|
| Greg's diary |
| September 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.
| Sunday, 2 September 2012 | Dereel | Images for 2 September 2012 |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
|
More DxO fun
|
Topic: photography, technology, opinion | Link here |
DxO release frequent updates of their DxO Optics "Pro" software, and I generally install them in the hope that something might speed it up. The latest version is 7.5.4, and I installed it a couple of days ago. Faster? Difficult to say: as soon as I try to process anything, all the images disappear from the image browser window, at least on the version I have running in VirtualBox. No message, just dead in the water.
And it's not consistent. In one scenario, the program runs for about a minute, then all files disappear from the Image Browser. The correction preview also disappears. Looking in the directory, I see that for each image a corresponding image has been added:
If I remove all files from this directory and replace them with other files, I still cannot get Optics Pro to display anything, even if I change to a different directory and then return. I need to stop and restart Optics Pro to get anything done, and I still can't set the presets. In some cases I have seen the text dummyNode below the directory name in the directory display, but I can't repeat this.
In another scenario, the images remain, but the correction preview does not occur automatically. I can't explain why the behaviour changes from one attempt to the next.
Looks like a bug to me, so I fought my way through the DxO web site to find the bug report page, which told me to include the files in My documents / DxO Optics Pro v7 log. With only a few minutes of searching I discovered that this referred to the directory \Documents and Settings\grog\My Documents\DxO Optics Pro v7 logs, not quite the same pathname. But it may make it clearer to me that \Documents and Settings is the Microsoft equivalent of /home under Unix.
In that directory there were basically two sets of log files, dopcor.txt and DXOOpticsPro.txt, with several older versions numbered dopcor.0.txt and so on. They had lots of things of interest, like this:
Clearly something has gone wrong, but what? I suppose it's typical of user-friendly programs in the Microsoft space that they don't report the error to the user, who might get frightened.
The reply came back today, telling me that my machine was too wimpy to run this. I replied asking what I should specifically upgrade, and got an answer back pretty quickly:
Continuing to upgrade on a computer (in this case two computers) that just met the minimum requirements to run (as in start-up and not crash) eventually leads to computers that cannot run the software at all. Read "upgrade" as more computations requiring more from the computer.
Try adding some RAM to your problem machine or go back to a prior version that both will run for the time being. Then think about what you want to upgrade to.
That's a lot of help. It's also not convincing: I was running the systems with “only” 2 GB of memory because the Task Manager showed that I wasn't using even that much. It seems that support also doesn't really know what's going on. Maybe I should think about upgrading to something that is less bloated.
| Monday, 3 September 2012 | Dereel | Images for 3 September 2012 |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
|
More DxO experimentation
|
Topic: technology, photography, opinion | Link here |
As it happened, I have just received another 4 GB of memory for eureka, meaning I could replace the 1 GB DIMMs with 2 GB and thus increase total RAM to 8 GB—just what I need to increase the memory size of my VirtualBox machines. So after yesterday's suggestion to increase memory, I put smart back up to 4 GB. No difference. And looking at the task manager, it's not surprising:
|
|
||||||||
Despite their slowness, the DxO processes aren't that big. There seem to be two processes: DopCor.exe, here using 189 MB, and DXOOpticsPro.exe (note that capitalization) using 272 MB. A far cry from 2 GB.
So what is the problem? Yesterday, based on the log messages, I suspected that it might be the network (SMB) file system, so I first copied the files to the local file system and tried like that. No problem (apart from the normal glacial processing speed). For 24 images I needed 48 minutes, 1 second, compared to 50:04 with 2 GB memory. So I tried again with only 1 GB. That did slow things down, but after about 1½ hours it managed to convert them without problems. So: it looks like some file system issue.
And the other things? It seems that DxO introduces new functionality in the second order upgrade releases. With 7.5.1 (I think) they came up with the idea of displaying all the images not once (in the Image Browser tab) but twice, also in the “Process” tab, and that out of sequence:
|
|
||||||||
This shows a run shortly before the end. The images with the green tick have been processed, the one bottom centre is being processed, and the one to the right has yet to be done. But the one to the right of that has already been processed. The file names show further cases where the names are out of sequence. Why do they add this kind of stuff? It's not only unnecessary, but also confusing. I can't find any release notes, and it's not clear how old the manual is, since there's no version number. But should I have to re-read the entire manual on every second level update?
These files with the .dop extension seem similar. They're not directly related to the problems, apparently a new feature (or a change in defaults). In any case, a clear violation of POLA.
| Tuesday, 4 September 2012 | Dereel → Ballarat → Dereel | Images for 4 September 2012 |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
|
Microsoft world: the pain
|
Topic: technology, opinion | Link here |
While at the Friends of the Ballarat Botanical Gardens, Lorraine Powell showed me the proofs of the new “Pictorial History of the Ballarat Botanical Gardens” book. She had it on a USB stick, and she wasn't connected to the net, so we moved it to Genevieve's machine and tried to email it to me. Gmail wouldn't come to the party: it was over 25 MB. OK, what's ftp for? But how do you find it? This machine doesn't even offer to give you a “Command Prompt”: I had to find the “Run” window and start COMMAND.COM manually. And yes, ftp is still there in Microsoft machines, but the on-system firewall blocks outgoing ftp. It offered to open it for me, but I had forgotten to set binary mode, so I said no. Then I set binary mode and tried again. But this time it didn't offer to open the port for me again. Dead in the water.
Why is Microsoft such a pain to use?
| Wednesday, 5 September 2012 | Dereel → Ballarat → Dereel | Images for 5 September 2012 |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
|
Microsoft 8: first impressions
|
Topic: technology, opinion, photography | Link here |
So it looks like the virtual hardware I use to run DxO Optics "Pro" is too wimpy: they prefer 64 bit machines, and clearly I should have multiprocessor support, which my version of Microsoft “Windows” XP doesn't have. But there's a prerelease version of “Windows” 8 available, and I downloaded it a few weeks ago. Time to install.
There's always something confusing about Microsoft. Their view of the world, particularly networking, is just plain bizarre. The first thing I had to do was to enter an email address—why not a normal ID? I'm not sure, but it seems that it might not have been a user ID at all, but an email for registration. And once again it knew better than I do what makes a good password: “Passwords must have at least 8 characters and contain at least two of the following: uppercase letters, lowercase letters, numbers, and symbols”. So I did that and got “The password contains characters that aren't allowed”. Which? Why don't they say? Presumably it was a space. But that's OK: BloodyStupidMicrosoft works fine.
Then it wanted things like the phone number. It knew that +61 is the country code for Australia, and it wanted the rest of the phone number. Not valid, it said. In the end it proved that it wanted the area code without the 0. And then:
WHY can't any of this software give sensible error messages? Tried again and finally got it running, but not before having to enter my birth date and gender (what business is that of Microsoft's?) and entering a Captcha!
|
|
||||||||
What earthly reason is there for that? And there's this funny start screen that everybody's talking about, lots of squares taking up more than the area of the screen. How do I find anything? Fortunately I had read a review in c't, indicating that you need to move the mouse cursor to top right to get a selection of things to do. And finally I discovered that pressing the Alt key would toggle between the start screen and the desktop. With a little more messing around I was able to get the things I needed on the start screen, and after that it started looking like the Microsoft I know. It seems they've just rearranged the window manager.
SMB shares? No obvious way of setting them from the control panel. But I've been there before. Mounting external file systems isn't a control panel feature, it's part of “My Computer” (now just called “Computer”), and things look pretty much the same except that I can't tell the machine to automatically remount on boot. Score 1 for my diary, 0 for Microsoft.
So: another Microsoft “box” up and running. Each seems more difficult to use than the one before.
|
Enfield radiation tower
|
Topic: general, technology | Link here |
On the way home, found somehting that I had been looking for for a while: the Enfield NBN radiation tower:
It wasn't quite where I had been expecting it, but yesterday I looked through the area and saw nothing. Today I noticed it from a few kilometres away, so presumably it has only just gone up. There have been no objections to the tower in Enfield, so it looks as if we might still have another 6 months to go.
|
DxO Optics Pro under Microsoft 8
|
Topic: photography, technology, opinion | Link here |
Back home and set to installing DxO Optics “Pro” on bigpain, the new Microsoft 8 box. What did I get? 64 bit version or 32 bit version? DxO have been advertising the speed advantages of the 64 bit version at some length, but there was only one image to download. Hopefully it includes both variants. Installation went relatively smoothly, but I couldn't activate it: it seems that there's an activation counter somewhere, and I had used it too much.
Still, there's always the 30 day free trial, so tried that. Yes, pretty much the same as before. With 4 CPUs it promised to convert images on average every 30 seconds—pretty much the same as what I would have got with the 32 bit version. All in all nothing obviously better.
| Thursday, 6 September 2012 | Dereel | |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
|
Another power glitch
|
Topic: general, technology | Link here |
Mains power has been relatively reliable lately. The last failure was on 4 April 2012. This evening it looked as if we had another one, but it was the very briefest of fluctuations, and even my bedside clock, usually the first to reset, made it through. But my main machine eureka didn't! And it's on a UPS! What went wrong there? My best bet is that it was a massive power spike, but it didn't hit any of the other UPSs.
| Friday, 7 September 2012 | Dereel | Images for 7 September 2012 |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
|
Rain gauge problems
|
Topic: technology, opinion | Link here |
So I've replaced the rain gauge on my weather station, it has rained and... nothing. What's the problem? This one doesn't seem to have any mechanical issues, and the electronic connection seems to be working: when I mounted it the vibration caused some false “rain”. So what is it? Do I care? It's very inaccurate anyway. But it would be nice to understand the problem.
|
Finding the Emacs screens
|
Topic: technology | Link here |
It's been over two months since I switched from i386 to amd64 (32 bit to 64 bit) FreeBSD, and there are still a number of irritations that I haven't fixed. One is that Emacs windows are positioned outside the display. I haven't found a solution for that; I suspect it's less a FreeBSD issue than an Emacs or X problem. But at least I've found one way to retrieve them when using fvwm2: select them via the WindowList menu, which will bring them back to top left:
|
|
||||||||
|
|
||||||||
I probably knew about this kind of thing decades ago, but I never found much use for it until now. It's also interesting because it gives exact details of the “location” of the window. I had specified 120x85-53+0, which on this screen should have translated to 120x85+1111+0, but instead I got 120x85+2263+0—an offset of 1152 pixels. Is that significant?
| Monday, 10 September 2012 | Dereel | Images for 10 September 2012 |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
|
Lame mouse syndrome returns
|
Topic: technology | Link here |
Into the office this morning to discover that my mouse was limping again. The web has a number of hits for the problem, but nothing that's obvious. Now that I'm running 2 X servers, I was able to confirm that it hit both of them, and that there's no obvious connection with CPU time, though it's possible that some single process might be sucking it. Today I restarted both servers, not without difficulty: another issue is that the mouse is completely inactive when I start X, and I have to do it yet again. This is anything but reliable. Next time I'll try stopping individual processes to see if it improves things.
| Tuesday, 11 September 2012 | Dereel | Images for 11 September 2012 |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
|
Navman: improving user experience
|
Topic: technology, opinion | Link here |
Message from David Corkery of Navman today:
We are currently undertaking Search Engine Optimisation for the NAVMAN website to make the experience even better.
We noticed that you have a link on your site back to www.navman.com.au on the following URL:
http://people.lemis.com/grog/diary-oct2011.phpIt's great that you've done this and we really value the reference you've provided. However, we would like to request a slight change to this existing link, if it's not too much trouble.
The current link text is 'Navman' we would like you to change this to 'Navman GPS'
OK, not a problem. Looking at the entry, it makes sense, so I did it. But he doesn't seem to have read it very carefully:
It proved to be a Navman GPS N196, a model so old that Navman no longer want to know anything about it, not even for map updates (in its turn a good reason to avoid Navman).
Clearly the best way to improve the user experience would be to include information above the device. Have they done so? Not yet. I sent David a reply; let's see if anything happens.
|
X hangs: more insights
|
Topic: technology, opinion | Link here |
One of the positive results of upgrading my computer system is that the new version now recognizes my USB flash card reader, which up to now I have had to use with the old (USB 1.0) Apple. Now I can get much higher speeds without firing up another machine.
Well, almost. Today I put it in the machine, it was recognized, but I got the dreaded mouse hang. Nothing else was wrong: after shooting down and restarting X, everything worked. So is this maybe an issue with the FreeBSD USB subsystem? Certainly it occurs under Linux as well, but maybe that's a different cause for the same symptom.
| Wednesday, 12 September 2012 | Dereel → Ballarat → Dereel | |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
|
More GPS fun
|
Topic: technology, opinion | Link here |
For no particular reason, dragged out my old Garmin GPS II device today, put some batteries in, and left it to find where it was. It took nearly an hour! And yes, it works. But what use is a GPS receiver without navigational aids?
And navigators are now so cheap that it's just not worth thinking about. Today Yvonne bought a new navigator at ALDI for $59. I've been buying every one they have on offer, about twice a year, in the hope that the maps will some day improve. Today might be the day: finally they've discovered Kleins Road, where I live. It's only been there for 100 years or so. But although they're still using the term “Go Cruise”, it's not clear what this means. My best guess is that it's an ALDI trademark, since my oldest (now defunct) one also used that name. The software used Nav N Go, but this one has a completely different user interface, which I first need to learn. That makes it really difficult to compare. More messing around required.
And if that wasn't enough GPS stuff, also got a reply from David Corkery , explaining that the device I had was older than the web site, so there's no mention of it, but that maps are indeed available for it, and that I can use it to trade in for a newer Navman navigator. All well and good, but wouldn't you think they'd put that info on the web site?
| Thursday, 13 September 2012 | Dereel → Ballarat → Dereel | Images for 13 September 2012 |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
|
More power pain
|
Topic: technology | Link here |
Into the office this morning to discover that both eureka and dereel had rebooted last night at 18:16. It appears to have been a power failure, but there was none. But at that time we blew a contact breaker on that circuit, which also supplies the kitchen. How I love underrated Australian power circuits!. Power stayed off for 30 seconds, clearly long enough to kill the UPS. What a pain these things are!
| Friday, 14 September 2012 | Dereel | Images for 14 September 2012 |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
|
More network pain
|
Topic: technology | Link here |
Came into the office this morning to discover that we have been off the network since shortly after midnight. Optus had sent a terminate request:
For some reason the ppp process didn't even try to reconnect, so we didn't get back on the net until 10:30. And my fun wasn't over then. I couldn't send mail: I got a message indicating that there was no space for the message. Subsequent investigation showed that coincidentally I had used up too much space on the remote server (with old, mouldy log files that I hadn't removed), and I needed to tidy that up first. Strangely, these messages didn't make it to the mail log files on either machine.
|
Where is the radiation tower?
|
Topic: technology, general | Link here |
Discussing my planned move of house on IRC today, and Callum Gibson asked whether I'd still be in the range of the radiation tower. The simple answer is “yes”. But looking at that map (which requires me to enter “Dereel” manually), I discover that it has changed in the last 6 months. At that time the map showed the tower (the orange marker with C for Commenced) rather inaccurately placed a little to the north-east of the correct place. We also noted that Chris Yeardley's house is covered, but other parts, including the plot of land she wanted to sell us, aren't. In the second image, the green area at bottom left indicates a lack of coverage:
|
|
||||||||
|
|
||||||||
But now the location has moved to Ferrers Road, and the coverage has been updated:
|
|
||||||||
|
|
||||||||
More interesting, though, is the plot of land we were looking at, to the south-west of Chris' house, now has complete coverage, whereas before there was none at all:
|
|
||||||||
Sent a message to Scott Weston, who replied with the observation that the maps were always inaccurate, quoting the towers in Enfield and Haddon, both of which have been erected, but not where the map shows them. That explains my surprise about the location of the Enfield tower: I had been looking for it where it was marked on the map. But that doesn't explain the updated coverage information.
|
NBN fixed wireless: first impressions
|
Topic: technology, opinion | Link here |
Yet another thing that Scott mentioned is that he now lives in Haddon, Victoria, and since a fortnight ago he has network access via NBN fixed wireless. He's described his experience here. The most interesting thing he has discovered is that the relatively low bandwidth of 12/1 Mb/s is per ISP, of which you can have up to 4. He also mentions rumours of a 25/5 Mb/s link coming in the not-too-distant future, which certainly makes things more interesting. Now if only VCAT would hear this complaint.
| Saturday, 15 September 2012 | Dereel | Images for 15 September 2012 |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
|
Slow photo processing
|
Topic: general, photography, technology | Link here |
House photo day today, with almost ideal conditions—except that braindeath, Chris Yeardley's loaner Microsoft box, appears to have died. I'm not convinced yet, but for today at any rate I had to run DxO Optics "Pro" in a VM. And that takes forever, especially since the latest version of DxO has problems with SMB shares and I had to copy the files physically onto the virtual disk, causing it to overflow. The processing itself was OK, but it took all day, and I still wasn't finished.
| Sunday, 16 September 2012 | Dereel | Images for 16 September 2012 |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
|
Microsoft "Windows" 8 performance and licensing
|
Topic: technology, photography | Link here |
My experience with Microsoft “Windows” XP yesterday was painful, more due to DxO Optics "Pro" than to Microsoft. The (virtual) machine only has a single processor, and it took about 130 seconds per image to process. But the “Windows” 8 installation uses all 4 cores, and DxO has advertised that the 64 bit version is much faster. So I ran that, and indeed VirtualBox showed it was using about 3 CPUs. The result? About 90 seconds per image, an improvement of only 30% with 3 times as much CPU power. Why is that? Virtual machine issues? I'm torn between buying a real, fast box just to run Microsoft, or giving up on DxO.
The license for the test version of “Windows” 8 expires in January. How do they enforce that? While looking at the converted files, discovered that they had a timestamp of last Thursday: that's what the clock was set to. So it seems that nothing is setting the time after sleeping, and it happily processes files with a date in the future. Wouldn't you expect a time-limited system to check that sort of thing?
|
Another dead monitor
|
Topic: technology | Link here |
My Sλmsung SyncMaster 2233SW monitor died today, not for the first time: it just didn't power on. Given that this was a replacement for a monitor that had a similar failure, it seems that this is a generic problem with the model. Is it worth repairing? I'll find out. But it's now 3½ years old, so in all likelihood it means a new monitor.
|
Joining AVI clips
|
Topic: multimedia, technology | Link here |
Yvonne wanted to edit some video clips today for upload to YouTube. They come from the camera in AVI format, and with a bit of pain we were able to extract specific scenes with avidemux2. But why must this software ignore the current working directory and put you into a completely unrelated directory that you used 9 months ago?
That was only the first part, of course. We ended up with 11 scenes which we wanted to convert to 2. Some formats, such as MPEG TS, allow simple concatenation, but AVI doesn't. We were in a bit of a hurry, but the stuff I found on the web wasn't very encouraging. In the end I found two ways: one with avidemux2, the other with mencoder. To do it right, they need to be put into a script, of course. Some day.
There seem to be some issues with the audio that I didn't have time enough to investigate. Probably there's more to be done.
|
ssh POLA violation
|
Topic: technology | Link here |
One of the problems I had with avidemux2 was that it wasn't installed on lagoon, Yvonne's computer, and it's so down-rev that I can't install it. So we ran it on eureka. But she couldn't connect: eureka refused her ssh keys. Regenerated them, but no luck:
What's that? I've never had that before, and it didn't happen when I tried accessing with my credentials. Searched on the web and found a large number of hits, mainly from Linux. But this one related directly to my situation: upgrade from FreeBSD 8 to 9. And it worked, though the answer they gave was truncated. It's as simple as:
=== root@eureka (/dev/pts/23) ~ 31 -> ssh-keygen -t ecdsa -f /etc/ssh/ssh_host_ecdsa_key -N ""
=== root@eureka (/dev/pts/23) ~ 32 -> /etc/rc.d/sshd restart
What puzzles me is that this only affected Yvonne, and not me.
| Monday, 17 September 2012 | Dereel | |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
|
Finding a high-definition monitor
|
Topic: technology, opinion | Link here |
So, do I repair my dead Sλmsung SyncMaster 2233SW monitor, or do I replace it? The first time it was replaced under warranty, but now it's way out of warranty. I could replace it with another 1920×1080 “high definition” monitor, but the writing is on the wall that higher definition is finally on its way. There are a number of surprisingly cheap 2560×1440 27" monitors available from Korea on eBay, and today I spent a lot of time investigating what is available.
There's a lot of information available on the web, of course. It seems that all monitors use the same IPS panel from Lucky Goldstar, and most of them have only DVI inputs. This report investigates the the Achieva QH270, one of the cheapest, and finds it surprisingly good.
After a lot of investigation, I've established:
The monitors are available at prices between (Australian) $280 and about $400. The specs are almost identical.
Some claim On-screen display, but it's not clear what they mean by it. This report describes the controls on his monitor: two buttons for brightness, two buttons for volume, and one on-off button. The image shows corresponding texts in Hangul. Other monitors also have 5 buttons, but they're arrows, suggesting some kind of navigation. I don't trust the item description to really mean what it says.
About the only differences are different claims about the maximum brightness (350 cd/m² or 380 cd/m²), which seem dubious to me, and whether the thing has loudspeakers or not.
It seems that all monitors come with documentation in Korean only.
There are differences in the number of dead pixels that the sellers guarantee. For an additional fee some of them will guarantee zero dead pixels, if you believe them: others make it clear that they just mean zero stuck bright pixels. The actual number that the sellers guarantee varies, the worst being 6 pixels. Others divide the screen horizontally and vertically into 9 sections and guarantee no more than one dead pixel in the middle and a total of 5 in the other 8 sections.
Some offer an HDMI connector as well. It's not immediately obvious that this requires more electronics, which also allow scaling of different display resolutions. These monitors also have a D-Sub connector, and while that might seem archaic, it has one great advantage: you don't need a dual link connection. As I saw years ago, I can drive two monitors with a dual link card via D-Sub, or only one with DVI. So the D-Sub connector looks useful, but it adds $50 to the price.
Wrote a web page comparing what's available right now. I still need to do some thinking before I make my decision.
In passing, it's interesting to notice that, while this resolution (3,686,400 pixels) is higher than anything I've had before, it still has fewer lines than the 2048×1536 CRT monitors I was using 10 years ago. Their 3,145,728 pixel resolution was only lower because of the width, not the height.
| Tuesday, 18 September 2012 | Dereel | Images for 18 September 2012 |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
|
Choosing a monitor, continued
|
Topic: technology | Link here |
More thoughts about a new 2560×1440 monitor today. The main question was: should I buy a version with HDMI and D-Sub connectors or just a basic one with DVI? It's not even clear whether my video cards (nVidia 9500GT) can generate 2560×1440 analogue outputs. My attempts failed, at any rate. In passing, it's interesting to note that I found a mode line for 2304×1728 in my configuration, a resolution of 3,981,312 pixels, 8% more than the 3,686,400 pixels that these monitors will do.
The card itself is OK: the eBay item descriptions all include it in their lists, but that's with DVI, and I can only drive one of them with my cards.
So why not a new card instead? Indeed, it could be cheaper: the HDMI versions cost at least $50 more than the straight DVI versions. But I don't have any slots free. So maybe replace it with a different card that has 2 dual-link interfaces? How do you find out how many links the cards can support? Many look like mine: two dual-link DVI connectors, but many only support one dual-link connection or two single link connections. With Andy Snow's help came up with this page, which shows how few of these specific chips handle two dual-link connections, and clearly there's even more searching to be done.
The other thing is: what would a 2560×1440 display via D-Sub look like? I haven't seen significant degradation at 1920×1080, but we're talking nearly 80% higher frequencies here. In the end I decided that I'll go for the cheaper solution and work out later how to support the fourth monitor. To make things more attractive, found a NEO 270WQ for only USD 299, which eBay conveniently converted for me:
|
|
||||||||
So I bought that. There's only one way to pay: PayPal. And what did PayPal charge me?
That's nearly $13 more than eBay told me, and I couldn't find it out until after I committed to buy! There's something seriously wrong here. Yes, we know that PayPal charges an arm and a leg for foreign currency transactions, but then eBay—really the same company—shouldn't lie about the price. In fact, neither price is plausible: OANDA tells me that at the correct exchange rate for today it should have been $284.36, so eBay is offering unrealistically low exchange rates. Clearly a little difference is understandable, but this is unacceptable. On the more positive side, the same seller was selling the same monitor for Australian $299, so I did save $4.64.
| Wednesday, 19 September 2012 | Dereel | |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
|
Alternative panorama processing sequence
|
Topic: photography, technology, opinion | Link here |
Creating images such as my verandah panorama takes a number of steps: first I take 20 bracketed sets of 5 photos at 1 EV intervals (because my Olympus won't give me 2 EV intervals) at 45° intervals, then I process three of them with align_image_stack and enfuse to a tone-mapped image, and finally I stitch them together with Hugin.
But there's a simpler way: Hugin can do the tone-mapping too. It's not easy to find out how. There are a number of tutorials, but none of them appears to address this particular issue. In fact, there is a tutorial there, with the unlikely name Creating 360° enfused panoramas. And it could do with an overhaul too: apart from going into irrelevant details such as how to correct for fisheye problems, it refers to an old version of Hugin, and things have changed enough that it's difficult to guess which boxes to tick. In particular, at the time the tutorial was written, the Stitcher tab looked like this:
Now it's like this:
|
||||||||||||||||||||||
The layout of the Output section has changed almost beyond recognition. By experiment I've established that the Blended panorama (enfuse) is now called Exposure fused from stacks, as shown in the second image. I'm not sure whether Blended exposure layers was needed at the time the tutorial was written; I only need to check the one box now.
With this selection, Hugin really did align the images and build a tone-mapped result. Here's the comparison, first the old way and then the new way. Running the mouse over either image (preferably after enlarging) shows the alternate image:
|
|||||||||||||||||||||||
|
|||||||||||||||||||||||
Is it worth it? The problem area in this panorama is over the sliding doors on the extreme left, which still end up being underexposed. With the old method I had to do considerable masking to get the best image quality, while the enfused version is out of the box. That's good, too, because the alternative is to apply the same mask to three images at a time, something that I think Hugin still can't do. But on the whole I think that it did a better job. On the other hand, the underexposed and overexposed images made life tougher for the control point detector; typically it found no control points at all between the images in the set with the greatest exposure differences. And the fast panorama preview looked noticeably different from the finished result (this time before postprocessing):
|
||||||||||||||||||||||
|
|||||||||||||||||||||||
The preview suggested I should have lightened the panorama, but in fact it was (marginally) too light already; that's probably why the shadows came out better.
| Thursday, 20 September 2012 | Dereel | Images for 20 September 2012 |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
|
Finding a dual port, dual link DVI display card
|
Topic: technology | Link here |
So now my monitor is on its way:
Then, 40 minutes later:
Looking at the times and the sender email addresses, this appears to be left hand, right hand syndrome, possibly combined with eBay's habit of marking items shipped before they really are.
So I can expect it some time next week. In the meantime, I need to work out how to drive my array of monitors. The new monitor will require one of my display cards by itself, which means I can't drive more than 3. To get more I need a card with two dual DVI ports.
OK, that should be simple enough, shouldn't it? Yes, it should. After over an hour of searching, I still don't know which nVidia cards can do it, nor even which are the current generation. Their web site is a catastrophe. I've found one page that tells me the information for the Quadro FX series. But how old is that? And where's a general product comparison? Their products page wants you to choose a product range before you know anything about it. Why can't they create a real comparison chart based on specifications?
|
Power supplies: more is less
|
Topic: technology, opinion | Link here |
Another thing I have bought is an Antec EA-550 power supply. It costs $115, and I could get a power supply that does the job for $35. But I don't know the efficiency of the el cheapo supply, only that it's under 80%, while the Antec does about 91%. I've been measuring the power consumption of eureka, my main machine, for a week or two now, and it uses between about 170 W idle and 250 W at full load. Assuming 70% efficiency, the power supply is delivering between 120 W and 175 W. To deliver the same power at 90% efficiency would consume 133 and 195 W, a saving of, say, 50 W, or 1.2 kW per day. At $0.25 per kWh, that's $0.30 a day, or $110 a year. Clearly it makes sense to pay more for a more efficient power supply.
In passing, it's interesting to note that Antec's web site appears to be trying to outdo nVidia in web site brokenness. It presents itself to me in incompletely translated German, and again there's no overview. If I select “Products” and „Netzteile“ (“power supplies”), I first have to choose a series, none of which look like EA. But I know the product number, so I searched for it and got this page, which does not contain a link to the product page. Following the Antec - EA-550 Platinum link I got an empty “Support Page”, partially in Chinese. In the end I found the product page by trial and error: it's the “EarthWatts” series.
|
Printing web pages: the pain
|
Topic: technology, opinion | Link here |
We're off to Melbourne on Sunday, and I'll pick up the power supply from CPL on the way. I have the address in my GPS navigator, but to be on the safe side it made sense to print out the location info page.
What a catastrophe! I don't know what it is about this page, but it took me about 10 attempts before I got anything even remotely resembling a copy of the page:
Firefox created a file that was completely illegible. It appears to have chosen its own fonts reminiscent of early dot-matrix printers:
|
|
||||||||
I've seen this before, and it seemed to be related to something in my firefox configuration, but I never found out what. On other profiles it printed correctly; but not today.
So I tried Opera, printing to a file. Even worse:
=== grog@eureka (/dev/pts/13) ~/public_html/Day/20120920 30 -> ghostview opera.ps
=== grog@eureka (/dev/pts/13) ~/public_html/Day/20120920 31 -> file opera.ps
I don't normally use chrome for printing, because it insists in printing only in PDF. But in this case I had little choice, so I printed to a file and converted it to PostScript. That worked, and the text was legible—but the map was missing!
|
||||||||||||||||||||||
On IRC Michael Ralston, the resident troll, suggested using Microsoft “Internet Explorer”. Under the circumstances, that wasn't such a bad idea, so I tried it with my evaluation version of Microsoft “Windows” 8. Of course, I had no printer set up, but IE offered to do it for me—only it didn't see the network printer. I had to go to the Control Panel to set it up. Returning to IE, it still didn't see the printer. I had to restart it, in the process getting into a full-screen display that I couldn't get rid of. It swallowed all the data, then went back to idle without printing. The second time I saw on the printer control panel: “Ignore data”. My best guess is that Microsoft installed an incompatible driver.
Next I tried on Apple. It didn't pretend to recognize a Brother HL2700-CN printer, but it did offer “Generic PostScript”, and that at least printed. But the layout was all wrong, and the map was partially blanked out. The latter seems to have something to with the driver. Spent some time trying to find out out how to get “Safari” to print to a file (hint: it's under the button labelled “PDF”):
|
|
||||||||
The map in the file looked normal enough, though the page was nearly double the width of the paper.
Finally I started a firefox from lagoon, Yvonne's computer, finding a spare display so that firefox wouldn't try to start a local version, and tried printing. Success! For small values of “success”. The left-hand bar was completely missing, but the information I needed was still there.
So: we've been using printers on computers for decades. Surely it should be simpler to print out web pages. I could use a screen grabber and print out the visible part of the page, but surely a print function should be better, not worse.
And yes, I'm sure that the page in question is broken. But that doesn't explain my biggest problem, the first version of the firefox output. I'm baffled.
| Saturday, 22 September 2012 | Dereel | Images for 22 September 2012 |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
|
Network connection: registration hops cease
|
Topic: technology, opinion | Link here |
While investigating the cause of my Internet connection problems last month, I discovered a continual stream of cell hopping every couple of minutes. It continued through times of good and bad connection quality—but this morning I discovered that it had stopped. The last hops were:
Why did it stop? Why did it happen in the first place? The connection quality is still motley, but acceptable.
| Sunday, 23 September 2012 | Dereel → Melbourne → Dereel | Images for 23 September 2012 |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
|
New power hardware
|
Topic: technology, opinion | Link here |
Then to CPL to pick up my power supply and yet another new UPS. An amazing place. A far cry from MSY: glossy, full of showcases, four people on duty doing I know not what. One of them served customers (2 in the 10 minutes we were there), another got the items, and the others sat in one of many offices. Despite the relationships, they were very slow. But I got my goods, and they look like what I wanted.
|
Web page rendering on tablets
|
Topic: photography, technology, opinion | Link here |
While at Ron and Steve's, showed some of my web-based photos. Some came out rotated by 90°, something I've never seen before, such as this one:
|
|||||||||||||||||||||||
The issue appears to be this EXIF tag:
That's a left-over from the way I took the images: camera mounted vertically, stitched together to make a landscape image, such as this one:
|
|||||||||||||||||||||||
I didn't know of any web browser that evaluated the EXIF data. Now it looks as if I'll have to reprocess thousands of images to remove that tag.
|
Cell-hopping starts again
|
Topic: technology | Link here |
Yesterday I noticed that the cell-hopping on my 3G network connection had stopped. For about 29 hours. Now it's back again:
Why? One possibility is that one of the towers was down.
| Monday, 24 September 2012 | Dereel | Images for 24 September 2012 |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
|
Fixing my photos for tablets
|
Topic: photography, technology | Link here |
Yesterday's rotated images on Steve's tablet were cause for concern. On IRC discovered that a number of people could reproduce it, and that it really did come from the Orientation EXIF tag. OK, that's simple enough, but how do I fix it? I had about 106,000 JPEG images to go through. How much traffic would it cost to upload the changes to my external web site? A short test shows that rsync handled the update pretty efficiently. Wrote a little script called disorientate with the following punchline:
Ran that across the photo tree, which ran for several hours, as did the rsync run:
343 MB seems a lot until you consider it's only a little over 3 kB per file.
|
Saturday's photos, continued
|
Topic: photography, technology | Link here |
The activities of the last few days have resulted in a significant backlog of photos to process. I still haven't written a web page for the flower photos of last Sunday, nor the house photos for Saturday. Continued with the latter today. The “garden centre” panorama was done with HDR, and because of the sun I had a number of images with my hands blocking out the sun.
Tried the new method for merging the images—all 66 of them. It found control points for all except one image, a particularly light component image that I was able to just get rid of. But some of the control points were not quite what you'd expect:
|
||||||||||||||||||||||
After removing that kind of error, got quite a good closure. But masking out things was a real pain—how do you apply exactly the same mask to three images?—and without it the result was less than acceptable. Here first the “new” method, then the “old” (with mouseover alternation). My hands are clearly visible at the edges of the “new” one, and it also has green lens flare at bottom left:
|
|||||||||||||||||||||||
|
|||||||||||||||||||||||
So at least for HDR images I'll stitch my panoramas the old way: first create tone-mapped images, then stitch them together. As a result, I'm still not done with Saturday's photos.
|
New monitor
|
Topic: general, technology | Link here |
I've been following the progress of my new monitor for a few days. It was sent with DHL and arrived in Australia on Saturday, after only 38½ hours. That seems better than UPS, though I've never had anything sent from Korea before. It's difficult to know how long it would have taken end to end if it had arrived during the week, but as it was, this morning was the earliest practical delivery date. And indeed we found a notification in the letterbox: to be picked up at Napoleons CPO. They could have delivered it to the door, but I'm sure they have a valid excuse. In any case, headed off to Napoleons after lunch—19.8 km each way—and picked it up. Well packaged and protected with a “Fragile” sticker with arrows pointing to the bottom.
Unpacked it and put it together. It looks good:
|
|||||||||||||||||||||||
The other monitor is the defective 22" monitor that it's replacing. I'm going to have to position it where there are no reflections, but I think the place I had chosen anyway will fit the bill. But that's as far as I could go today: X configuration is going to be fun, and there's also the power supply to replace in eureka, so I'll have to plan some time when I'm less occupied.
| Tuesday, 25 September 2012 | Dereel | Images for 25 September 2012 |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
|
Preparing to install the new monitor
|
Topic: technology | Link here |
My new monitor has been on the table outside the office for over a day now, and I still haven't installed it, much to Michael Ralston's disgust. But I want to have a smooth transition. I'm reminded of this cartoon from xkcd (click to enlarge):
The first thing is what to do with the fourth display in the short term. The obvious thing to do is to connect it to dereel, but for some reason the nvidia driver doesn't work on dereel, something I encountered and ignored months ago:
=== root@dereel (/dev/pts/1) /usr/src 35 -> kldload nvidia
That's what you'd get if the module is already loaded, but it wasn't. The real reason was hidden in the log file:
Reinstalling the driver didn't help, so I decided to rebuild the system, which was in need of it anyway. And that took the rest of the afternoon. Mañana.
| Wednesday, 26 September 2012 | Dereel | Images for 26 September 2012 |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
|
System upgrade: the sharks
|
Topic: technology | Link here |
Continued with my reconfiguration today. After updating the system on dereel, I was able to load the nvidia driver with no further problems, and I got one monitor up and running in native resolution. Jürgen Lock suspected a mismatch between kernel and /sys. That's possible, though I didn't think so, but after rebuilding the system there's no evidence left.
So: the next steps were to replace UPS and power supply and then integrate the new monitor. The UPS was a surprise: I had assumed it was defective, because minor power fluctuations killed eureka 3 weeks ago. But when I disconnected the power to the UPS, it continued to supply power on battery. And I discovered that nerd-gw, which hadn't failed, was on the same UPS. So it seems something else was in play—possibly the power supply I also replaced, combined with inadequate filtering by the UPS. The filtering of the new unit should be better, as should the new power supply.
Next came the power supply. It's smaller than the old one (both in dimensions and power rating), but it's also heavier:
|
|||||||||||||||||||||||
|
|||||||||||||||||||||||
In passing it's interesting that the current and voltage ratings on the old power supply (ostensibly 720 W) add up to only 363.5 W. And Antec's obfuscation continues with the markings on the packaging and the instructions. It's an Antec EA-550, but you won't find “EA-550” anywhere except on the power supply itself; elsewhere it's this silly “EarthWatts” name, which suggests a ground loop to me. It's also interesting that there's a sticker on the packaging: “No Power Cord”. Not a big issue, though you'd think that a high quality power supply wouldn't skimp there. But it doesn't: it did come with a cord, which I didn't need.
Getting at eureka involves moving a monitor, and I needed to move some anyway, so I did it then. Previously I had monitors with 1920 pixels horizontally and 1080, 1200, 1080 and 1080 respectively. I'm replacing the third monitor with the new 2560×1440 model, which would leave no two 1080 monitors adjacent to each other. So decided to rearrange 1080, 1080, 1440 and 1200, the latter (currently) connected to dereel.
Replaced the power supply, connected up the connections and turned on. Machine booted happily, but—no display. Nothing to do with FreeBSD: I didn't even get the power on display. Everything else was normal.
What was the problem? Suggestions ran around on IRC that it was the sharks from the xkcd cartoon. Spent nearly two hours investigating, including removing the second display card, replacing the old power supply, connecting up the auxiliary PCIe power, all to no avail. Then gradually narrowed it down to the DVI—VGA adapters, and in the end to a single incompatibility: the BenQ E2200HD didn't seem to work with the adapter, at least not on boot. Connecting the other monitor as console solved the problem, and for some reason there was no issue starting X on the BenQ, just like I have always done. What a weird bug!
|
Matrix NEO 270WQ monitor: first impressions
|
Topic: technology, opinion | Link here |
After that harrowing experience, I should have known better than to try to set up the new monitor. After all, it has a somewhat restricted interface—no other modes than 2560×1440 will work At All. And I didn't have any mode lines for the device. But of course, egged on by Michael Ralston, I did put it in there. It didn't start well: I pressed on the start button, and nothing happened. Not even when I held it down for a long time.
It took me a while to realize that the buttons are underneath the monitor, and not even very well aligned with the markings. Perfectly normal push switches, and it came on immediately. Ran X -configure. It found the other two monitors, but not the new one.
I was about to put it away for another day, but I thought that it wouldn't do any harm to just try to start X normally. Bingo! It Just Worked. Looking at the log file shows:
I suppose it's typical that the name FRT DIGITAL (DFP-0) doesn't match anything I've seen before. The “No valid modes” lines refer to what I asked for in the configuration file. So it had at least enough EDID information for X to determine what it is. So now the desktop looks like this:
|
|||||||||||||||||||||||
Spent a bit more time tuning the X configuration, but the real question still needs to be answered: how should I use this real estate? I've been running two X servers recently, :0 with four separate displays, and :1 with two displays each spanning two monitors. The latter proved to be inconvenient in normal use, particularly since many programs pop up dialogue boxes in the middle of the display—straddling the two monitors. But maybe I should do that with the two leftmost monitors. For the time being I have more real estate than for a long time, a total of 10,137,600 pixels—still a far cry from the 17,382,400 I had years ago.
Another issue I hadn't expected: how do I orient the monitor? It subtends such a large angle that the edges are pointing away from me. It really should be a concave dish. I suppose I'll have to get used to that, or sit further away.
| Thursday, 27 September 2012 | Dereel | Images for 27 September 2012 |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
|
Power problems not resolved
|
Topic: technology | Link here |
The first power failure had another result: once again, eureka failed immediately, though the UPS showed that it had enough power for 45 minutes, enough to weather the failure completely, as nerd-gw did. So what's causing the failure? It can't be the UPS, and it can't be the power supply. I'm still guessing that it's some kind of transient, but why does it only affect eureka?
|
DxO acceleration
|
Topic: photography, technology, opinion | Link here |
I've been running DxO Optics "Pro" in a Virtual Machine with a prerelease of Microsoft “Windows” 8 for some time now, and haven't been exactly happy with the speed. DxO claim that the 64 bit version is significantly faster than the older 32 bit versions such as the Microsoft XP I was running before. That version was single processor only, and it took a little over two minutes to process an image. You'd expect it to take a little over 30 seconds running on all 4 CPUs. But the new 64 bit version with “Windows” 8 took about 80 seconds per image. Why? It does indeed use all 4 CPUs, though it won't let me select more than one image in parallel—presumably a bug.
And now, suddenly, it's faster! I'm really getting images processed in about 34 seconds, as fast but no faster than you'd expect from XP on 4 CPUs. Why? I have no idea. I've tuned some minor settings, but nothing that would explain that much acceleration. It's still glacially slow, of course—I think the theoretical maximum speed on the fastest generally available CPUs would be about 10 seconds per image, not exactly blindingly fast.
| Friday, 28 September 2012 | Dereel | Images for 28 September 2012 |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
|
Reinventing my X configuration
|
Topic: technology | Link here |
So now I have my new monitor up and running well. That's the easy part. For well over 20 years I've been continually refining my X desktop for my personal taste. For at least 20 of those years it has been a multi-head setup, and I've gradually come to the conclusion that 4 monitors are enough. But now I only have 3 on eureka, and the fourth on dereel proves to be a pain, in particular because it has its own screen saver timer.
So, the first thing should be to find a way to connect a fourth monitor to eureka. There are several options: I have PCI and a PCIe 1x slots free. Could I find a cheap display card that would fit there? In fact, I have an nVidia PCI card that would theoretically be acceptable, but the driver doesn't want to know:
So a newer nVidia card, or a different manufacturer? I also have a SiS PCI card, but it's even older, and I think I bought the nVidia cards because the drivers for the SiS had atrophied, though I can't find any reference in my diary. But low-end display cards aren't expensive, so went looking for one on eBay.
That's like a needle in a haystack. Looking for “PCI” also finds “PCIe” and similar—as far as I can tell, only the latter. Have standard PCI cards died out? Similar considerations apply to PCIe 1x. In the end, gave up for the day without making any progress. There's still the option of finding a normal PCIe 16x display card with two dual-link ports, but that's difficult, and they're more expensive. I've heard that PCIe cards will work in slots with fewer lanes, but it's not clear how well supported that is. I suppose I should try it out.
The other issue was the layout of the three monitors I do have. Unlike the old CRT days, I didn't have to worry about what resolution to choose, but I still need to reconfigure my fvwm2 configuration files to match, and I'm still not sure whether I want to join the two 1920×1080 monitors into a single display. For the moment I'm doing it, but I'm not sure it'll stay that way.
|
X hang bug: more insights
|
Topic: technology | Link here |
While configuring X, Yvonne came with her camera and wanted files read off it. Last time I read the files on eureka, it triggered this horrible X hang bug, where the mouse cursor jumps back and forth between two screens. I'm gradually coming to the conclusion that this could be a FreeBSD bug after all. This time switched to a VTY before inserting the card. No luck. When I returned to X, it hung anyway. But at least it seems that I'm finding a way to reproduce it. Now I suppose I should try with a PS/2 mouse.
| Saturday, 29 September 2012 | Dereel | Images for 29 September 2012 |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
|
RootBSD: keeping it up
|
Topic: technology, opinion | Link here |
Through much of my life, keeping it up has been important. Forty years ago at UNIVAC, it was a particular challenge, because the technology of the day required routine maintenance. But the 1108 was a multiprocessor system, and individual components could be maintained without taking down the entire system. Then Tandem Computers raised the whole thing to an art form, and uptimes of over 12 months were commonplace, limited only by the requirement of software upgrades.
Part of my job at Tandem was to ensure highest uptime, and I'm always very reluctant to reboot a machine if there's any alternative. Yet another reason to hate Microsoft. And the other day Powercor, with help from a defective UPS, managed to take down a machine that had been up for 170 days.
All that pales into insignificance, though, in comparison with my external server at RootBSD:
=== grog (/dev/ttyp0) ~ 2 -> date; uptime
That's four years and a few minutes. The idle process for CPU 0 has also used the longest CPU time I have ever seen:
That's 1249 days of CPU time. People say that this means I'm vulnerable to various security problems. Maybe I am, but I haven't been able to find any.
|
Fun processing photos
|
Topic: photography, technology | Link here |
House photo day again today, and despite the filthy weather decided to do it today rather than tomorrow. But because of the wind took a different approach to the garden centre panorama: instead of HDR images made from sets of 5 exposures, took a single exposure with flash to lighten the relatively close dark areas. A combination of that and the unexplained increase in processing speed DxO Optics "Pro" (now barely 30 seconds per image) meant that I was finished much faster than usual, despite the expected problems with control points due to the wind.
One thing I didn't expect was when masking one of the panoramas. It had a very good fit, but the exposure blending in one area was less than optimal, so I chopped off a corner of one of the images:
And what happened? Suddenly a very bad fit at that point:
|
|
||||||||
Or at least that's what both fast preview and normal preview said. On stitching, everything was OK:
|
|||||||||||||||||||||||
|
|||||||||||||||||||||||
This was with Hugin. I wonder if commercial panorama software does any better.
|
More monitor investigations
|
Topic: technology | Link here |
One of the things that was clear after rearranging my monitors was that :0.1 (later the right half of :0.0) did not have optimal display settings. Went off looking for web sites helping with monitor calibrations, but it wasn't until Andy Farkas reminded me that I found this LCD test site, which is really quite impressive. And it gave the new monitor a clean bill of health: the settings were as good as perfect. The only thing I couldn't check properly was the black level, because it requires a really dark environment. But that, too, seems to be OK.
The most interesting test was the clock and phase test, which on the new monitor worked fine, as it did on the ΛOC monitor, but both BenQ monitors showed significant flickering. This is clearly a disadvantage of the analogue connection. I suppose I should spend some time to see if I can improve things.
|
EDID information for the new monitor
|
Topic: technology | Link here |
The other thing that I had noticed was the EDID information for the new monitor. There's more than reported by the X server, but how do I display it? Went looking and found edid-decode. Installed it and ran it. No output: it just hung. Looked for the documentation. No documentation, anywhere, not even UTSL. The source shows that if started without parameters, it reads from stdin. The first parameter, if supplied, is a file name, and clearly it supplies the EDID information. But where does it come from? No idea. Even minimal documentation would help, but as it is, I really have no idea how to use it.
Looked in the Wikipedia page, which at least told me that the X server will dump the EDID information to the log file if started with the option -logverbose 6. So did that and came up with:
There's some interesting stuff in there, but also obvious errors. There's no way that this display can be nearly 4 years old (2009, week 7).
| Sunday, 30 September 2012 | Dereel | Images for 30 September 2012 |
| Top of page | ||
| previous day |
|
X hang problems: solved!
|
Topic: technology | Link here |
For well over a year I've been complaining about sporadic hangs with X, where the mouse cursor would get “stuck” bouncing between two screens. I've suspected blame on the part of the nVidia driver, the mouse driver, and even the FreeBSD USB stack. Today I got a message from Andrew Hout telling me that the bug had been identified and fixed. Only two weeks ago, as shown by this bug report, which includes a very good summary of the problem, which was in the nVidia driver after all. The latest version was released only earlier this week. And of course it had nothing to do with FreeBSD, as evidenced by the other reports on the web.
So: installed the new driver with portupgrade, which conveniently didn't load the new driver. There's an issue there: since this is a kernel module, it's difficult to know which version is loaded. The information is there, but you need to know where. In my case, after loading the new driver, it reports (at line 100):
The other way to check is with kldstat, which isn't very informative at that level:
=== root@eureka (/dev/pts/10) /usr/ports/x11/nvidia-driver 16 -> kldstat
After unloading and reloading the driver, however, the size changes, and it moves to the end of the list, leaving the previous slot (8) empty:
It's interesting that it's smaller than the old version.
| Top of page | Previous month | Greg's home page | Today's diary entry | Next month | Greg's photos | Copyright information |