|
| Greg's diary |
|
||||||||
|
|
|
||||||||||||||||
| Monday, 1 January 2024 | Today's diary | Today's images | top | next | last |
Another new year, and with it the New Year's Letter. In past years I have just sent a brief message with the URL, at least partially because the idea of creating a batch of photos with MIME attachments was too hard. But this year I found out how. My send script now looks something like
for i in `sed < $CHRISTMAS_LIST 's:.*<::; s:>$::' | sort -u`; do
mutt -s "Happy New Year from the Leheys" < Christmas-message -a New-Year.pdf -- $i
done
Christmas-message (a name chosen for hysterical raisins) contains the message with the URL. CHRISTMAS_LIST is a shell variable pointing to the list of recipients. I print the web page to a PDF file and attach it with the rather strange syntax shown. The -- is required after the file names and before the recipient.
So: test with local recipients (Yvonne and me), works. Go live at exactly 11:00 (0:00 UTC). Off it goes, but precious little came through.
Oh:
Jan 1 00:07:05 lax postfix/smtp[82271]: DC75D2810F: to=<groggyhimself@hydra.lemis.com>, relay=mx0.lemis.com[121.200.11.253]:25, delay=1.7, delays=0.56/0/0.55/0.6, dsn=5.4.0, status=bounced (host mx0.lemis.com[121.200.11.253] said: 554 5.4.0 Error: too many hops (in reply to end of DATA command))
One of the side effects of the batch approach was that the messages went out with the From: address groggyhimself@hydra.lemis.com rather than the groggyhimself@lemis.com that I normally use. And this frustrating Postfix configuration requires configuration for every system for which it should accept mail—or at least I haven't found a way round it:
mydestination = $myhostname, localhost.$mydomain, $mydomain,
mail.$mydomain, www.$mydomain, lax.$mydomain, ffm.$mydomain,
aussie-gw.$mydomain, ftp.$mydomain, pasocentral.org,
tiwi.$mydomain, lagoon.$mydomain, hydra.$mydomain, dereel.$mydomain, teevee.$mydomain, localhost
When I fixed that, I found the usual rejections, but this time for different reasons:
<pooruser@bigpond.com>: message size 40440345 exceeds size limit
31457280 of server extmail.bigpond.com[203.42.40.138]
Message size 40440345? What's that? Part of it is the inefficient MIME coding, but it's only a 5 page PDF document! And I had printed it before and checked the size. 4 MB, far too big, of course. So where do the 40 MB come from? But looking at the files, I saw two different ones, printed by different versions of firefox:
-rw-r--r-- 1 grog lemis 29,509,353 1 Jan 13:36 Greg's Happy New Year 2023-2024.pdf
-rw-r--r-- 1 grog lemis 4,055,865 1 Jan 13:37 New-Year.pdf
The first (nicely with special characters to annoy shells) is from the new version 120.0 on hydra, and the second was probably from the old version 49.0. Who says there isn't progress?
And of course the mail was going out at a snail's pace, at least relative to the size of the messages. All first to mx1.lemis.com at 5 Mb/s, all 247 of them. And I don't know how to safely remove queued messages from a Postfix queue. So off they all went, taking a total of 4½ hours, and about half of them were rejected. I can fix that. Mañana. At least it showed that I can keep a 5 Mb/s uplink saturated; I got rates of up to 8 Mb/s, but even that would require 40 seconds per message.
Yesterday's issues with MediathekView were irritating, but solvable, once I found out how to install Java. So today I did that. Then used my old invocation:
=== grog@hydra (/dev/pts/38) /usr/local/MediathekView 123 -> java -DproxySet=true -DproxyHost=allmine.lemis.com -DproxyPort=3128 -jar /home/local/MediathekView/MediathekView.jar
JavaFX wurde nicht im klassenpfad gefunden.
Stellen Sie sicher, dass Sie ein Java JRE ab Version 8 benutzen.
Falls Sie Linux nutzen, installieren Sie das openjfx-Paket ihres Package-Managers,
oder nutzen Sie eine eigene JRE-Installation.
“JavaFX” (whatever that might be) not found in the class path. OK, what does the official start method say?
=== grog@hydra (/dev/pts/38) /usr/local/MediathekView 124 -> MediathekView
. Configuring for non-portable mode
. Programmstart: 2024-01-01T15:50:44.54388937
. Version: 14.0.0
. === Java Information ===
. Vendor: OpenJDK BSD Porting Team
. VMname: OpenJDK 64-Bit Server VM
. Version: 20.0.2
. Runtime Version: 20.0.2+9-1
. Maximum Memory: 97824 MB
. Operating System: FreeBSD
. OS Version: 13.2-STABLE
. OS Arch: amd64
Jan 01, 2024 3:50:44 PM com.sun.javafx.application.PlatformImpl startup
WARNING: Unsupported JavaFX configuration: classes were loaded from 'unnamed module @54422e18'
. OS Dark Mode detection not supported
. OS Available Processors: 32
Exception in thread "Thread-2" java.lang.UnsupportedOperationException: FreeBSD is not supported
at com.sun.javafx.tk.Toolkit.getDefaultToolkit(Toolkit.java:195)
at com.sun.javafx.tk.Toolkit.getToolkit(Toolkit.java:232)
at com.sun.javafx.application.PlatformImpl.startup(PlatformImpl.java:290)
at com.sun.javafx.application.PlatformImpl.startup(PlatformImpl.java:162)
at com.sun.javafx.application.LauncherImpl.startToolkit(LauncherImpl.java:651)
at com.sun.javafx.application.LauncherImpl.launchApplication1(LauncherImpl.java:671)
at com.sun.javafx.application.LauncherImpl.lambda$launchApplication$2(LauncherImpl.java:195)
at java.base/java.lang.Thread.run(Thread.java:1623)
. Verzeichnis Einstellungen: /home/grog/.mediathek3
. Konfig existiert nicht!
. Es gibt kein Backup
. Weder Konfig noch Backup konnte geladen werden!
It started and produced windows wanting all sorts of configuration information (path to VLC-player, for example, which I don't need and which I suspect is misspelt. But after a whole lot more input, it came up with this:
|
|
|||||||||
“You're running this program on an unsupported operating system. Stopping the program”.
What a cop-out! I thought the whole point of Java was to be platform independent. Looking back at the stack trace above, I see the message
Exception in thread "Thread-2" java.lang.UnsupportedOperationException: FreeBSD is not supported
Is that from Java or from MediathekView? In either case it's annoying, especially since I've already established that the people who wrote it aren't interested.
Yesterday I discovered that MediathekView throughput under Microsoft was round 10 times as fast as under FreeBSD. I recalled having done some half-hearted tuning attempts, and discovered that I had done some ten years ago. At the time I had issues with the RSP, which is probably why I didn't follow them up. But clearly now I should investigate again.
I still haven't done anything about the presumed scam that we fell into trying to buy riding boots last week. There's a bare chance, despite all indications, that they're legitimate, and for $50 I'm not sure I want any more pain.
But today, just before going to bed, I had another as a result. DigitalOcean sent me a mail saying that my credit card had been refused. Yes, of course: I had locked it. Off to the web site, fought my way through the billing pages (which claimed that I owed $0.00) to find the right page, and changed the credit card number.
Apart from the appalling web site, all went well. But the bill was nearly 4 times as high as usual. I get 1 TB free per month, but this month, it seems, I used a total of 3.5 TB, incurring an additional charge of USD 25.07. Where did that come from? It corresponds to an average traffic rate of 1.31 MB/s. If this continues, I should check what's causing it.
| Tuesday, 2 January 2024 | Today's diary | Today's images | top | previous | next | last |
Took my first photos of the New Year with my Olympus E-30 yesterday. Processed them. The time didn't show.
What's wrong there? Did I forget to set the clock? No, the time was set correctly, and it had chosen a file name based on the date (101 in the second to fourth place in the file name), but:
=== grog@eureka (/dev/pts/6) ~/Photos/20240101/orig 14 -> exiftool 51012647.ORF|grep Date
File Modification Date/Time : 2008:01:01 10:00:00+11:00
File Access Date/Time : 2024:01:03 11:29:53+11:00
File Inode Change Date/Time : 2024:01:02 10:39:43+11:00
Modify Date : 0000:00:00 00:00:00
Date/Time Original : 0000:00:00 00:00:00
Create Date : 0000:00:00 00:00:00
The E-30 was released in 2008. Can it be that somewhere internally there's a 4 bit year counter?
Bad weather forecast today: heavy rain and thunderstorms. That says one thing to me: grid power failures. To be on the safe side, charged up the batteries of the PV system to 95%.
And sure enough, at 13:15 we had a complaint from a couple of these toy UPS systems that I bought a year ago. Nothing further.
But then at 13:33 I came back into my office to find eureka displaying the first screen of the BIOS setup. Apart from the obvious conclusion that there had been a power failure that the PV system didn't handle, that's funny; normally it doesn't stop there.
While I was watching it, it reset again. And again. OK, nothing to do now. Let's wait until Powercor gets its act together.
That started only 6 minutes later, but the course of the afternoon I received a bewildering series of contradictory messages:
13:39 (only 6 minutes later). SMS from Powercod, restoration expected at 16:30:
|
|
|||||||||
13:46: No outages known, and my postcode is invalid:
|
|
|||||||||
14:32: Their map shows an expected restoration of 15:30:
|
|
|||||||||
At some point my SMS got updated without getting a new send time, but at 14:55 I saw:
Ugh! In my experience, yes, it will be restored before then, but if not, it's time to drag out the candles. Though we removed all heavy power consumers, the sky was so dark that most of the time we had no PV power at all, and it seemed unlikely that we could hold out that long.
I kept an eye on the inverter display, of course; it shows a red LED when the grid is out. At 16:30, the LED was out: we had power again! It proved to have come back at 16:12 after a brief hiccough (red line):
|
|
|||||||||
At 16:40, I received another SMS:
|
|
|||||||||
That is so typical of Powercor. We have power (and based on the grid topology I'm sure that all of STONES ROAD DEREEL had too), but half an hour after restoration they claimed that it would take another 7 hours!
Finally, at 18:10, they caught up with reality:
|
|
|||||||||
This is really amazing, even for Powercor. While accessing their information, I was asked to fill out a survey about the information. Since I had nothing to do, I painstakingly went through it, even entering text on this horrible glass keyboard. And I had to choose answers that didn't match my experience.
So why did I find eureka stuck in the initial BIOS screen? When the power came back, tried again. No, it wouldn't go any further, and it didn't react to the keyboard. Dead in the water.
OK, from decades ago the sequence is: remove everything connected and see if something is holding it up. Started with the USB connections. Bingo! It came up and started doing normal things.
But I need the USB stuff, so started replacing connections one by one. The 7 port hub was the culprit. Change for the 4 port hub that I had lying around. Still it didn't work.
OK, connect the disks (all 4 of them) directly to the system. No space for camera cables. The first three disks probed fine, but the last (my photo backup disk) didn't show at all. Dead? It powered up happily when I connected it. More elimination, and I discovered good news, bad news: the disks are all OK, but the last USB port is dead.
What happened? The power supply for the USB hub was very hot and smelt burnt. Clearly it was destroyed by a Powercor surge that fought its way through the PV system. And the last USB port? I put the mouse dongle in it, and it probed happily, but didn't work properly, though the mouse stumbled a bit when I tried it. A different USB port? No. Another dongle? Also no reaction. It's not clear why, but it looks like I need a new mouse too.
OK, finally get eureka up and started X. Surprise, surprise. Screen 0, whose monitor is currently not connected, came up on monitor 3, completely disregarding what I specified in the configuration file. And of course all the fonts and things are incorrect, because I was in the process of reconfiguring X (and have been for the past month or two). High time to finally finish the migration to hydra, which is still a fair amount of work.
Done! After only 1½ hours of frustration. And it seems that only eureka went down. Check the NFS mounts on the other system. hydra OK. lagoon OK. tiwi... no response, though rwho told me that it, too, had rebooted.
Into the lounge room to see what was going on. No display! Not again! This drives me mad. Why do I have so much difficulty with this machine deciding which display to use? Dragged in a monitor to the lounge room, pulled tiwi out of its shelf and connected things up. No display.
Of course, these ThinkCentres only even try to display if a monitor was connected at boot time. OK, boot, watch.
Unable to mount dereel:/ on /dereel
Fool! I shut down dereel last week, and I should have checked for file systems mounted from other systems because of the problems I had the following day. Set /dereel to noauto in /etc/fstab and all should be well.
But why did that stop it coming up with a display on the TV (HDMI output)? I don't know. That choice is made during the initial BIOS probe, and NFS mounts come much later. Rebooted, and it confirmed my expectations: it came up on both the TV and the monitor, with a multi-headed X configuration that I hadn't expected. Remove the monitor, reboot again, and all was back to normal. Only 2 hours recovery.
Two hours so far. I don't have a mouse on eureka. Isn't this God's way of telling me that I should finally complete the migration to hydra?
| Wednesday, 3 January 2024 | Today's diary | Today's images | top | previous | next | last |
After yesterday's catastrophic failure, spent most of today documenting the problems that occurred, and recovering from them.
The main issue today was the mouse. I'm sure I had a second Logitech M705 mouse, but I couldn't find it. But I tried out both dongles on hydra, and how about that, one worked. The other was presumably the one that Powercor destroyed. So at least I have one mouse.
I left that on hydra and tried out the other mice and dongles. Two appeared to be the same kind, and probe as
Jan 3 14:38:54 eureka kernel: ugen0.11: <DaKai> at usbus0
Jan 3 14:38:54 eureka kernel: ukbd1: <DaKai 2.4G RX, class 0/0, rev 1.10/3.11, addr 17> on usbus0
Jan 3 14:38:54 eureka kernel: kbd3 at ukbd1
Jan 3 14:38:54 eureka kernel: ums1: <DaKai 2.4G RX, class 0/0, rev 1.10/3.11, addr 17> on usbus0
Jan 3 14:38:54 eureka kernel: ums1: 5 buttons and [XYZT] coordinates ID=1
That looked like they would match the two “Jenkins” mice (an Officeworks brand) that I had lying around, but no, no response. The third probes as:
Jan 3 14:39:37 eureka kernel: ugen0.3: <vendor 0x0430> at usbus0
Jan 3 14:39:37 eureka kernel: ukbd0: <vendor 0x0430 Sun USB Keyboard, class 0/0, rev 2.00/1.04, addr 20> on usbus0
Jan 3 14:39:37 eureka kernel: kbd2 at ukbd0
Jan 3 14:39:37 eureka devd: Executing '/etc/rc.d/syscons setkeyboard /dev/ukbd0'
Jan 3 14:39:37 eureka kernel: ugen0.11: <MOSART Semi.> at usbus0
Jan 3 14:39:37 eureka kernel: ukbd1: <MOSART Semi. 2.4G Keyboard Mouse, class 0/0, rev 1.10/1.02, addr 21> on usbus0
Jan 3 14:39:37 eureka kernel: kbd3 at ukbd1
Jan 3 14:39:37 eureka kernel: ums1: <MOSART Semi. 2.4G Keyboard Mouse, class 0/0, rev 1.10/1.02, addr 21> on usbus0
Jan 3 14:39:37 eureka kernel: ums1: 5 buttons and [XYZT] coordinates ID=3
But I couldn't get any mouse to talk to any of the dongles.
And then there's more: the Logitech mouse has a number of buttons on the side. On eureka, I have it configured so that one of the buttons mapped to button 2. But the same configuration on hydra just didn't work. I have some recollection of seeing this before, but it's just one more thing to chase. I was able to set the simulation with buttons 1 and 3, like I do on tiwi.
And that's about all I had the energy for. I really should finish the X reconfiguration and move the monitors around.
Message from Alan Kennington today, mentioning strange log messages from his mail server:
2024-01-01T20:45:40.804505+11:00 condor postfix/qmgr[28568]: 32A9C120D94: from=<>, size=3436, nrcpt=1 (queue active)
2024-01-01T20:45:41.432611+11:00 condor postfix/smtp[21039]: connect to hydra.lemis.com[192.109.197.129]:25: No route to host
2024-01-01T20:45:41.634015+11:00 condor postfix/smtp[21039]: 32A9C120D94: to=<grog@hydra.lemis.com>, relay=none, delay=17157, delays=17157/0.01/0.62/0, dsn=4.4.1, status=deferred (connect to hydra.lemis.com[192.109.197.129]:25: No route to host)
It was quickly followed up by similar messages from a number of other people.
Oh. hydra shouldn't be visible to the mail system. I had already noticed this shortly after sending it, but of course it affects replies too. But why “no route to host”? This looks like an incorrect error message. Yes, hydra is visible on the net, but it's not listening. The correct message should have been “Connection refused”.
What do I do about it? Add an MX record, of course:
=== grog@freefall (/dev/pts/0) ~ 3 -> host hydra.lemis.com
hydra.lemis.com mail is handled by 20 mx0.lemis.com.
Tested that, and it worked. So presumably the retries should come through, but so far none have.
I've been meaning to send out mail to the recipients who bounced, but clearly there's more work to be done.
| Thursday, 4 January 2024 | Today's diary | Today's images | top | previous | next | last |
So how do I map the side keys on my Logitech M705 mouse? Off searching and came up with this response, which conveniently included a script showing that the replyer has a directory called /temp. With a bit of modification and testing, it works. Put this into my .xinitrc:
# Logitech unwired: find the ID
MOUSEID=$(xinput | grep -m 1 "Logitech USB" | sed 's/^.*id=\([0-9]*\)[ \t].*$/\1/')
# And set middle button to 8 and 9
xinput set-button-map $MOUSEID 1 2 3 4 5 6 7 2 2 10 11 12 13 14 15 16
So at least one problem has been solved. I just wish I understood what good moused is any more. If I don't run it, I can't communicate with the mouse. But mapping flags don't seem to do anything.
I'm almost ready to move my main desktop monitors to hydra; all I need to do is to fix my window manager configuration. But before I got very far with that, I discovered that another issue still hadn't gone away: from time to time the window manager crashes, which initially I blamed on x2x. But I'm not doing that any more, and today the window managers for hydra:0.1 and hydra:0.2 crashed, and nothing I could do would restart them:
=== grog@hydra (/dev/pts/15) ~ 30 -> fvwm2 -s -display :0.2 -f /home/local/X/.fvwm/fvwm2rc-hydra:0.2
[fvwm][FlocaleGetFontSet]: (-*-*-medium-r-*-sans-12-*-*-*-p-*-*-*) Missing font charsets:
ISO8859-5, KOI8-R, ISO8859-7, JISX0208.1983-0, KSC5601.1987-0, GB2312.1980-0, JISX0201.1976-0
Assertion failed: (ret != inval_id), function _XAllocID, file xcb_io.c, line 626.
Abort trap
Are the “missing font” messages trying to tell me something relevant? They seem to crop up all the time, so they're probably “harmless”. Is it maybe related to my configuration files? Nope:
=== grog@hydra (/dev/pts/15) ~/public_html/recipes/videos 33 -> fvwm2 -s -display :0.2
[fvwm.2][CMD_EdgeResistance]: <<DEPRECATED>> The command EdgeResistance with three arguments is obsolete. Please use the following commands instead:
EdgeResistance 250
Style * EdgeMoveDelay 250
Style * EdgeMoveResistance 10
[fvwm][FlocaleGetFontSet]: (-adobe-times-bold-r-*-*-14-*-*-*-*-*-*-*) Missing font charsets:
ISO8859-5, KOI8-R, ISO8859-7, JISX0208.1983-0, KSC5601.1987-0, GB2312.1980-0, JISX0201.1976-0
[fvwm][FlocaleGetFontSet]: (-adobe-times-bold-r-*-*-18-*-*-*-*-*-*-*) Missing font charsets:
ISO8859-5, KOI8-R, ISO8859-7, JISX0208.1983-0, KSC5601.1987-0, GB2312.1980-0, JISX0201.1976-0
Assertion failed: (ret != inval_id), function _XAllocID, file xcb_io.c, line 626.
Abort trap
Where did it get those outdated commands from? Something old and mouldy? Is it maybe a real bug with the current version of fvwm2? Can I use the old version on eureka?
=== grog@eureka (/dev/pts/14) ~ 5 -> fvwm -display hydra:0.2=== grog@eureka (/dev/pts/14) ~ 6 ->
[fvwm][main]: <<ERROR>> can't open display freebie:0.0
[fvwm][main]: <<ERROR>> can't open display freebie:0.1=== grog@eureka (/dev/pts/14) ~ 7 -> DISPLAY=hydra:0.2 fvwm
[fvwm][main]: <<ERROR>> can't open display freebie:0.1
[fvwm][main]: <<ERROR>> can't open display freebie:0.0
Where did that come from? Yes, once I had a system freebie.lemis.com, though I'm not sure it was in this millennium. But where is it getting the name from, and why is it using it?
OK, Google, what's going on here? Came up with surprisingly few hits—only 5, of which the last one was this diary. But this one, which points to this page, suggests that it's not just my problem. And the interesting part was:
The dirty fix for me was to switch from fvwm2 to the developing fvwm3 window manager.
OK, I can do that too. And how about that, it works (sort of) with the same configuration files, and it doesn't even complain about the fonts (nor anything else, for that matter). But it comes up in monochrome, so it's not just a drop-in replacement. Before I continue, I should check if it, too, isn't just waiting to crash.
Over the weekend I bought a new camera. According to Australia Post tracking, it was in Delacombe yesterday afternoon, and it would be delivered on Monday, 4 days' time. Do I believe it? Not a bit. It has only been a month since I last had that kind of claim. But this time was different. A little later I received email:
57 N 03-01-2024 To ebay@lemis.c ( 907) eBay N 📦 ORDER DELIVERED: Olympus PEN E-PM1 Full Spectr…
58 N 03-01-2024 To ebay@lemis.c ( 700) Australia Post N Your parcel is ready for collection from the Post Office
That's in the order in which they arrived. “Delivered” (or “DELIVERED") is clearly an approximate term. But of course it was there, so off and picked it up.
My new camera is an Olympus E-PM1. Why that? I already have one, which by coincidence I received 7 years ago today, and it wasn't a spectacular camera when it was new, over 10 years ago. But this one is special: it has been converted for “full spectrum”, infrared and normal light. That's better than cameras that only see infrared, since I can convert it to that by simply putting an infrared filter on it. And the price was right: $90, where these things normally cost round $400, more than I'm interested in paying. And of course, since it's a Micro Four Thirds system camera, I have plenty of lenses for it. I had been warned that the image stabilization didn't work, but that's not so big a deal.
First thing, of course, was to compare it to my existing E-PM1. Externally, they're just a different colour. And both of them forget the date when the battery is removed. The only real thing is the sensor. Here the old PM1 and then the “new” camera (run the cursor over an image to compare it with its neighbour):
|
|
|||||||||
|
|
|||||||||
It wasn't until I looked at these photos that I saw the scratch on the sensor of the new camera. So far I haven't seen any evidence in the images.
OK, take a photo:
|
|
|||||||||
Looks pretty much like the ones I took with other cameras and an infrared filter, including an image that was still on the SD card of the other E-PM1. But in the meantime I have seen instructions that ask for setting white balance. Bingo!
|
|
|||||||||
By comparison, the other E-PM1 gives, for almost the same view,
|
|
|||||||||
As in so many cases, nothing to write home about. But other foliage looks more interesting:
|
|
|||||||||
Somehow, though, everything seems to be slightly out of focus:
|
|
|||||||||
That was taken at f/1.4 with the Leica Summilux 25 mm f/1.4, but something should be in focus. I need to keep an eye on that.
While watching TV this evening (Die jungen Ärzte, 2/29), saw this:
|
|
|||||||||
The subtitles translate as “Oh, yes, then I fell into this silly cable”. And there it hangs, on the left of the display, and it continues down to the floor!
Yesterday I found a solution for the mail replies to hydra.lemis.com: add an MX record. And it worked, at least for me.
I couldn't put in mx1.lemis.com, because it doesn't know hydra. But it's still broke! Only mx1 can access mx0. I'm not sure how I tested it, but “it worked for me”. And for nobody else.
So, the correct workaround is (finally!)
=== grog@freefall (/dev/pts/0) ~ 8 -> host hydra.lemis.com
hydra.lemis.com mail is handled by 20 mx1.lemis.com.
hydra.lemis.com mail is handled by 10 mx0.lemis.com.
And finally all the waiting mail came from. It's interesting to note how many people contacted me about it, assuming that it was a permanent error, though they typically got messages like this saying that the server will retry for 5 days:
####################################################################
# THIS IS A WARNING ONLY. YOU DO NOT NEED TO RESEND YOUR MESSAGE. #
####################################################################
I'm sorry to have to inform you that your message has not yet
been delivered to one or more recipients. It's attached below.
<groggyhimself@hydra.lemis.com> delayed: host mx0.lemis.com (121.200.11.253): Network
error: Could not connect: Connection refused
Only one person got a “modern” message:
|
|
|||||||||
It's 140 kB in size, compared to about 1 kB for the typical message. To make up for that, it omits almost all information: the name of the server (though it appears to be in the Google domain), the time of the message, whether it would retry or not (it seems that it didn't, unlike all other servers). We're living in modern times!
| Friday, 5 January 2024 | Today's diary | Today's images | top | previous | next | last |
Somehow I spent most of the day catching up with documenting the problems I've had so far this year. And I'm still dawdling with moving my computer display to hydra, at least partially because it's going to be a lot of moving monitors around. But it's uncomfortable the way it is now, so I should really do so.
Using a mouse with a cable is a pain! Time to buy a new one. I've been using the Logitech M705 for nearly 11 years, without being overly happy with it. But it works. Still, that's a long time. What has changed?
Off searching the web for “best mouse” and such things. but didn't come up with anything overly interesting. Mice are beginning to look even more bizarre, but I couldn't find the details I wanted. I could, however, pay up to $200 for a “gaming” mouse (is that a wild mouse?). In the end, it looks like I should just get a new M705.
As the result of the lack of carburettor, our lawn mower is still out of action, after nearly 2 months in a season with particularly virulent grass growth. Garry Marriott has lent us his mower, and today Paul Donaghy came by and mowed the lawn for us. The mower is a particularly large one, and he had difficulty with the corners. Garry doesn't have anything like that, so it works fine for him.
Shepherd's pie for dinner again this evening. Last time I had finally established the correct quantities, so there wasn't much to do. I had 250 g (OK, 248 g) of lamb, and for that I needed a dish with a surface of 250 cm². For a circular dish, that's a diameter of 17.8 cm. The one I found had a diameter of 19 cm, but it was only just big enough:
|
|
|||||||||
|
|
|||||||||
That fitted fairly well. And the gravy? 1.5 times meat makes 375 g. I tripped over a fencepost and made some with 375 g of water, a total of 413 g, and it was probably a little too much.
The good news is that the sachets of mashed potatoes (the ones with contradictory quantities) weigh a “metric” 115 g, very close to 4 oz avoirdupois (113.4 g). They want 500 ml of water, making a total very close to the 600 g in the recipe (though here today it should really have been 500 g). Still, that's close enough.
Next, how long and how hot do I cook it? 30 minutes at 180°. But no, after that time is was still quite pale. I had to raise the temperature to 190° and put it on grill, after which things were quite acceptable:
|
|
|||||||||
And somehow it tasted better today.
| Saturday, 6 January 2024 | Today's diary | Today's images | top | previous | next | last |
Jesse Walsh along today to do some more gardening. He's a fast worker, but it still took a long time. We should consider reshaping the garden to make things easier.
I'm using hydra more and more, which currently means that I have a total of 7 active displays. But that's only part of the desktop:
|
|||||||||||||||||||||||
From left to right, that's:
So the only thing stopping me from connecting all these displays together as one system is that I didn't connect eureka:0.0, and that I didn't reboot euroa. Then I could have connected them all together with x2x. That gives me the largest display surface that I have ever had, (finally) beating the 17,382,400 pixels that I had 22 years ago:
| Number | Resolution | Pixels | Total pixels | |||
| 2 | 1366x768 | 1,049,088 | 2,098,176 | |||
| 5 | 1920x1080 | 2,073,600 | 10,368,000 | |||
| 1 | 1920x1200 | 2,304,000 | 2,304,000 | |||
| 1 | 2560x1440 | 3,584,000 | 3,584,000 | |||
| 1 | 3840x2160 | 8,294,400 | 8,294,400 | |||
| 26,648,576 | ||||||
Why didn't I connect up the stragglers? It's not worth the trouble. This kind of setup is just plain uncomfortable, covering an angle of about 200°. Even without the laptops, the different angle of view for hydra and eureka is physically painful, to the point that I've stopped using x2x and am now using separate keyboard and mouse for each system.
House photo day, and once again I tried doing it on hydra.
Success! Well, sort of. These strange screen blank events on the root window still occur, but they prove to be harmless, and xearth soon clears up the mess. About the only problem that occurred is something that I have seen before. For some reason, this view continually gives me problems with newer versions of Hugin. Here as stitched with the same commands on hydra (Hugin version 2023.0.0) and eureka (Hugin version 2018.0.0). Run the cursor over an image to compare it with its neighbour:
|
|||||||||||||||||||||||
|
|||||||||||||||||||||||
But that's nothing to do with X, just another hurdle to cross when the time is right. The actual stitching is done in batch mode, so theoretically I could do this with the old software. So there's one less obstacle to moving my displays to hydra. About the last remaining one is physically moving the connections, including the new monitor. And that will require new connections, since the single DisplayPort cable that I have is too short for the new location.
| Sunday, 7 January 2024 | Today's diary | Today's images | top | previous | next | last |
I really should be moving my displays to hydra, and if I had more time, I would do so. But maybe the delay isn't such a bad idea after all. It's becoming clear that while the new LG 27” 27UP850-W 4K UHD monitor is excellent for photos, it causes problems for much software, in particular “modern” things like web browsers, which can't scale correctly. But how do I even resize the cursor?
So maybe it's a better idea to move it to position 0 in the new layout, where I don't use it all the time, but where it's available for photos. And for that, I need to get Nvidia to talk to the old Matrix monitor.
I've had my “new” Olympus E-PM1 for a couple of days now, and it looks as if it might have serious problems, but I've been too busy to check. Finally got round to it today: take comparison photos at minimum and maximum focal lengths and and widest aperture and f/22, with and without an infrared filter on both E-PM1s, using the old Zuiko Digital ED 12-60 mm f/2.8-4.0 SWD on it, because it's one of the few that can take my infrared filter.
The new E-PM1 had great difficulty focusing, and it was happy to take photos anyway. But why the difficulty? It seems that it's the camera/lens combination. The other one also had problems. In the middle of the testing, the battery in the old camera ran out, and I had to remove the tripod mount to change it. I don't have time for this! Let's wait until I'm back from Melbourne.
Finally yt-dlp has been updated to download videos from ARD. A good thing, too. I've been through a whole set of alternatives, all with their own problems:
But finally it's done, at least for this obstacle.
We've been in Dereel for nearly 17 years, well over a fifth of my life. And we're seeing change over that time. This afternoon I went to the Elfords to borrow their mobile fridge again, and on the way noticed a number of changes:
In our old house in Kleins Road they are changing things to look less livable and more commercial:
|
|
||||||||||
|
|
||||||||||
And across the road, a new Jenny (Furneaux) has moved in where Jenny Bartlett once lived. It seems that they have started a riding school, but it would be difficult to tell from the appearance. Only the high-roofed hall behind the house could have something to do with it:
|
|
||||||||||
|
|
||||||||||
|
|
||||||||||
And then down Harrisons Road a new house is under construction, near the one on the corner which was built 4 years ago.
|
|
||||||||||
We had looked at a property in that area a couple of days before Stones Road, if I'm interpreting my diary correctly. The issue was at least that there was no electricity, and five years ago I wrote:
I wonder what they'll do for electricity.
But now there is power, at least to the new house and potentially to the old one. I wonder what they had to pay for that.
Tomorrow we're off to Melbourne again, for the first time in nearly 3 years, for Yvonne to have another EUS and possibly FNA. And it had to be a Monday!
The process takes forever (well, 30 minutes). But then there's the interminable waiting time, 90 minutes recovery and of course 4 hours driving. Last time I went shopping and looked round the places of my childhood.
So, where do we go shopping? All the markets are closed, but there are other places. Setting up my route, I discovered that Fleischers (nomen est omen) Butchers in Boronia have closed down. What other German butchers are there? Years ago I went to the Wursthütte in Malvern, but they're not open on Mondays either. And then there's the Malverrn Continental Butchers, which appear to be the same thing with a different web site, also generally closed on Mondays and in this case also until 22 January.
What's left? Box Hill Central for Chinese food, Minh Phat (a company which seems to have lost its web site) for Vietnamese and general South-East Asian food, Casa Iberica for Spanish and Mexican food. Not much for an itinerary.
Three years ago I was pleasantly surprised by the new ceramic frying pan that I had bought: it was really non-stick:
|
|
|||||||||
|
|
|||||||||
But that was 3 years ago. Sadly, like so many ceramic things, it seems, it's no longer even average non-stick:
|
|
|||||||||
Even removing the residue and putting it through the dish washer wasn't enough:
|
|
||||||||||
That's a pity. I wonder if newer pans will be better.
Lamb biriani for dinner this evening. We had originally planned it for a group dinner nearly 3 years ago. But then a COVID-19 lockdown was imposed, and since then we no longer have dinner guests.
The dish is remarkably complicated, though today I found a few minor improvements: you don't need to soak saffron in water if you're then going to put it in a broth, for example. And the spice proportions (which I'm sure I had greatly increased relative to the original recipe) still seemed far too low, and I increased some of them by up to double.
It didn't taste bad. But somehow it didn't taste nearly as good as I had recalled, and certainly not as much to justify the work it took.
| Monday, 8 January 2024 | Today's diary | Today's images | top | previous | next | last |
Off to Melbourne as planned this morning, through some of the worst weather I've seen in Australia: fog with visibility down to 100 m, and drizzling rain all day long, making driving on the freeway the kind of agony that I haven't seen in decades.
Dropped Yvonne at the Epworth Hospital and on to Minh Phat, not far away, where Google Maps carefully guided me into a dead-end delivery bay. Back out and round the corner, fighting my way through the rain, and actually found quite a bit of stuff, though I had planned to do most of my shopping in Box Hill Central.
Then off to Box Hill Central, where Google Maps obligingly brought me to the underground car park, and found a parking place close to the stairs. Upstairs, nothing looked familiar. Yes, it's been 6 years since I was last there, but I shouldn't have forgotten things completely.
After discussing with a bloke trying to raise money for guide dogs, found my way up another flight of stairs. Still nothing looked the way it should have done. Found a sign pointing to the exit, but it was to the east. It should have been to the north. Had I lost my orientation since coming into the car park? No, it proved that I was in a different building! Last time I was there, I thought there was only one, but they have extended with further buildings.
Into the main area, which was as I remembered it. Except that the supermarkets weren't anything like as good as I recalled. On the whole, I think that Minh Phat matches my requirements better. Bought some choi sam and something with a name like “mustard caixing”, which looked similar. I should have taken a photo. And that was all.
Walked around the area round Box Hill. I don't know how long the predominantly Chinese centre has been there, but my father was in the area 90 years ago, going to Box Hill High School. And it's one of the first places I went to when I returned to Australia on 12 March 1997. No mention of a Chinese centre there at the time.
Getting out was fun: the payment machines for the parking are really confusing, and if they're capable of issuing offered a receipt, they do a good job of hiding it. And when I got to the exit, the barrier wouldn't accept the ticket. I couldn't even get it to try to read it. Pressed the “Help” button, with no immediate reply. And then I saw that the barrier was open! My guess is that it recognized my registration (which it had written on the ticket when I entered). But they're too polite to assume that I wouldn't know that.
OK, where next? Casa Iberica in Alphington? Sorry, says Google Maps, they're closed for the day—and that round 15:10. OK, Fitzroy? They close at 16:00, and you'll need 20 minutes to get there. OK, off, getting misled on the way, and arrived round 15:40, where at least I was able to get some chorizo and masa lista, along with some Manchego cheese for the fun of it.
And now? Round the corner to the Melbourne Museum? They close at 17:00, so I wouldn't have much time there. In the end I gave up and went back to the hospital, where, though Yvonne had finished her procedure, I wasn't allowed to see her for another hour.
Finally she was allowed out. Down to the car park, where they had the same kind of parking machines as in Box Hill. But there I paid $2 for 2 hours. Here it was $24 for 1½ hours! And interestingly, Yvonne says that she saw it display a brief text in German. Sadly, I didn't, but that once happened to me in Box Hill 5 years ago, and on that occasion I thought that I might have hallucinated. And at the exit, once again my card wasn't accepted. This time it pulled it in, spat it out and asked me to repeat. Tried all four orientations to no avail. Finally called help, and was asked if I had paid, if I had tried all orientations. Then the barrier went up. What a mess!
Off home. What a day!
| 10:30 | Set off for Melbourne | |
| 12:30 | Yvonne arrives at hospital | |
| 15:15 | Yvonne admitted to procedure | |
| 15:45 | Procedure complete, start recovery | |
| 17:30 | Recovery (“1½ hours") complete | |
| 17:45 | Yvonne dismissed | |
| 18:00 | Leave for home | |
| 19:45 | Arrive home |
The bad news is that yes, there is some anomaly in Yvonne's pancreas, and it might need an operation that would remove most of it. We'll have to discuss that with Kon Shimokawa soon.
I spent over an hour at the hospital waiting for Yvonne. What could I do? Play around with my phone. I had tried to send her some photos of fish in the Box Hill shopping centre, but I hadn't been able to work out how to do it. Finally I found out. I should have written it down: it's so non-intuitive that I have forgotten again.
The other surprise was the amount of credit that mobile date uses. It's difficult to get an
overview from ALDI Mobile, but it must
have been about $20. Given that I used to get by for a year on a $15 recharge, that's
clearly a lot. My tariff “plan” is cheap for phone calls, but really expensive
for mobile data. Should I change? No, I don't think so. Today I could have used one of
their $20 data “plans”, but that's so seldom that it wouldn't be worthwhile. And the other
costs, including parking, made it relatively insignificant.
| Tuesday, 9 January 2024 | Today's diary | top | previous | next | last |
A lot has happened over the last couple of days, and I didn't get a chance to write my diary yesterday. As a result, I barely got it finished before dinner. Once again the question rises whether I shouldn't just be less verbose. But no, I think it's worth my while.
Finally got round to looking at the photos I took on Monday. The good news seems to be that the focus isn't that bad after all. And after trying to take photos with the Zuiko Digital ED 12-60 mm f/2.8-4.0 SWD, I discovered that my two newest general purpose lenses, the M.Zuiko Digital ED 12-100 mm f/4.0 IS PRO and the Olympus M.Zuiko Digital ED 12-200 mm f/3.5-6.3, both have the same 72 mm filter thread, so next time I can try them.
First recognition: yes, the new camera does focus correctly:
|
|
||||||||||
That's taken at f/4 and a focal length of 60 mm.
And the infrared filter does make a difference. Here without, then with (run the cursor over an image to compare it with its neighbour):
|
|
||||||||||
|
|
||||||||||
Surprisingly, the version with the infrared filter looks more “natural” than the one with full spectrum. I should investigate the influence of ultraviolet light.
| Wednesday, 10 January 2024 | Today's diary | top | previous | next | last |
Spent most of today not quite as busy as yesterday, but there's still a lot of work to catch up with; my benchmark mail inbox was sitting at round 100 messages, nearly 2 pages of display.
Despite my overflowing mail inbox, it seemed that I wasn't getting as much mail as usual, in particular from news sources. And Quora seems to have stopped altogether.
OK, check the mail logs. To my surprise, I found a whole lot of things like this:
Jan 5 15:53:31 eureka postfix/smtpd[25111]: connect from mx2.freebsd.org[96.47.72.81]
Jan 5 15:53:31 eureka postfix/spawn[25649]: warning: command /usr/local/bin/perl exit status 2
Jan 5 15:53:31 eureka postfix/smtpd[25111]: warning: premature end-of-input on private/bld-policy while reading input attribute name
Jan 5 15:53:32 eureka postfix/spawn[25649]: warning: command /usr/local/bin/perl exit status 2
Jan 5 15:53:32 eureka postfix/smtpd[25111]: warning: premature end-of-input on private/bld-policy while reading input attribute name
Jan 5 15:53:32 eureka postfix/smtpd[25111]: warning: problem talking to server private/bld-policy: No error: 0
Jan 5 15:53:32 eureka postfix/smtpd[25111]: NOQUEUE: reject: RCPT from mx2.freebsd.org[96.47.72.81]: 451 4.3.5 <groggyhimself@lemis.com>: Recipient address rejected: Server configuration problem; from=<owner-src-committers@freebsd.org> to=<groggyhimself@lemis.com> proto=ESMTP helo=<mx2.freebsd.org>
Jan 5 15:53:32 eureka postfix/smtpd[25111]: disconnect from mx2.freebsd.org[96.47.72.81] ehlo=1 mail=1 rcpt=0/1 data=0/1 rset=1 quit=1 commands=4/6
That was the first occurrence, but it's 5 days ago. What is it? What's private/bld-policy? Have I changed something in my mail configuration? Or DNS? No, nothing corresponds. And the strange thing is that not all email is rejected.
Off looking for private/bld-policy. Two hits:
=== grog@eureka (/dev/pts/4) ~ 6 -> locate bld-policy | xargs ls -l
-rw------- 1 root wheel 0 13 Feb 2013 /home/oldroot/var/spool/postfix/pid/unix.bld-policy
-rw------- 1 root postfix 0 2 Jan 2017 /var/spool/postfix/pid/unix.bld-policy
Clearly neither of those has anything to do with the issue. What is it? It took a while to think of searching the Postfix) configuration files. And there I found:
smtpd_recipient_restrictions = permit_mynetworks,
check_recipient_access hash:/usr/local/etc/postfix/accept-to,
reject_unauth_destination
check_policy_service unix:private/bld-policy
What's that? It must have been there for ever. On searching, it seems that I added it relatively recently and implicitly on 13 January 2007, apparently as the result of installing the now-defunct Blacklist Daemon—thus the bld in the name. Removing the last line of the specification fixed the problem.
So why did it turn up now? My guess is that bld went away when I installed the current version of eureka. So far I have been receiving mail only relayed from mx1.lemis.com, so permit_mynetworks allowed everything through. But last Friday I changed my firewall rules to allow direct mail access, and that triggered the issue. And why did some mail get through anyway? Presumably, for some reason, it was relayed via mx1.
And Quora? They really seem to have given up trying.
| Thursday, 11 January 2024 | Today's diary | top | previous | next | last |
Looking at the weather observations today, discovered something disturbing: no data. The entire directory /var/db/mysql/weather was empty! The modification timestamp of the directory showed that it had last been changed at 13:14.
How can that happen? My programs can't do that, because they don't delete entire directories, and they don't know about some of the files in there. In any case, nothing for it: that's what backups are for, so restored all my MySQL tables, including the ones in other databases that weren't deleted, and noted with a certain satisfaction that I didn't need to restart MySQL.
And then they were gone again, at 15:48! Another restart, and some head-scratching. I had just received two emails from CERT-bund warning me of open ports on my firewall. Ah, yes, I should close them. Could they be the clue to the lost data? One of them was NetBIOS, but that still requires authentication. As the mail said, the main vulnerability is elsewhere:
Over the past months, systems responding to NetBIOS nameservice requests from anywhere on the Internet have been increasingly abused for DDoS reflection attacks against third parties.
And if it was an attack from outside, why just my weather data? Well, do I know that it's just my weather data? I don't see anything else, but that's something to check tomorrow. In any case, I didn't have any further issues with the weather data.
I was brought up in the mid-20th century, at a time when the treatment of the Jews in Hitler Germany was a shock that still hadn't died down. The newly formed United Nations had held a Genocide Convention as a result. Never again should this happen.
And the Jews finally found their homeland, as the British had wished for a long time. The country flourished. The only fly in the ointment was the jealous neighbours who wanted (and in some cases still want) to destroy the country. Bad neighbours!
But gradually my sympathy for israel has dampened. It has been clear for some decades that they have no intention of releasing the territory that they have occupied, and there was more than enough evidence of their mistreatment of the rightful owners of the land even before 7 October 2023. Under those circumstances, it's not surprising (though also not pardonable) that the Palestinians arise and attack the Israelis.
But wait, the Israelis can do better. You kill 1200 Israelis? We can kill 25,000 Palestinians! And we won't stop! Out, damned Palestinians!
It's not surprising that other countries are opposed to their brutality and destruction, and finally South Africa has gone to the International Court of Justice and accused Israel of genocide—the very crime that was created because of the way the Jews were treated under Hitler.
We'll see how that pans out. The whole thing turns my stomach. I'm reminded of verse 7 of Deuteronomy:
When your God brings you to the land that you are about to enter and possess, and [God] dislodges many nations before you—the Hittites, Girgashites, Amorites, Canaanites, Perizzites, Hivites, and Jebusites, seven nations much larger than you—and your God delivers them to you and you defeat them, you must doom them to destruction: grant them no terms and give them no quarter.
But one thing has been becoming clear to me: what I learnt at school was not the unbiased truth. The Palestinians weren't just jealous, they were angry for having been driven out of their home and into refugee camps that have now been holding “refugees” for longer than I have been alive. And even then the peace-loving Israelis had used genocidal tactics, such as the atrocities.
So I've been trying to find more about this for some time. A while back I participated in a course produced by some Israeli institution. It was interesting, but I've forgotten most of it, and I don't recall much discussion of violence. On the other hand, Deir Yassin wasn't an invention.
So I've been looking for, and finding, videos about the matter. There's this one:
But it's not very long (10 minutes), and it suggests that nothing seriously went wrong before the Six Day War. What led up to Deir Yassin? More searching brought me to this series of videos, a total of 3 hours, produced by Al Jazeera:
That paints a very different picture. On the face of it, it all looks very plausible, and there are details in there that explain what happened later. The bottom line I got was: the Jewish (Zionist) settlers were anything but friendly towards the people they displaced, and atrocities like the ones we're seeing now also occurred before the Second World War. The difference seems to be the press coverage, possibly one of the reasons why the Israelis seem to delight in killing reporters.
But is it balanced? Al Jazeera is an Arabic broadcaster, one run by the government of Qatar. On the whole I have a good opinion of them, but that doesn't mean that they're always neutral in what they present. It would be nice to find something as plausible and as detailed presented by “the other side”.
| Friday, 12 January 2024 | Today's diary | Today's images | top | previous | next | last |
I've heard the term mustard greens before, but never really quite understood it. But it seems that whatever I bought on Monday is one of many kinds.
Petra Gietz knows of others; in fact, she grows them, and today she brought some along for comparison:
|
|
|||||||||
That's mine at the top, hers are the bottom two. It's interesting to note that they all have serrated leaves, and they have a somewhat spicy taste. But hers are more strongly flavoured than mine, and she tells me that they grow like weeds. So possibly I should investigate further.
The last few years have been pretty terrible for the garden, but lately there have been some encouraging signs. The curry tree is finally (just) flowering:
|
|
|||||||||
And the Hibiscus rosa-sinensis and transplanted Fuchsia are also promising:
|
|
|||||||||
|
|
|||||||||
So is the Corymbia ficifolia, which must have 500 buds:
|
|
|||||||||
And even the Clematis that was looking so sick six weeks ago has picked up:
|
|
|||||||||
So what caused yesterday's loss of data? Was it really a security breach? And did more get lost than I thought? How do I check?
The most important things, of course, are my photos. And that proved comparatively easy. I do an automated backup every night, and this morning the output showed:
Fri 12 Jan 2024 05:05:01 AEDT Greg Photo Backup disk 8
-rw-r--r-- 1 root wheel 25 11 Jan 05:34:05 2024 /photobackup/Iam
Filesystem 1048576-blocks Used Avail Capacity iused ifree %iused Mounted on
/dev/da3p1 7,629,565 7,217,012 336,257 96% 1,743,234 1,500,412 54% /photobackup
sent 34,480,155 bytes received 190 bytes 36,123.99 bytes/sec
/dev/da3p1 7,629,565 7,217,012 336,257 96% 1,743,234 1,500,412 54% /photobackup
Fri 12 Jan 2024 05:21:40 AEDT Photo backup ended
~
By chance I hadn't processed any photos yesterday. The important thing here is the iused column in the df output: no change. So my photos are safe unless somebody went and replaced the contents.
What about elsewhere? If files get deleted, the directory is modified (remove the file name entry). So this should help:
=== root@eureka (/dev/pts/33) /usr/local/etc/postfix 35 -> find -x / -type d -mtime -2 > /var/tmp/toucheddirs=== root@eureka (/dev/pts/33) /usr/local/etc/postfix 36 -> find -x /home -type d -mtime -2 >> /var/tmp/toucheddirs=== root@eureka (/dev/pts/33) /usr/local/etc/postfix 37 -> xargs < /var/tmp/toucheddirs ls -lrdt | less
This finds all directories that have been modified in the last 48 hours and lists them by modification date. The only relevant entries were:
drwx------ 4 mysql mysql 1536 11 Jan 17:33 /home/var/db/mysql/weather
drwxr-xr-x 2 root mysql 1024 11 Jan 17:33 /home/var/db/mysql/weather/saveweather
That's the time when I restored the contents and also created a backup directory. So it seems that nothing else was changed.
But was it an attack from outside? /var/log/messages is full of messages like
Jan 11 15:52:22 eureka smbd[13647]: [2024/01/11 15:52:22.505008, 0] ../source3/lib/access.c:338(allow_access)
Jan 11 15:52:22 eureka smbd[13647]: Denied connection from 41.208.71.226 (41.208.71.226)
Jan 11 15:52:22 eureka smbd[13647]: [2024/01/11 15:52:22.505356, 0] ../lib/util/pidfile.c:153(pidfile_unlink)
Jan 11 15:52:22 eureka smbd[13647]: Failed to delete pidfile /var/run/samba4/smbd.pid. Error was No such file or directory
Jan 11 16:15:41 eureka smbd[59185]: [2024/01/11 16:15:41.244124, 0] ../source3/lib/access.c:338(allow_access)
...
That covers the time of the second deletion, and it doesn't show anything different from the rest of the day. And now, of course, it's silent.
So: there was a vulnerability that doesn't seem to have been exploited, and the localized nature of the data loss also doesn't seem to match. One that got away?
But then, syncing my diary, I got this message:
...
12900 files...
13600 files...
13993 files to consider
./
diary-aug1966.php
102,332 100% 1.57MB/s 0:00:00 (xfr#2, to-chk=13805/13993)
13,993 files? That should be over 15,000! What got lost?
Oh. Symlink pain. That was on hydra. Try again on eureka and I got
15313 files to consider
I really need to complete this transition.
Gaza is term that's in everybody's mind at the moment. But looking back in this diary, I found the alternate spelling Ghazza—by me! Why?
Looking at the original spelling غَزَّةَ, it's clear that it starts with the letter غَََّ, also written medial ـغـ and final ـغ. That's transliterated as ghayn in the Wikipedia Arabic alphabet, though I recall it as ghain. And that's the letter normally transliterated as gh, like in names for Baghdad and Afghanistan. The letter normally representing g is ...
Oh. It seems that there isn't one. I learnt (as far as it went) Arabic script via the Malay Jawi script, and the alphabet contains some letters that don't exist in Arabic. Letters like ڽ (nya) and ڠ (nga), representing the ny and ng sounds, are understandable, but ڤ (pa), representing the p sound, is less understandable. I knew about that because over 60 years ago my friend Fadel Abdul Khadir, from Kuwait, couldn't read ڤيراق (Perak) on stamps in my stamp collection, pronouncing it Ferak instead. But only now do I discover that ݢ (ga) is another of these basic sounds that have no representation in the Arabic alphabet.
So: Gaza, Ghaza or Ghazza? I don't see any explanation for the doubled z. And from what I've heard in Arabic, Gaza is probably closer. But it took a while to find that out.
Intrestingly, there's a Ghazzeh ( غزة)) in Lebanon. Bar the diacritics, which I don't understand, the spelling is the same.
Does my “new” Olympus E-PM1 work well enough to keep? Should I return it or ask for a reduction in price? Spent some time looking at videos about the topic, and this one in particular seems useful:
I'm still making my way through it, but so far it has given me a better overview of the topics, and also the difference between “only infrared” and full spectrum imaging. It also shows that I misinterpreted the photos I took Nowadays “only infrared” is effectively monochrome, like this photo I took today, using my R72 infrared filter:
|
|
|||||||||
And here it is again without the filter:
|
|
|||||||||
In passing, it's interesting to note the difference in exposure: the R72 filter reduces exposure by about 1 EV for this camera.
Other details: I stopped while he was talking about infrared photography with unmodified cameras. I've been there before nearly 10 years ago, but on that occasion I didn't know about changing the white balance, something that I'm going to have to look at now. So far he hasn't made the important point that the results depend on the sensor/internal filter combination, as I had noted at the time. But he did bring back the spectre of hot spots and gave a link to a table of good and bad lenses. Not surprisingly, the Zuiko Digital ED 12-60 mm f/2.8-4.0 SWD that I used last time figures in the “bad” column, but it claims that the M.Zuiko Digital ED 12-100 mm f/4.0 IS PRO is good. On the other hand, one of the comments states “very bad hot spot” for that lens.
The other interesting point is that he recommends only brand name R72 infrared filters, specifically Hoya. Went looking. In my diameter (coincidentally 72 mm), the cheapest I could find cost $92.75, and one (from Austria) cost $186 plus $90 postage! That quite relativizes the $90 I paid for the camera.
And how much difference do the scratches on the sensor make? Took a photo without a lens of the backdrop at the back of my office and got:
|
|
|||||||||
Yes, it's not even, but I can't see any evidence of scratches. Why not?
| Saturday, 13 January 2024 | Today's diary | Today's images | top | previous | next | last |
Seen this morning:
|
|
|||||||||
That's the best Buddleja_×_weyeriana I've seen since we've been living here.
Spent a lot of time this afternoon looking at further videos on full spectrum photography. In particular, Rob Shea produces a surprising amount of content. Today I looked at a video titled DxO PhotoLab 7 for Infrared Photography :
I'm still using DxO PhotoLab 5, and so far I haven't seen any reason to upgrade. Is this the reason? No, I don't think so. Nearly all the functionality he describes is in version 5 as well. And some of the ideas are worthwhile, like this transition (run the cursor over an image to compare it with its neighbour):
|
|
|||||||||
|
|
|||||||||
|
|
|||||||||
This is an image I took 10 years ago, and with which I wasn't very impressed (first image). The second is after applying some kind of white balance, and the third after a little frobbing with DxO. They're still not good, but it shows some promise.
Every day I sync my diary and other web files to the external web server www.lemis.com. It usually take a couple of minutes. But today it hung in the middle. Why? Yet another thing to investigate.
| Sunday, 14 January 2024 | Today's diary | top | previous | next | last |
Why was my network upload so slow yesterday? Did some checking, first with my network statistics page. It showed complete nonsense. Then with traceroute, which showed latency of round 250 ms down my network link. Problems at Aussie Broadband? First I need to understand why my network statistics were so crazy.
That was simple: firewall misconfiguration. I had blocked all incoming ICMP packets, and that included outgoing pings. I do need to do something to limit the amount of incoming ICMPs, which are round 300 per second, but in principle it doesn't help to just reject them. That would only work if the pingers decide to give up.
Second best: reject all packets for systems that don't exist. The same consideration applies, but at least that will show if it makes a difference. All my IP addresses are in the upper half of my net block, 192.109.197.128/25, so I can reject anything to 192.109.197.0/25. I also thought that I could reject 192.109.197.192/26, but I have some things in that range. It'll be a while before I see any results there.
Gradually performance improved—I think part of the latency issue was that I was downloading a video at the time—and I started looking at other things, such as this diary. Finish it, sync...
/: write failed, filesystem is full
mkdtemp: private socket dir: No space left on device
Oh. That's on the external server. Tried to connect, which took for ever. OK, first find the disk hog. Log files, of course. This machine only has about 50 GB of disk, and over 15 GB were in old log files. I didn't want to delete the last one, just compress it, but for that I need more space....
Finally it was done, but performance remained terrible. top told me something like:
238 processes: 62 running, 176 sleeping
CPU: 94.8% user, 0.0% nice, 3.5% system, 1.7% interrupt, 0.0% idle
Mem: 486M Active, 2589M Inact, 976K Laundry, 739M Wired, 375M Buf, 389M Free
Swap: 3048M Total, 391M Used, 2657M Free, 12% Inuse
PID USERNAME THR PRI NICE SIZE RES STATE C TIME WCPU COMMAND
63576 www 1 23 0 28M 13M RUN 1 0:02 6.42% httpd
63591 www 1 24 0 30M 13M RUN 0 0:01 5.73% httpd
63536 www 1 23 0 28M 13M lockf 0 0:02 5.72% httpd
The web server was maxed out:
=== root@lax (/dev/pts/6) /home/grog 4 -> ps aux | grep httpd | wc -l
111
111 httpd processes! What does restarting the web server do? Nothing. The call hung. OK, hard shutdown:
=== root@lax (/dev/pts/8) ~ 1 -> killall httpd
No matching processes were found
I've seen that a lot recently. What's behind it? In any case, this did it:
=== root@lax (/dev/pts/6) ~ 9 -> ps aux | grep httpd | awk '{print "kill -9 " $2}' | sh
But restarting the server quickly shot the load up again:
=== grog@lax (/dev/pts/5) ~ 5 -> ps aux|grep httpd|wc -l
127
What is all this stuff? The log file shows:
8.219.75.135 - - [14/Jan/2024:03:33:07 +0000] "GET /grog/diary-apr2008.php?dirdate=20080426&imagesizes=111211112111111111112111111111111111111111111111212 HTTP/1.1" 200 190059 "-" "Mozilla/5.0 (Macintosh; Intel Mac OS X 10_15_7) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/116.0.0.0 Safari/537.36"
8.219.105.244 - - [14/Jan/2024:03:33:12 +0000] "GET /grog/diary-apr2008.php?dirdate=20080406&imagesizes=1111112111221111111111111111111111111111111112 HTTP/1.1" 200 243494 "-" "Mozilla/5.0 (Macintosh; Intel Mac OS X 10_15_7) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/116.0.0.0 Safari/537.36"
8.222.203.212 - - [14/Jan/2024:03:33:13 +0000] "GET /grog/diary-apr2008.php?dirdate=20080423&imagesizes=2111111111111111111112111111111011111111111111111113 HTTP/1.1" 200 240241 "-" "Mozilla/5.0 (Macintosh; Intel Mac OS X 10_15_7) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/116.0.0.0 Safari/537.36"
8.219.75.135 - - [14/Jan/2024:03:33:12 +0000] "GET /grog/diary-jul2008.php?dirdate=20080713&imagesizes=11111121111111111111111111111211111111111112 HTTP/1.1" 200 306536 "-" "Mozilla/5.0 (Macintosh; Intel Mac OS X 10_15_7) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/116.0.0.0 Safari/537.36"
8.219.198.76 - - [14/Jan/2024:03:33:11 +0000] "GET /grog/diary-apr2008.php?dirdate=20080420&imagesizes=1111112111111111111311212 HTTP/1.1" 200 243415 "-" "Mozilla/5.0 (Macintosh; Intel Mac OS X 10_15_7) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/116.0.0.0 Safari/537.36"
8
I've seen this before. Those are web crawlers, and they're iterating through all image sizes in my diary. At the time I had identified the names of the crawlers and redirected, but these crawlers are too polite to have reverse lookup. I can employ heuristics, but there must be an easier way. I need to think about it.
The good news is that the load average gradually dropped during the course of the day, though it remained over 20. Can it be that the crawlers adapt to when they expect the network load to be the lowest?
| Monday, 15 January 2024 | Today's diary | Today's images | top | previous | next | last |
While in Melbourne on Monday, I bought some prawn (“shrimp”) pastes that I hadn't seen before. This one, I thought, would be like petis udang.
|
|
|||||||||
|
|
|||||||||
It wasn't, but possibly close enough. The other I had intended to use to make some prawn noodle breakfast, possibly like the too-pedas Tean's mi udang paste:
|
|
|||||||||
But it's in oil. Maybe something like KL Hokkien Mee? With that vague idea, I tried:
| quantity | ingredient | step | ||
| 60 g | squid tubes, cut into strips | 1 | ||
| 60 g | raw prawns | 1 | ||
| 1 | small chili, chooped | 2 | ||
| 10 g | garlic | 2 | ||
| 5 g | ginger | 2 | ||
| 15 g | “shrimp” paste | 2 | ||
| 60 g | choi sam | 3 | ||
| 180 g | cooked spaghetti | 4 | ||
| 50 g | medium prawns (4 to 6) | 4 | ||
| soya sauce | 5 | |||
Unlike with KL Hokkien Mee, I didn't fry the squid tubes; I fried the choi sam stalks with the paste, garlic and ginger, then added the prawns, squid and the rest of the choi sam. The result looked a little dry, so I added a little soya sauce:
|
|
|||||||||
The result? Not bad, not spectacular. It proved that the paste was pedas enough by itself, so if I do it again, I'll leave out the chili.
So how can I fix the load situation on my external web server? The issue is at least that crawlers overload it by requesting the same page multiple times with different image sizes. The obvious solution is to migrate the image resizing to the browser. That means Javascript, something that I barely know. Spent some time looking at the function oneimage () that prepares the images. It's 750 lines of particularly convoluted code, some of which may be obsolete. But converting it is non-trivial, and for the moment the load is down, so I'll put that on the tuit queue.
I've been running a mixed X environment for some time now, trying to decide what to do next. My issues are:
So what do I do? What is the problem with fvwm2, anyway? Does it behave the same if I build it from source? That's easy enough to test.
No, the behaviour is different. Previously it crashed every day or two. Now it crashes every time. But that has an advantage that I can test it. What happens if I don't give it a configuration file? It works! So it has something to do with the configuration file, which explains part of how it could get away.
More playing with the configuration files. Somehow it's related to the xterm starting menus that I've been working on. But it's not the entries themselves; just a frame of a menu is enough to cause it to crash.
OK, what about other versions of fvwm2? I have them on all my systems, and they don't crash. So: run tiwi's fvwm2. It crashes! The crash itself is somewhere in an X library, so that in itself is not surprising, but it does help localize the bug.
So run fvwm2 on eureka, just pointing to hydra. That works. The only issue is that I can't start an xterm natively on hydra; I need ssh. For the time being I'll have to put up with that, though it seems silly to run input to a local xterm via another system.
So: update my genxterms.el to not single out the local system for special treatment if that system is hydra:
But it doesn't work! While balancing my parens, I got messages like
Wrong number of arguments: (lambda (entry) (car entry)), 0
How do you debug that? That's the body of the called function, but the name itself is missing. All I have is entry. With a bit of searching, found
(defun system-entry (entry)
(car entry) )
But all the calls appeared to be complete. Yes, Elisp has debugging features, but this runs in batch mode, and I have no idea how to run it interactively. So off to debug with (print), which was, for some reason, particularly difficult. It seems that the error messages are wrong. In the end I gave up for the day with a configuration file that used ssh for every system. Grr.
| Tuesday, 16 January 2024 | Today's diary | Today's images | top | previous | next | last |
Why am I having so much difficulty configuring fvwm? Part of it is the historical background, coming on 28 years, part of it is my lack of good Elisp debugging—something that I should investigate more—but it seems that part of it is simply that the original was also buggy. Why did I write this script in Elisp? It's predominantly string-based, and something like PHP would have done better, but when I wrote it I didn't know PHP. But since then it has been through 116 revisions and more than tripled in size.
So: after much pain I've established a bug in the original, but ran out of energy to fix it. Will that be the end of it?
The Fuchsia that I replanted last month is recovering as well as can be expected. At least it's bringing some foliage and buds:
|
|
|||||||||
But then I discovered that I had repotted it “relatively” recently, only a year ago, and at the time it looked much better:
|
|
|||||||||
So what went wrong?
| Wednesday, 17 January 2024 | Today's diary | Today's images | top | previous | next | last |
in Melbourne on Monday I bought these frozen prawn wontons:
|
|
|||||||||
Put them in a mee soup for breakfast. They became enormous:
|
|
|||||||||
In fact, together they weighed about 60 g before and 84 g afterwards, but they're so much bigger! I think 2 would be enough. And yes, they don't taste bad in the mee. They're a little bit pedas, which I hadn't expected.
Decades ago, exposure for a photo was mainly guesswork, helped with little tables like this one that came with my FED-1, since most people didn't have any kind of exposure meter:
|
|
|||||||||
How accurate are they? Looking back through the photos I took at the time, I got reasonably good results, though they could be out by 1 or 2 EV.
Today I don't need that, but I can still mess up exposure, like with this photo I took today:
|
|
|||||||||
I had taken that with the camera set to manual exposure. What I really wanted was this:
|
|
|
|||||||||
Not much difference? That's postprocessing for you. The first image was taken at 1/250 s and f/5.6 (EV 12.9). The second was taken with exposure metering, 1/80 s at f/4.5 (EV 10.7). It's surprising how little difference it made, but the colour gradation in the soil is one place to look.
The www.lemis.com web server continues to suffer under extreme overload, with load averages up to 120. And for once I let my daily sync run. Normally it runs for 10 seconds, checking through the 15,000 odd files. Today, though, I had:
=== grog@eureka (/dev/pts/41) /home/local/X/.fvwm 21 -> time syncgrog
Wed 17 Jan 2024 13:04:44 AEDT
...
sent 494,660 bytes received 256,471 bytes 89.92 bytes/sec
total size is 4,487,460,324 speedup is 5,974.27
Wed 17 Jan 2024 15:24:13 AEDT
real 139m29.308s
user 0m2.956s
sys 0m0.278s
2 hours, 20 minutes! The surprising thing is that the server responds to web page requests relatively promptly.
So I've established why my Elisp scripts were misbehaving. Time to fix things. But while thinking about it, discovered that running fvwm2 on eureka to control a hydra window was even more suboptimal than I thought: every program that it starts is on eureka, not hydra. So it looks as if that isn't the solution. Should I really go to the trouble of finding the bug? Compared to “modern” software, fvwm2 is positively tiny: when I built it, it happened so fast that I thought it hadn't done anything. I should find out again how to build a debug version and at least find out what is triggering the bug.
What can I do with my new full spectrum Olympus E-PM1? I've watched a lot of videos of varying quality, but nothing tells me what sort of motives to use, nor how to process the images for best effect. Full sunlight, lots of greenery, and that's about it.
OK, how about tracking my weekly house photos? What can I do in the settings? Spent some time RTFMing, and discovered that the camera has some vestiges of long-forgotten features like panoramas and 3D images, but not (apparently) exposure bracketing. Still, the settings were worthwhile: it was set to take photos before focus was complete, something that's easy enough to change. And white balance? It's meaningless for raw images (which, all sources agree for once, is the only way to go), but it certainly helps compose. Here the default display on the viewfinder and then the modified one:
|
|
|||||||||
|
|
|||||||||
|
|
|||||||||
|
|
|||||||||
I suppose I could change it further, but so far it's not clear what colours I should be aiming for. At least here I can see colour differences. Frustratingly, it seems that the camera has a further problem: the external VF-2 viewfinder doesn't work.
| Thursday, 18 January 2024 | Today's diary | Today's images | top | previous | next | last |
It seems that it has been for ever since we've had Hibiscus rosa-sinensis flowers. Why? I'm not sure. Part of it was the bad winter two years ago, but even the potted one didn't want to flower. But now the first of many buds has opened:
|
|
|||||||||
|
|
|||||||||
And the outside bush is also gradually recovering, though I wonder if it isn't about time to replace it.
Into town today to talk to Kontoku Shimokawa about the results of last week's investigations of Yvonne's pancreas. In principle nothing that I didn't already know after reading the report. She has a main duct intraductal papillary mucinous neoplasm with maximum diameter 13 mm, and like Wikipedia, he tells us that this is a benign tumour that can develop into cancer. The questions are if and when. He appeared quite nervous, and his answers to my questions were vaguer than I would have liked, but it seems that there is no immediate danger. On the other hand, pancreatic resection appears to be one of the most difficult operations there are, 7 hours or so, and the chances of survival are only about 95%. The real issue is catching it in time if it does become cancerous, and for that she'll initially have an ultrasound in 3 months and an MRI in 6 months.
| Friday, 19 January 2024 | Today's diary | top | previous |
Somehow I can't get myself to migrate further to hydra. An obvious reason is because it more or less works as it is. But it's inconvenient to have two sets of keyboard, mouse and monitor.
So what's holding me up? Little details. While I'm quite happy with the new LG 27UP850-W monitor, but it's clear that “modern” software—and even some older software—has issues with it. A web browser is a catastrophe, and I haven't found a way to tell the window manager how to enlarge its icons and things. Even finding the mouse cursor is non-trivial.
So switch it down to 2560×1440, like the old monitor? No, sometimes it's really useful to have the resolution. Just not for the main display. So: set up the old “Matrix” monitor, as planned? nvidia-setup doesn't want to know about it, and though I could give it the EDID data, what further problems are waiting around the corner? How much longer does it have to survive? Should I maybe buy a new 2560×1440 monitor, like I had planned 2 months ago? I decided against it then, and after looking at pricing today, I think I would decide against it now. In the meantime, do nothing.
Watching further videos on full spectrum (“infrared”) photography today. I was quite positively impressed by Rob Shea's videos, and the videos by others help maintain this impression: they're frequently badly produced and badly informed.
The real question: what do you want to do with infrared photography? You can't reproduce it the way it was, because you can't see infrared. So at the very least you need to remap the colours. But how? Rob Shea is also not perfect in this respect. To really understand it, you've got to consider the technology: modern digital cameras have sensors that respond roughly to three different colours of visible light. The way that they do it doesn't match human eyes very well—somewhere I have commented on this, but I can't find it.
This page shows the spectral response of an unnamed camera sensor:
But that's very different from the spectral response of a human eye, here from Wikipedia's page on Spectral sensitivity:
And of course the camera curve already factors in the infrared and ultraviolet filters in front of the sensor. What do these sensors do without those filters? I don't know, and it seems that none of the video producers have even thought of the idea.
In any case, it's clear from my experiments ten years ago that all sensors are not equal, not even with my limited sample of Olympus sensors. So the real issue is how to remap the "colours” to something that humans can see. Kodak made something like that with their Aerochrome film, with fixed remapping of course. But people seem to misunderstand that to the point that one of the videos, by Vincent Versace, claimed that there was no advantage whatsoever to Infrared film, and that digital could do anything that film could. I haven't finished watching the video, and I may not: it's long, and despite the fact that it was produced in cooperation with DxO, it seems to emphasize processing software that I don't use.
So: wouldn't it be nice to find something that discussed colour remapping, in particular what mapping makes sense? Rob Shea touches on this idea, but in a manner that emphasizes process rather than intention.
Tomorrow's house photo day. I'll try a couple of full spectrum panoramas along with the normal processing.
Seen on an Emacs today:
Error running timer `mouse-avoidance-fancy': (wrong-type-argument number-or-marker-p nil)
What does that mean? My guess would be another mouse driver issue.
Do you have a comment about something I have written? This is a diary, not a “blog”, and there is deliberately no provision for directly adding comments. It's also not a vehicle for third-party content. But I welcome feedback and try to reply to all messages I receive. See the diary overview for more details. If you do send me a message relating to something I have written, please indicate whether you'd prefer me not to mention your name. Otherwise I'll assume that it's OK to do so.
| Top of page | Greg's home page | Today's diary entry | Greg's photos | Copyright information |
