a number of depressing things:
It was apparently stuck in an fsck/reboot loop. This is a guess, but it fits the situation.
The onboard NIC was dead. I had to put in an old Ethernet card to get it back on the net.
In the process, ran fsck manually (thus also removing the chance of proving my hypothesis about the reboot loop). But the fsck output seemed to confirm it:
|
|||||||||||||||||||||||
So: yesterday's power failure mortally wounded the motherboard and may have taken a big chunk out of the projector lamp's life. It still works, but we've seen this before. High time to claim all the damages Powercor has done over the years.
It wasn't until some time later that I discovered a whole lot of processor dumps in teevee's /var/crash:
-rw------- 1 root wheel 183701504 Jul 10 18:55 vmcore.10
-rw------- 1 root wheel 183427072 Jul 10 19:16 vmcore.11
-rw------- 1 root wheel 183537664 Jul 10 19:38 vmcore.12
-rw------- 1 root wheel 178221056 Jul 10 19:58 vmcore.13
-rw------- 1 root wheel 183984128 Jul 10 16:21 vmcore.3
-rw------- 1 root wheel 183750656 Jul 10 16:46 vmcore.4
-rw------- 1 root wheel 183603200 Jul 10 17:06 vmcore.5
-rw------- 1 root wheel 181481472 Jul 10 17:29 vmcore.6
-rw------- 1 root wheel 183406592 Jul 10 17:50 vmcore.7
-rw------- 1 root wheel 183844864 Jul 10 18:12 vmcore.8
-rw------- 1 root wheel 183590912 Jul 10 18:33 vmcore.9
Unfortunately this machine doesn't have a kernel with debugging symbols, so I couldn't get much more than this for the first dump (vmcore.3):
(kgdb) bt
#0 0xc07996cc in doadump ()
#1 0xc0799db7 in boot ()
#2 0xc079a089 in panic ()
#3 0xc098be3d in ffs_blkfree ()
#4 0xc099f518 in indir_trunc ()
#5 0xc099f4d7 in indir_trunc ()
#6 0xc099f7e8 in handle_workitem_freeblocks ()
#7 0xc09a08c8 in process_worklist_item ()
#8 0xc09a2372 in softdep_process_worklist ()
#9 0xc09a4d97 in softdep_flush ()
#10 0xc0775929 in fork_exit ()
#11 0xc0a930e0 in fork_trampoline ()
After that, all the remaining dumps failed like this:
(kgdb) bt
#0 0xc07996cc in doadump ()
#1 0x00000000 in ?? ()
#2 0xe4a3aa08 in ?? ()
#3 0xc0805a4c in bread ()
Previous frame inner to this frame (corrupt stack?)
That's consistent with my comments at the time. There was indeed a reboot loop, and it went on for over 12 hours. The dumps stopped at 19:58 not because the system recovered, but because the disk was full:
Jul 23 13:33:31 teevee savecore: reboot after panic: ffs_blkfree: freeing free block
Jul 23 13:33:31 teevee savecore: no dump, not enough free space on device (108108 available, need 179138)
Jul 23 13:33:31 teevee savecore: unsaved dumps found but not saved
That date isn't the date of the dump; the dump was in the swap partition since 10 July.
|
Apple: insecurity through obscurity
|
Topic: technology, opinion, gardening | Link here |
I'm still working on a way to make the upload of files to the Friends of the Ballarat Botanical Gardens web site both secure and easy for non-computer people. Previously it seems that the files got uploaded via a web interface that at least uses https, but I didn't really want to implement that. What's wrong with good old scp? Jenny Burrell, so far the only person who updates the web site, knows more about computers than most gardeners, and fortunately she uses an Apple, so scp should be available.
But it seems that Jenny didn't use this interface after all. As she writes,
The ftp program I use is Fetch 5.6 on Mac OS 10.6.7.
I didn't know that fetch had been ported to Mac OS X—it wasn't on my machine. But it doesn't upload; it only downloads. And what's that “5.6”? Checked and found that there's a commercial product called FETCH that does upload—and it costs $29. And it does some secure versions of ftp that I had barely heard of, including SFTP, which proves to be short for SSH File Transfer Protocol. Clearly that's secure enough, but just requires me to set it up. But looking through the feature list mentions nothing of public key cryptography: maybe it only supports SFTP with password authentication.
But why? Isn't scp on the machine? It is on mine. So I asked Jenny, and knowing the vagaries of Apple's misnamed “Finder”, asked her to try starting it from a shell. But she didn't do that:
I still can't locate scp on this computer with a Spotlight search. Perhaps they took it out or replaced it with something else. I've looked in the system folder in Sharing and Security.
OK, so “Spotlight” isn't Finder, but it belongs to the same problem set. I asked her again to start a shell, and sure enough, it's there.
So how is any Apple user to know that? There's no documentation, and the GUI that floats somewhere above the system doesn't want any contact. Yes, you can go in with “Finder” and select the “folder” /usr/bin, page down through a lot of grey icons, and find it. But who wants to do that? And what's the point of putting software on a machine if nobody can find it? Once again, I'm completely baffled by the way Apple does things.
|
Improving firefox
|
Topic: technology, opinion | Link here |
Anybody who reads this diary could be excused for believing that I'm an old fart, out of touch with reality, who refuses to accept progress. Sometimes I wonder if they're not right. My complaints about firefox are an example: why would anybody want to let the window manager manage firefox's windows when firefox, not a window manager, offers tabs, a less useful substitute?
I've been using tabs for several weeks now, and firefox is much more stable for it. It also uses less memory, which I have suggested is related. It's also a pain, though I'm coming to terms with it. But today Daniel O'Connor posted the URL of an article on one aspect of the new firefox:
... the folks at Mozilla uncovered what was starting to look like a solution for Firefox's Achilles heel, resource usage.
So it is known. It's just that, so far, nobody cared enough. Score one point for the old farts.
|
Aligning time-sequence images, revisited
|
Topic: photography, technology | Link here |
A few months ago I spent some time trying to align multiple images, without success. Since then there has been a lot of activity on the Google hugin-ptx forum about this kind of topic, so posted my own question (why can't I use email? I don't get any error messages, but it never arrives), and got at least one suggestion. It seems that once again Hugin does different things depending on how you ask it. If you use the “Assistant”, you get seriously sub-optimal results, like the ones that I had in March. In all these groups of three photos below, the first will change to the third when the mouse cursor goes over it:
|
|
||||||||
|
|
|||||||||
|
|
|||||||||
|
|
|||||||||
But if you select “Create control points” from the “Images” menu, and you haven't let the “Assistant” mess around, the results from the optimizer are much better:
|
|
||||||||
|
|||||||||||||||||||||||
|
|||||||||||||||||||||||
|
|||||||||||||||||||||||
That gives better results, but the automatic control point generator makes some bad assumptions: for some reason it selects foliage, which can move, over bricks and walls, which don't. This one should be easy enough to do manually, so I tried that. And the results were terrible!
|
|
||||||||
That proved to be the result of allowing the “Assistant” to mess around with the images first. Although I removed all control points, some information must have remained. After starting Hugin again, things looked much better, but still not as good as the second approach. In particular, when adding control points, I get some significant deviations for points which most definitely correspond:
|
||||||||||||||||||||||
|
|
||||||||
The result of this is a somewhat worse correlation than with automatic control point generation:
|
|
||||||||
|
|||||||||||||||||||||||
|
|||||||||||||||||||||||
|
|||||||||||||||||||||||
In particular, try as I might, the second two images are rotated compared to the first, something that also happened with my very first attempts. I've tried this several times, and it happens every time.
One of the loose ends that's been hanging around here is: how does Hugin know that the images were taken at different focal lengths? From the EXIF data, as it reports in the “Camera and Lens” tab:
|
|
||||||||
But I've already established that the Olympus EXIF data is not very accurate. This particular lens (12-60 mm) appears to have only about 32 values for focal length, so the increments are more than 1 mm (there's no 25 mm setting, for example). If Hugin is basing the sizes on the reported focal length, this could be the problem. But what's the solution?
|
TV reception: back to two tuners
|
Topic: multimedia, technology | Link here |
Clearly my current TV reception isn't good enough. It got worse when I put in the third tuner, though the deterioration applied to all tuners. I don't need a third tuner for some time (only about every couple of weeks), so took it out again and tried some recordings on the problem channels, 7TWO and 7mate. And, indeed, the results were better:
| Programme | Start | End | Daisy chain | File | Number of | |||||||||
| name | time | time | Channel | Tuner | position | size (GB) | recoding errors | |||||||
| Murphy Brown | 14:13 | 15:00 | 2062 (7TWO) | 2 | 1 | 1.7 | 1 | |||||||
| Malcolm & Eddie | 14:13 | 15:00 | 2063 (7mate) | 1 | 2 | 3.5 | 2 | |||||||
So why is that? Is it possible that the tuners aren't doing proper impedance matching, but just hard-wired daisy-chaining through? I could believe that. I'll try with one tuner to see what happens there.
| Tuesday, 12 July 2011 | Dereel | Images for 12 July 2011 |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
|
New navigator data
|
Topic: general, technology | Link here |
As promised, received a 4 GB microSD card today with newer maps:
The serial number seems to be a truncated release date, i.e. 26 January 2011, interesting because it was a public holiday. The data is still pretty inaccurate: Enfield is still a couple of kilometres to the north, and the town centre of Dereel is placed at the junction of the imaginary roads I noted a month ago:
|
|||||||||||||||||||||||
|
|
|||||||||
That was looking south, and the following map is oriented north, but there are still massive errors, as a comparison with the Google map, itself very inaccurate, shows:
|
|||||||||||||||||||||||
|
|||||||||||||||||||||||
|
|
||||||||
In particular, the area shown as Dereel is pretty empty (as my photo above shows), and most of the roads (Elder Street, Russell Street, and the mirrors on the east of the main road) just plain don't exist. Interestingly, “Elder Street” seems to have grown longer since the previous maps. Ferrers Road starts at the east of the main road, and the road on the other side is called Swamp Road. To make up for it, Kleins Road (where I live) and Camms Road (where Peter Dilley lives) are still not on the map. It's clear, though, that the errors are related to those in Google Maps, in particular this egg-shaped area for Dereel.
Still, the map information contains one other hint: studly “WhereiS”. Presumably I could find out whether the maps were up to date there. Went looking there to see what I could find, but of course it's not that simple. Finally found the correct web site, which looked good. But first I had to tell it the model of my GPS receiver—why?—and of course it wasn't there, not even a reference to iGO My way. And without a device, I couldn't find out what the latest maps were. So I guessed some random Uniden device and was told that I had to ask Uniden. Selecting Garmin was not much more help: I could select my country, only one choice: Australia and New Zealand, no choice at all, but the only information I got was that it would cost me money, even before I knew if my maps needed updating or not. What a pain!
The other issue I had was the application crash with certain kinds of POI:
|
|
|||||||||
They sent me new POI data too. But the same old crash. You'd think they'd check this stuff before sending it out. Another message, asking them to fix it and test that it's fixed before sending me another memory card.
|
More image alignment experiments
|
Topic: photography, technology | Link here |
Google Groups has finally agreed to deliver my mail, after somewhat more than 48 hours, and a couple of people have responded on the thread. Some seem contradictory, but there was a bit of a sense there that the focal length could be part of the issue. The focal lengths of the two images were given as 21 mm (44.72°) and 23 mm (41.18°). The EXIF data reports a maximum of one focal length value between these two--maybe. Maybe there's nothing at all between the two values. If Hugin is relying on this information, there's a good chance that it will be wrong. I suppose I could guess at closer focal length or field of view values, but there must be a simpler way.
For the fun of it, I tried changing the focal length spec for the second and third images. I left the first one at 23 mm, but by changing the focal length of the other two. At 22.421 mm, the errors dropped to an average of 0.3 pixel, with a maximum of 0.92 pixel. Wonderful!
Unfortunately, the results didn't look at all wonderful. All three images had a different size—on one occasion. But it wasn't repeatable, and now I have this, which looks pretty much the same as before:
|
|||||||||||||||||||||||
|
|||||||||||||||||||||||
|
|||||||||||||||||||||||
I'm experimenting further, but it's beginning to look as if the Hugin or the control point detectors need to find another way of determining the relationship between the image sizes than relying on the EXIF data.
| Thursday, 14 July 2011 | Dereel | Images for 14 July 2011 |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
|
More insight into reception problems
|
Topic: technology, multimedia | Link here |
More investigation of the recording problems today, for the last few days. It's proving useful to present this data in tabular form; clearly the next thing to do is to store it in a database, so I can do correlations between the file size and the number of errors (an obvious one) or time of day (less obvious). But even without that, it's becoming very clear that I have a problem with the TS that carries the channel 7 programmes (the first three below):
| Programme | Start | End | Daisy chain | File | Number of | |||||||||||
| name | Date | time | time | Channel | Tuner | position | size (GB) | recoding errors | ||||||||
| To Gillian | 11 July 2011 | 0:57 | 3:30 | 2062 (7TWO) | 2 | 1 | 5.5 | 23 | ||||||||
| Dawn Anna | 12 July 2011 | 11:57 | 14:30 | 2006 (PRIME7) | 1 | 2 | 3.9 | Died at beginning | ||||||||
| Once Upon a Mattress | 12 July 2011 | 15:57 | 18:30 | 2062 (7TWO) | 1 | 2 | 5.5 | 10 | ||||||||
| Truth and other lies | 12 July 2011 | 2:37 | 2:35 | 2203 (SBS HD) | 1 | 2 | 8.7 | 0 | ||||||||
| Behind the force | 12 July 2011 | 20:27 | 22:00 | 2203 (SBS HD) | 2 | 1 | 6.5 | 0 | ||||||||
| Köter Rex | 13 July 2011 | 19:27 | 20:30 | 2203 (SBS HD) | 1 | 2 | 3.7 | 1 | ||||||||
| What happens in Vegas | 13 July 2011 | 21:36 | 0:13 | 2088 (GO!) | 1 | 2 | 5.9 | 2 | ||||||||
| Communism | 14 July 2011 | 10:56 | 12:30 | 2002 (ABC1) | 1 | 2 | 3.4 | 0 | ||||||||
| Man about house | 14 July 2011 | 11:57 | 14:30 | 2080 (GEM) | 1 | 2 | 12 | 0 | ||||||||
|
FETCH does key-based authentication
|
Topic: technology | Link here |
Got a reply to my message to FETCH support. Yes, it can do key-based authentication, as a bit of an afterthought: you still need a (dummy) password, which you can store, but you can't store the pass phrase for the key. Or you can start ssh-agent, of course. But how do you do that on an Apple?
|
Working around the last GPS navigator bug
|
Topic: technology, general | Link here |
Finally got the file update that I needed to eliminate this bug on my GPS navigator:
|
|
|
|||||||||
As I had already discovered, the file is stored in an archive branding.zip, and indeed, it's present. So “failed to open” probably refers to a bad language use of the word open, in this case presumably “process”. They sent me a new branding.zip which was considerably smaller than the old one:
Further investigation showed that the file poi_brand.spr was missing. So if it's there, you get the message “can't open”, and the application dies. If it's not there, everything works fine. Now doesn't that make sense?
| Friday, 15 July 2011 | Dereel | Images for 15 July 2011 |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
|
Installing Firefox 5
|
Topic: technology, opinion | Link here |
I've been less than happy with firefox release 4—not that I've ever been very happy with firefox—but it seems to be the only game in town. Were the developers also unhappy with it? Given that firefox 4 was the current released version for all of a month or two, you might get the impression. Still, there are issues that came in with release 4 that I might hope would be gone in release 5, so today set out to install it.
The good news: I haven't seen any problems with release 5 yet, and it's possible than one problem I've had (incorrect repositioning after enlarging photos) might have gone away, though it's too early to be sure. But installing it was a pain! That's the fault of the FreeBSD Ports Collection not of firefox.
It started with the dependency checking: it determined that I should install perl version 5.12.4:
OK, so we need a perl upgrade. Why didn't it happen automatically? Ran portmaster against perl, and got:
That's not one error, it's two. First, it's trying to build the current version, and secondly it's failing. So I tried again to install perl version 5.12.4. Same problem:
Round about here I decided to give up trying to build perl and load it from the latest and greatest binary package from ftp://ftp.freebsd.org/. That reinstalled perl 5.10.3, but on IRC I heard:
That was only part of the problem. It seems that the perl port should have updated /etc/make.conf with a line:
But there was nothing there. And why did the default perl port install such an old version if it had been changed 5 years ago? Why is this version still on the ftp site? Looking further, I see:
The build date for firefox is the beginning of this month, but the version is 3.6.18. And the current version in the Ports Collection is version 5. What's wrong here?
In any case, it proved that the real issue was the missing perl version in /etc/make.conf. Installing the package fixed that, and I was able to rebuild firefox without any further problems. But what a pain! I think I was right years ago when I decided that I should reinstall ports rather than upgrade them.
|
Reception problems: further deterioration
|
Topic: multimedia, technology, opinion | Link here |
Somehow investigating my TV reception problems doesn't make them go away. Today was worse than ever. After restarting the new firefox, I discovered that a recording I had started had only recorded about 500 MB in over an hour. Looking at it confirmed: it was broken beyond recognition. Why? Did a number of tests, with little results except to confirm that something was very wrong. Tested the cabling, but in contrast to earlier attempts, wasn't able to provoke any particular change—with one exception: the USB tuner (number 1) stopped working altogether, and I had to reboot to get it to work. But it did disprove my suspicion that the daisy chain was part of the problem: disconnecting the cable out of the physical first tuner made no difference to the problems. I'm still baffled.
| Programme | Date | Start | End | Daisy chain | File | Number of | ||||||||||
| name | time | Channel | Tuner | position | size (GB) | recoding errors | ||||||||||
| Food Safari | 14 July 2011 | 19:27 | 20:30 | 2203 (SBS HD) | 1 | 2 | 4.4 | 0 | ||||||||
| Crownies | 14 July 2011 | 20:30 | 22:57 | 2002 (ABC1) | 2 | 1 | 5.2 | 3 | ||||||||
| Dangerous Minds | 14 July 2011 | 21:32 | 0:15 | 2062 (7TWO) | 1 | 2 | 5.9 | 0 | ||||||||
| We joined the Navy | 15 July 2011 | 11:57 | 14:30 | 2080 (GEM) | 1 | 2 | 5.4 | failed at start | ||||||||
| The Hills | 15 July 2011 | 13:28 | 14:00 | 2088 (GO!) | 2 | 1 | 1.2 | 9 | ||||||||
| Test recording | 15 July 2011 | 14:00 | 15:00 | 2080 (GEM) | 2 | 1 | 4.6 | 18 | ||||||||
| PBS News | 15 July 2011 | 16:42 | 16:50 | 2203 (SBS HD) | 2 | 1 | 0.6 | 0 | ||||||||
| Kind of Magic | 15 July 2011 | 16:44 | 16:50 | 2002 (ABC1) | 1 | 2 | 0 | file empty | ||||||||
| (reboot) | ||||||||||||||||
| PBS News | 15 July 2011 | 16:54 | 18:00 | 2203 (SBS HD) | 1 | 2 | 4.6 | 8 | ||||||||
| Kind of Magic | 15 July 2011 | 16:54 | 17:08 | 2002 (ABC1) | 2 | 1 | 0.5 | failed at start | ||||||||
| Saturday, 16 July 2011 | Dereel | Images for 16 July 2011 |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
|
Firefox 5: better?
|
Topic: technology, opinion | Link here |
One of the things that I had to do to live with firefox release 4 was to give up using multiple windows, because firefox crashed frequently if I did. What about version 5? I had a couple of cases where it looped for a considerable period of time (in the order of 10 seconds CPU time, frequently using more than 100% of a processor):
On one occasion it went on so long that I had to shoot it down. But so far no crashes, for what that's worth.
| Sunday, 17 July 2011 | Dereel | Images for 17 July 2011 |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
|
Still more reception problems
|
Topic: multimedia, technology | Link here |
The reception problems aren't going away. Here's the latest collection of problems:
| Programme | Date | Start | End | Daisy chain | File | Number of | ||||||||||
| name | time | time | Channel | Tuner | position | size (GB) | recoding errors | |||||||||
| Hustle | 15 July 2011 | 21:30 | 23:03 | 2002 (ABC 1) | 1 | 2 | 3.5 | 0 | ||||||||
| Light Brigade | 16 July 2011 | 1:12 | 4:30 | 2062 (7TWO) | 1 | 2 | 7.2 | 0 | ||||||||
| Food Safari | 16 July 2011 | 17:30 | 18:04 | 2203 (SBS HD) | 1 | 2 | 2.5 | 0 | ||||||||
| Nim's Island | 16 July 2011 | 18:27 | 21:50 | 2006 (Prime 7) | 1 | 2 | 8 | 2 | ||||||||
| Gardening Austraya | 16 July 2011 | 18:29 | 19:00 | 2002 (ABC 1) | 2 | 1 | 1.3 | 0 | ||||||||
| New York | 16 July 2011 | 20:27 | 23:38 | 2022 (ABC 2) | 2 | 1 | 5.4 | 6 | ||||||||
| Alfie | 16 July 2011 | 23:57 | 2:35 | 2005 (SC 10) | 1 | 2 | 5.6 | 32, died after 3.8% | ||||||||
| Britannia Hospital | 17 July 2011 | 1:47 | 4:30 | 2080 (GEM) | 2 | 1 | 12 | 0 | ||||||||
| Miracles | 17 July 2011 | 12:42 | 16:10 | 2062 (7TWO) | 1 | 2 | 7.6 | 0 | ||||||||
| Typeface | 17 July 2011 | 14:56 | 16:31 | 2002 (ABC 1) | 2 | 1 | 3.7 | 5 | ||||||||
Even the ones without recoding errors are now showing problems, like the “Gardening Australia” recording. On the whole, though, it's not too bad, with the exception of “Alfie”, which was unrecognizable. But that was with SC 10, a Transport stream I've had problems with in the past, and which I haven't recorded for a long time. But I wish I could find out what it is. Maybe it's really just marginal signal.
| Monday, 18 July 2011 | Dereel | |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
|
More GPT pain
|
Topic: technology, photography | Link here |
So now I have all three 2 TB disks set up: one as my /Photos file system, and two for backups. Only one problem: the first of the backup disks was still partitioned with MBR. That's a problem not just for consistency, but also because the partition name is different, so I can't mount them from entries in /etc/fstab. That's OK: I still have two other backups (the other one is the old /Photos disk, which still barely has enough space, and which is still online), so I wiped the MBR disk and partitioned with GPT instead.
This time it had other surprises in store for me. Because of the USB problems I have on dereel, I did it on teevee, the TV computer. And it's running an old version of FreeBSD that requires explicit offset and length for the add command. Instead of the first variant below, I needed the second:
That's not difficult, but I'm superstitious, and there's always the potential to get offset or length wrong, so I moved the drive to dereel. And though dereel generally handles other USB disks OK, it didn't want to know about this one. OK, we have more computers. Put it on lagoon, Yvonne's computer, and created the GPT partition and file system. Then I moved it to teevee.
And teevee didn't want to know! It claimed that there was no GEOM on the disk. Repeat, same thing. Finally I gave up and partitioned it on teevee, offset, length and all, and finally got things moving. Over 100 Mb/s Ethernet, that'll take about 24 hours. Hopefully it will still be readable on other machines when we're done.
| Tuesday, 19 July 2011 | Dereel | Images for 19 July 2011 |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
|
Administering the Friends' computers
|
Topic: gardening, technology | Link here |
Somehow spent all day today with the Friends of the Ballarat Botanical Gardens. In the morning, various preparations, mainly writing and trying to get Jenny Burrell's public key. That was yet another case of working around “user-friendly” software.
Basically, all I wanted was for her to generate a pair of RSA keys and send me the public key. Completely straightforward:
And indeed, that worked. Almost. I never got the key. It's presumably sitting around on Jenny's machine waiting for her to configure the MTA. Normally she uses Apple Mail, which apparently doesn't use local MTAs. So I asked her to send it to me using Apple Mail. First tried it out on my own Apple machine. Not simple. Firstly it took me 10 minutes trying to work out how to create a new mail message. About the only thing that looked likely was the button “New” on the main window, but when I tried that I got some message about the certificate from dereel.lemis.com, and no create window. Then I searched all the menus and found nothing.
Tried Apple's non-help. The article on “Create a new message” has all sorts of detailed information, but it doesn't tell you how to start. There was certainly no “Create” button. Finally asked on IRC. Yes, it is the “New” button, since renamed to “New Message”. Why didn't I get a compose window? Maybe it got placed below another window, which Apple seems to like to do. I later checked the menus again, and there was definitely a “New message” selection in File menu (the others are in the “Message” menu, as you might expect). But I could have sworn that it wasn't the first time I looked, and they were definitely the mail menus, and not the “Finder” menus. Either I'm losing my mind, or it's a case of “Now you see me, now you don't”.
But my problems weren't over. In principle, you click on the “Attach” button. And it brought up one of these silly graphic directory displays—without the dot files. So I couldn't select .ssh/id_rsa.pub. Discussed various methods on IRC, in the process discovering that you can open a “dialog” [sic] box by pressing ~ where it asks for a file. That's definitely a great workaround from the GUI, but it wasn't the choice I had already made; instead I got Jenny to copy the file to her home directory with the good old tools that always work:
Then she was able to attach it and send it to me. Isn't Apple nice? I note that they no longer advertise with “The computer you can learn to use in a day”.
Also checked the pH of the potting mix of the Hibiscus rosa-sinensis with the yellow leaves. I had been suspecting it was too alkaline, a typical cause of iron deficiency despite the presence of iron. But it was very acid, round pH 4.5. Could that still be the case? Took a leaf to show to people at the Friends.
In the afternoon to the Gardens, mainly to try out a laser printer in replacement of an ink-jet printer that was reportedly too slow. Also took some left-over Cannas with me for sale there.
The weather wasn't the best, and not many people were there. Found Lorraine Powell on the machine I was supposed to be looking at, and she gave me a very different perspective: the problem with the thing wasn't that it's slow, but that it just frequently prints badly. That proved to be typical ink-jet smudge. The printer doesn't get used much, so we'll probably replace it with a laser printer some time.
The big surprise, though: I had been told that we can't get ADSL here, in the middle of Ballarat, so I had been planning for wireless, though I first wanted to confirm that we can't get ADSL. But what did I find? We have an ADSL connection already! I don't know yet how fast or how expensive, but that's a great advantage.
Also a lot of talking about other details. Nothing's difficult, but it's a new situation for me to cater to people who really don't know much about computers and just want to get the job done.
Also showed the leaf of the Hibiscus rosa-sinensis to Yvonne, who confirmed that it's iron deficiency. Dropped in to Bunnings on the way home and picked up some iron chelate. They're horrendously expensive; hopefully they'll do the job. Also some orchid fertilizer and a new spray unit.
| Wednesday, 20 July 2011 | Dereel | Images for 20 July 2011 |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
|
Redirecting web pages, the right way
|
Topic: technology | Link here |
Yesterday at the Friends Mike Sorrell asked me how I got the nickname Groggy. I went to a computer, brought up my home page, and followed the link to the page. But I got:
|
|
||||||||
Brought it up manually—strangely, that worked—and left the investigation until I got home. Here, everything worked correctly. So I tried with Microsoft “Internet Explorer”. Failure.
Where's the bug? After a bit of investigation, I found that the link was incorrect: grog.html instead of grog.php. In the course of time, I have changed the structure of my web site, and also many pages have changed from a .html to a .php suffix. But of course many of the older URLs have been indexed, so I have a 404 document which tries to redirect to the correct page, and it seems to work. Except for “Internet Explorer”.
Further investigation, with the help of people on IRC who know more about HTTP than I do, showed that the redirect page I was returning was correct, but I was still returning an error code 404. According to the specification, my method should work (here quoted from Wikipedia):
At the HTTP level, a 404 response code is followed by a human-readable "reason phrase". The HTTP specification suggests the phrase "Not Found" and many web servers by default issue an HTML page that includes both the 404 code and the "Not Found" phrase.
A bit of searching came up with this page, showing that, though my page should have worked, there's a better way, and “Internet Explorer” even understands that one: just return an HTTP redirect reply. Once again I have learnt something.
|
Food Safari 13, Groggy 1
|
Topic: cooking, opinion, technology | Link here |
After commenting on the French Food Safari last week, I followed up with a somewhat milder comment on the web site:
Looking at the big firm cheese with the big holes in it at the beginning of the episode, I thought it was (Dutch) Maasdam. But looking more carefully, it appears to be a French copy of Swiss Emmental. Why did you show that? I'm sure it wasn't because the French hosts wanted you to. And I see you still talk about "Speck" as if it were a French word (it's German). The French word is "lard". I find it disappointing that these extreme inaccuracies mar an otherwise interesting series. Greg Lehey
The text was all in one line, despite the empty lines that I had left in the hope of having something like proper text layout, but they were removed. More interestingly, though: the site has “Agreee” and “Disagree” buttons for people to press. And I got only one agreement and 13 disagreements. These were the only disagreements on the page.
OK, I don't write my opinions to solicit public support, but why? Nobody came up with a follow-up (far too difficult, let's just press a button), and it's difficult to see what there is to disagree about. Or was it the production team? I almost hope so: that would mean that they have taken notice, and that can only mean that they'll do more research next time.
In passing, it's interesting to note that http://www.sbs.com.au/shows/frenchfoodsafari has significant rendering problems, not only on FreeBSD, but also under Microsoft:
|
||||||||||||||||||||||
The big orange area should be an image offering to play the video, but it fails here both on firefox and “Internet Explorer”.
| Thursday, 21 July 2011 | Dereel | Images for 21 July 2011 |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
|
Copying DVDs
|
Topic: multimedia, technology | Link here |
Two weeks ago I borrowed a DVD from the Geelong Regional Libraries. teevee.lemis.com, my TV computer, doesn't have a DVD drive, and the machine is located where a DVD drive wouldn't make any sense. So to watch the DVD, I first need to copy it to disk, one of the most primitive operations you can perform with a computer. We've been copying data since computing began. But both the media industry and “clever” programmers have made copying DVDs a minefield.
You can mount a DVD on the computer and access the files with no difficulty. By definition, they're in the subdirectory video_ts:
=== grog@dereel (/dev/pts/6) ~ 66 -> l /cdrom/video_ts/
The files vts_01_?.vob are a total of 5.5 GB long, and clearly the payload of this DVD. They're split into chunks of 1 GB to avoid signed 32 bit overflows. Theoretically you could create a copy like this:
This works as long as the DVD isn't encrypted. But frequently it is, as in this case, and you need libdvdcss to decrypt it. There's a program called vobcopy that performs the copy. It's a little strange in its interface, it goes for the biggest file on the DVD, and it has this remnant of the last century where it won't create a file more than 2 GB in size unless you ask it very nicely. But normally it works.
Not so today:
=== grog@dereel (/dev/pts/23) /src/ISOs 20 -> vobcopy
The “8” in the final message is the numeric value for SIGFPE. What went wrong there? Hard to say. I can view it with mplayer, but that doesn't copy. Instead I need the companion program mencoder. And there the fun starts. What options?
=== grog@dereel (/dev/pts/23) /src/ISOs 24 -> mencoder dvd://1 -o mydvd
So, I need to work out at least one command option. Time to RTFM. But:
=== grog@dereel (/dev/pts/23) /src/ISOs 27 -> man mencoder | wc
At 60 lines per page, the man “page” is 153 pages long! Yes, you read this sort of thing with less, but it's really difficult to read, and it's not always particularly helpful:
lavc (-lavcopts)
Many libavcodec (lavc for short) options are tersely documented. Read the source for full details.
So I went off to the web, where I found a number of pages describing what to do. The most promising seemed to be this tutorial. But what a pain! First you need to run mplayer to find the crop values, then run mencoder twice with lots of coding options. Here the examples without the convenient scrolling boxes which hide half the text:
And that's a copy operation? Things weren't that bad in the bad old days of the 1960s!
Of course, that's only part of the explanation. It's not really a copy, it's transcoding, not what I want. More to the point, though, it doesn't work for me. First, I need to change some of the parameters, and when I do, the combination doesn't seem to be correct. But it's far too slow anyway: each pass runs only slightly faster than playing the DVD in the first place.
Next I found a reference in the mplayer manual, something that I had forgotten about, but which I should have read first. But it, too, talked about transcoding. And I had difficulty deciding what kind of audio I needed, so tried the “copy” option. Here the payload of the script I wrote to automate it:
It worked—but it reduced the size of the image by 80%, and the quality was distinctly worse. Also, it changed the aspect ratio. Tried various other options, and gradually it became clear that if I could copy the audio, I could copy the video:
That worked, and mplayer displayed it in the correct aspect ratio. But the audio was in Italian, probably because that was the last language in the list. OK, I was getting the hang of it:
It's still not perfect: there's some difference between the output file and the contents of a DVD that confuses my play script and causes it to select 4:3 aspect ratio. That's my problem, not mplayer's, since it displays fine with raw mplayer.
But what a pain! Yes, there are hundreds of variables, but it shouldn't be that difficult. In principle, vobcopy should have done the job, but it looks as if the DVD is scratched, and that returned invalid data to vobcopy, which didn't check carefully enough and crashed. I'm sure I'll have to refine things further before I'm happy.
|
New Friends' web site
|
Topic: gardening, technology | Link here |
I haven't got much feedback from the Friends of the Ballarat Botanical Gardens about the web site, so decided that today was the day. The new site is online, and the transition was completely smooth. Now we just need to get round to closing down the old one, which is no longer accessible.
| Friday, 22 July 2011 | Dereel | Images for 22 July 2011 |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
|
Hunting down the reception problems
|
Topic: multimedia, technology | Link here |
There were three things to record concurrently tonight, so put the third tuner back in ceeveear. I've been concerned that this might be part of my ongoing reception problems, so after putting it in I started recording the same programme on SC 10 (the channel with the worst reception, and fortunately also the worst content) on all three tuners at the same time. Playing them back showed clearly that the problems were not confined to one tuner. In this image, the left side of the announcer's forehead has been torn out on all three recordings:
|
||||||||||||||||||||||
|
|
||||||||
That was typical of nearly every such artefact, but it's difficult to keep the players in sync to get better photos. It's almost certainly not the only problem—I've seen issues with SBS in the past where the cause was clearly different, but the effect was the same—but it's fairly convincing in this particular case.
|
Copying DVDs: the simpler way
|
Topic: multimedia, technology | Link here |
Discussing yesterday's efforts to copy DVDs on IRC today. There should be a simpler solution. To quote H. L. Mencken:
For every complex problem, there is a solution that is simple, neat, and wrong.
In this case, Callum Gibson asked why I couldn't just copy the DVD with dd and use mplayer to play the image. I've known for a long time why not: the DVD decryption takes place when reading the data from the DVD, so the data that dd reads is not the data that libdvdcss reads. But of course, there's a simple way to prove or disprove that: try it out. So I did. And it worked.
Why? Is it because I had already used libdvdcss to set the decryption keys, or am I just plain wrong? On the face of it, it's the latter, but I've been bitten in the behind too many times to just accept that. More experimentation required, but not now.
The other suggestion, from Edwin Groothuis, was that if I can play the DVD with mplayer without any complicated options, why not just use mplayer -dumpstream to extract the VOB? Tried that, and that works too. That's so much simpler than the “solutions” I found yesterday:
The only thing to beware of is the message at the end:
That's a clever kiddie showing off. Normal completion, no core dump.
| Monday, 25 July 2011 | Dereel | Images for 25 July 2011 |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
|
Olympus “Viewer” 2
|
Topic: photography, technology | Link here |
Olympus have brought out a new version of “Viewer”, their program for image processing. I hadn't installed the previous version on smart, my Microsoft-in-a-VM, so did that today. But they wanted a machine with between 1 GB and 2 GB of memory, and I've had trouble getting even 1 GB in the past. Today was no different: the maximum memory I could set was a little over 800 MB. Made a copy of the disk to be on the safe side, created a new VM to use it, but wasn't able to start it: the copy had the same UUID as the one from which it was copied, of course, and VirtualBox refused to accept it. Somewhere there's a correct way to do it, but I couldn't be bothered looking.
Running “Viewer” was much the same as before, with the same quirks. But I persevered and managed to get it to convert all the images for last Saturday into TIFF, correcting lens distortion as it went. The whole thing took 2½ hours. Is that a lot or little for 309 photos? And of course, the bug is still there that I can't scroll to the bottom of my photo directories. In general I don't see much difference from the previous version.
|
Hugin development version
|
Topic: photography, technology | Link here |
I've been subscribed to the Hugin development mailing list for a while now. Quite a bit is happening: Yuval Levy, the chief developer, has spent the summer adding lots of new features and at least one bug, and since he's going back to university in the Swiss autumn, it's liable to stay like that for a while. But how to find the bug? On some machines, notably Microsoft, but also some Linux, opening the fast preview window causes the process to loop, and they're still looking for the reason.
Decided to jump in and see if I could see anything. That involved minimal work with Mercurial, not as bad as I had feared. But the configuration failed, and after a bit of work discovered that they had removed some of the files, apparently to fix a Ubuntu build failure, but in the process they broke the FreeBSD build.
Finally got it built by reverting changeset 5090. It runs, and I don't have any problems with the fast preview window. So I might as well stick with it.
| Tuesday, 26 July 2011 | Dereel | Images for 26 July 2011 |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
|
Panoramas with distortion correction
|
Topic: photography, technology | Link here |
After converting my photos with Olympus “Viewer” yesterday, I had 32 GB of images to process. The converted images were TIFF format, which required some changes to my scripts—fewer than I had expected. Spent most of the day playing around with them, in the process noting the good fit of most of the images. Went back and compared with the old ones, and discovered that, although the fit was different, there wasn't a really clear advantage of the new ones. Here a comparison:
| Raw | images | JPEG | images | |||||
| Image | Average | Maximum | Average | Maximum | ||||
| verandah-centre | 2.1 | 15.4 | 1.8 | 5.3 | ||||
| garden-centre | 0.8 | 2.6 | 1.0 | 3.4 | ||||
| garden-path-ne | 0.7 | 2.1 | 0.7 | 2.2 | ||||
| garden-se | 1.0 | 3.9 | 1.1 | 3.2 | ||||
| house-nw | 0.8 | 1.9 | 1.0 | 3.2 | ||||
| garden-n | 0.7 | 2.1 | 0.7 | 2.0 | ||||
| north-view-panorama | 0.8 | 2.8 | 0.8 | 2.6 | ||||
| garden-path-se | 0.8 | 2.4 | 0.8 | 2.2 | ||||
| w-to-house | 0.9 | 2.7 | 0.9 | 3.3 | ||||
| dam-panorama | 1.2 | 3.0 | 1.1 | 3.3 | ||||
| house-ne | 1.0 | 3.6 | 1.3 | 2.7 | ||||
That by itself doesn't say very much, though. In particular, it looks as if the first image, verandah-centre, fitted better the old way, but in fact the new one was noticeably better. Here both of them, first the old one, then the new. Running the cursor over the new one will replace it with the old for a better comparison:
|
|||||||||||||||||||||||
|
|||||||||||||||||||||||
In particular, there's a noticeable discontinuity in the beam at top centre:
|
|
|||||||||
|
|
|||||||||
The match is much better, but the colour fringes remain. In this crop it looks as if the old one has fewer colour fringes, but that is probably due to the choice of image at this point. Clearly “Viewer” doesn't do much correction of chromatic aberration with the lens corrections. About the only advantage of the old one is a somewhat more pleasant overall colour. But that can be improved.
The results of the other images vary. Put them on a separate page for later examination.
|
The captcha to end all captchas
|
Topic: technology | Link here |
I hate captchas, especially the fuzzy ones. But today I found one that completely blew my mind: who can write Φ²?
|
|
||||||||
At first I thought it was a joke, but I filled it out (with some difficulty) and sure enough, it was accepted. Maybe the hk should have been subscripted, but it accepted it without that. But how can they expect people to enter that kind of text?
| Wednesday, 27 July 2011 | Dereel | Images for 27 July 2011 |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
|
Inconclusive image processing
|
Topic: photography, technology | Link here |
Spent a lot more time today trying to process images in different ways and compare them. It took all day and was inconclusive. I suspect that somewhere I have confused the output of two different processes, but it's difficult to tell. Still, I'm spending too much time on this. If I have to sit in front of a computer, I could at least finish my weather station software.
| Thursday, 28 July 2011 | Dereel | Images for 28 July 2011 |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
|
Reception problems narrowed down
|
Topic: multimedia, technology | Link here |
Yvonne back with a tiny “set-top box” (in other words, a TV tuner) from ALDI today. I bought it in the hope that it might shed some light on my continuing TV reception problems, and indeed it did. First I connected it at the end of the antenna daisy chain, behind the two PCI tuners, and checked the reception. Pretty much identical to the problems that I've been having lately: in particular, ABC1 was very poor. Played around with the contacts at the antenna amplifier power connector and it got marginally better, barely enough to suggest that it was an improvement. Connected it directly to the connector, without the computer in between: no difference.
So, there's one thing clear, at least for the time being. The problem isn't the computer. So what remains? Connectors, cables (rats?), the antenna, its mounting or the transmitters. The antenna location is really suboptimal:
|
|||||||||||||||||||||||
There are other alternatives, such as south by the garage:
|
|||||||||||||||||||||||
But would that make such a big difference? Anyway, at the very least I can now refine my search.
| Friday, 29 July 2011 | Dereel | Images for 29 July 2011 |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
|
60 years of commercial computing
|
Topic: technology, history, opinion | Link here |
Last year I noted that it had been 64 years
since ENIAC was first demonstrated. But it
wasn't for another 5 years that computers came out of the closetlaboratory
and UNIVAC (later called UNIVAC I),
confusingly called a “commercial computer” (something later reserved for
non-scientific computers) was “delivered” to the United States Census Bureau. That was just over 60 years ago, and closer examination shows that it wasn't actually
shipped to the Bureau until the following year.
Looking at the intervening time, a couple of other important milestones occurred in a year ending in 1: in 1971, after a third of the intervening time, the first microprocessor was delivered, and in 1981, after half the time, the IBM 5150 was announced. None of these had the decency to happen on the same day, of course, but the average is close enough to today that it seems like a good time to stand back and look at the historical perspective. It's easy to divide it in to 3 periods of 20 years. In this article I'll concentrate on the middle period.
The first 20 years of commercial computing were characterized by extremely fast changes in architecture. UNIVAC had thermionic valves, the equivalent of 9 kB of delay line memory, and an instruction execution time of about 500 µs. Only 3 years later the IBM 704 had up to 36 kB of 12 µs core storage and an instruction execution time of about 24 µs. And while UNIVAC had a “repeat” instruction, the 704 had index registers. The manual of operation looks almost modern.
In the 1960s transistors gradually came to replace valves, core storage (much more of it) became the standard, and computers developed more and more of the features which we now consider modern: in 1961 the Burroughs B5000 had a hardware stack, the Burroughs D825 of 1962 was a 4 CPU SMP machine, and also in 1962 the Atlas introduced virtual memory. By the end of this period, computer architecture was relatively well-defined. But computers were still big and expensive—even “minicomputers” were too expensive for individuals to buy. That changed with the introduction of the microprocessor.
When was the first microprocessor introduced? It depends so much on what you mean. What does “microprocessor” mean? What does “introduce” mean?
My definition of microprocessor is a CPU on a single chip, not the only definition that was in use in the 1970s. My understanding was that the first real microprocessor was the Intel 4004, and it was introduced in 1971. But what does “introduce” mean? I'd consider it to mean that the first systems were built with it. The Wikipedia article went by the first advertisement, which in my book is wrong, though there's a clear date for that: 15 November 1971. But elsewhere there's mention of production quantities being shipped in March 1971. The truth is probably pretty close to 40 years ago today.
This 20 year period was also full of architectural development, and it's convenient to split it into 4 5-year periods:
1971 to 1975: There was amazing development. The 4004 was followed by the Intel 4040 and the Intel 8008 within a year. Unlike the 4 bit processors, the 8008 was almost a real computer. I wrote a program for one in 1975. And in 1974 the Intel 8080 was introduced. Some of the design compromises made for that processor are still with us today.
With the 8080 it became possible to build real computers. That wasn't what the industry had expected: they were all designed for embedded use, and it took the industry a while to reorient themselves. But CP/M changed all that. Microcomputers were built based on the Intel chips, but also the Motorola 6800 and Mostek 6502.
1976 to 1980: Things continued to improve, but the pace dropped. The next obvious step was to 16 bit processors, and Intel introduced the 8086 in 1978. And nobody paid much attention: the 8 bit processors had the software, and the 8086 wasn't compatible enough. Motorola went one better and introduced the 32 bit Motorola 68000 in 1979. I had predicted an IBM/360 on a chip by 1980. That didn't happen, but the complexity of the 68000 was similar, so I felt vindicated.
Still, 16 bit processors were interesting. The big problem was the operating system: the de facto standard was CP/M, and it only ran on the 8080, 8085 and Z-80. They had a 16 bit version under way, but it took forever to appear. Finally I ended up buying a system in mid-1980, from a company called Seattle Computer Products. They had not just the CPU and support boards (S-100 bus) but also an operating system, 86-DOS.
The price was right, but who had ever heard of Seattle Computer Products? I called them up and was connected to their President, whose name I recall at George Paterson (another indication that this wasn't the biggest company in the world), though I note that other reliable documentation refers to Tim Paterson. I asked him what he thought would happen when Digital Research brought out CP/M 86, and he said “We're feeling confident. Big companies have bought in to the system. If I told you how big, you'd know who they are”. It wasn't until mid-1981 (and the next section of this overview) that the IBM 5150 was announced, with PC-DOS, son of 86-DOS, that I understood what he meant.
1981 to 1986 was clearly the time when computers first became accepted by non-geeks. As the documentary makers would say, it changed the face of computing forever. I discovered that when I found that our admin people were using the same kinds of computers that we had thought the preserve of us techies. Technically, though, 16 bits were here to stay, and Unix was being run on various new 32 bit designs. It seemed that every man and his dog was bringing out new processor designs. And they were developed with that hallmark of big computers, virtual memory.
1986 to 1991 showed further refinement and a bit of a shakeup in the microprocessor market. A new architecture came out: RISC, which would solve the problems of the old CISC machines. Consolidation began, and a number of previously successful designs faded away, such as the Motorola 68000 series. But that wasn't a matter of just replacing CISC with RISC: market factors began to play an increasing factor, and the replacement Motorola 88000 machine was never a commercial success.
And round this time microprocessors escaped from microcomputers and started being used in larger machines. The era of discrete CPUs was as good as over. Even big machines like the IBM 360 series got microprocessor-based CPUs. The small computer became the norm, and many large computer companies switched or went out of business.
Looking back 20 years later, it's difficult to see any fundamental change. RISC computing turned out not to be the silver bullet that everybody thought it was, and CISC is still alive and well, even growing. The change from assembler to higher-level languages meant that the only people who had to put up with emetic instruction sets were the compiler writers. Increasingly, users were isolated from the computers, so the biggest factor in the success of a platform was marketing.
On the software side, a serious split has developed: toy operating systems for the illiterate masses, where the mouse replaces the keyboard, and “hobbyist” operating systems like FreeBSD and Linux for enthusiasts—and a growing proportion of the illiterate masses. Despite the significant increase in the number of people in software development, it seems that important technical issues are not solved as well as they once were: the incredible increase in computing power makes it unnecessary. That's why firefox wouldn't run on a Control Data 7600 supercomputer: it doesn't need to.
That doesn't mean that there hasn't been technical progress, of course, but it's mainly just details. Neither hardware architecture nor operating systems have really changed. Recently Apple brought out a new version of Mac OS X. I read an article about it in c't, a magazine I respect. Most of it was a description of how the appearance had changed, and some minor details have been improved. But is the integration between the operating system and the GUI any better? It wasn't mentioned, but I wouldn't have very high hopes. On the other hand, my interface to my computer is not significantly different from the one I had in 1991. I have multiple monitors and a different window manager, and that's about it. Even the monitor resolution peaked about 10 years ago and has been dropping since then.
Processor complexity and performance have increased markedly, of course. The fastest Intel 80386 of 1991 managed about 10 MIPS. The fastest modern processors now are probably 1000 times faster, achieved not just by jacking up the processor clock, but by increased parallelism. They have integrated cache, multiple processors and—the same basic instruction set. And they run in machines that still bear a strong resemblance to the original IBM 5150. Until recently I ran an Intel 80486-based machine, made in 1992, for my brewing system. I stopped only because it failed, and who needs old computers? People give them away on every street corner.
That's not to say that nothing has happened in the last 20 years, of course. But in my book it's no longer a technical revolution. There are a few new developments, but most of them—even the ever-increasing prevalence of embedded devices—based on technology that was in place decades ago. It's the decrease in size, power and price that makes it possible to market things that would have been impossible 20 years ago. That's technical evolution.
| Sunday, 31 July 2011 | Dereel | Images for 31 July 2011 |
| Top of page | ||
| previous day |
|
More panorama experiments
|
Topic: photography, technology | Link here |
As promised, the weather was better today, and I was able to take my weekly “house photos”, a term that I need to improve: they're mainly of the garden. Did a number of experiments today:
During some of my tests last week, I discovered that I could get almost acceptable panoramas with only half the images. The test was on the verandah, which is stitched together from 24 images: two rows at 30° intervals. The trouble with going to 60° is that the height is lessened, and in practice I've found that 45° is just about enough for panoramas taken in landscape mode. Today I tried the only intermediate step, 36°, which gave me a total of 20 images instead of 24. And in the process, discovered that my rotator doesn't seem to start at 0°. With rotations of 30° or 45°, I can start at 0°. With 36° I can't. I wonder if the previous owner took it apart and replaced the components out of alignment.
My current way of making panoramas involves two processing steps: first I merge two or three images taken with different exposures in the same direction to make a tone-mapped image. Then I stitch the tone-mapped images together. The problem there is that on days like today, when there's a lot of wind, the merged images don't match exactly, and leaves and things are clearly duplicated or triplicated.
So: what happens when I give Hugin the task of merging the images? Put in all 60 images and let it work it out? Tried that: yes, Hugin lines them up correctly, but when it comes to stitching them, it runs into some obscure exposure problem. Possibly that's my choice of processing parameters, but for the time being it's back to the old method.
Carried on from the experiments with Olympus “Viewer” that I did last week. Discovered one of the reasons why the images didn't look that much different: yes, “Viewer” can compensate for lens distortions, but you need to tell it twice, both “automatic” and “enable”. I had only had “automatic” set. Also experimented with chromatic aberration, which, strangely, Olympus does manually even with its own lenses. The results certainly looked better.
“Viewer” takes about 30 seconds per image to convert an image, at least with the VM I'm using for the task; others tell me that on a real machine it's much faster. I took 328 images today, which would have taken about 2¾ hours to convert. But I only planned to use 248 of them. I create a list in a file called makejpeg, which I already use to convert only the ones I want.
Added a Makefile target foroly to pick out the same files and store them in a directory 00-Oly (so that I can find the thing at all in the toy directory tree display), which made things a lot easier. Once again I converted to TIFF, probably a bad idea: try as I may I can't get any useful EXIF information in them. “Viewer” offers the choice “Exif-TIFF”, but it lies.
And exiftool has its problems too. It copies the data, but not all of it, at least the way I invoke it. That's even more of a problem than it seems: I use the EXIF data to set the timestamps on the resultant files, and I order them by time. Without correct timestamps they were out of order, which makes panorama stitching more difficult. In addition, the EXIF data didn't include focal length or focal plane diagonal, so I had to type it in every time. I think I should go back to JPEG.
By evening I had stitched two panoramas, made no easier by the fact that there was sunshine again in the photos, and despite the tricks I used, I managed to get a lot of flare. I wonder if there's some issue with Hugin.
|
More reception problems
|
Topic: multimedia, technology | Link here |
So, I had finally decided that my reception problems lie in the antenna system, and I'm wondering whether to approach it myself or get somebody in to do it for me. What speaks against the latter is the suspicion that they won't necessarily do it well enough. It looks as if Barry Robinson positioned the antenna incorrectly when he installed it.
But that doesn't seem to be the only problem. Came in this morning and found one “recording” completely empty, and a second one still “recording”, having stored nothing. Both were on GEM, so it looked like a smoking gun. Tried recording the same programme on another tuner and—it worked. After a lot of further investigation discovered that tuner 1 (the USB stick) had hung itself up. Reboot didn't help. Power cycling didn't help. I had to physically remove the device and replace it, and now it works again. Somehow there's too much flaky hardware out there, and much of it seems to be USB-related.
| Top of page | Previous month | Greg's home page | Today's diary entry | Next month | Greg's photos | Copyright information |
