< | >

Hacker's Diary

A rough account of what I did with Emacs recently.

June 21
Server hijinks... in the process of trying to add a new drive to my EC2 instance (this server!) I almost wiped out /home and had to do an emerency shutdown to clean up. When the server came back, dovecot wouldn't run, claiming that all its ports were already in use. After much digging around in config files, turning things off and on, and even using strace to see if I'd somehow configured it to fight with itself, I discovered portreserve, installed May 2020 as a dependency of dovecot, which apparently was added to the dovecot bundle without including the necessary integration to stop portreserve from hanging onto the dovecot ports when it was starting up. Yeesh.

While digging around in the pile of still-not-unpacked things I found my old, OLD iPod. It's not a first-gen, but it's still pretty damned old. Plugged it into the laptop, loaded up Music, and there's a nice option there to randomly fill the iPod. Clicked on Sync Settings and... a Finder window appeared, with most of what looks like the old iTunes UI features, and a conflicting UI for filling the iPod. Er. No idea what happens now.

June 14
More migration fun: MySQL 5.7.15 on old machine. I want it to be on the new machine. Setting up replication seems the obvious approach to do a clean switchover. Installing MySQL 8 (which is the version after 5, because Oracle never really got the hang of the whole version thing) on the new machine was evidently the wrong thing to do because there was apparently no way to make it work. Currently trying to just clone the install from one machine to the other and use that as the basis for replication. It's really tedious that I can't simply set up a replication server by creating the necessary conditions for replication on the primary - essentially, a server ID and a replication account - and then pointing a suitable secondary at it and letting it do whatever it needs to do to get up to speed. Like, you know, bulk download all tables while not accepting any connections, then switch to online replication mode, instead of making me do all of that manually.

Ok, that actually worked, with only a minor glitch (the command at the start of the database dump which synchronises the secondary with the primary didn't work, for some reason, until I ran it a second time. I guess I didn't really mean it the first time).

Department of What Fresh Hell Is This?: having successfully moved everything in the network to sit behind an Airport Express with the upstream cable modem in bridge mode, it appears that something on the Airport runs out of memory periodically, but not badly enough to kill the device, just enough to stop it from functioning. I don't actually want to purchase more hardware, so I may wind up strapping some sort of watchdog reset onto this, but this is how I discovered that Apple no longer sell wireless access points, and in fact stopped doing so in 2018. Disappointing, since I only just recently discovered that the Airport pops up in the Server app that comes with macOS Server, so you can enable things like port-forwarding for specific services without using the Airport Utility. The Airport's logging is unhelpfully configured to not give you much useful information at all (timestamps, process names or IDs) so I've resorted to running top in a terminal session to see if I can see what eventually eats all the RAM. Best estimate is that it'll take a week or so to break.

One more piece of weirdness fixed but not understood: moving OpenHAB to the new box while MySQL is still on the old box caused an offset of 3.5 hours in the timestamps. Telling OpenHAB to use database server time rather than its own time - and OpenHAB's idea of the time looked perfectly fine in the logs - fixed the problem. NO idea what that's about. There's an old, well-known variant of this from my early days at the current job where something was persistently off by 6.5 hours, and the reason was that one thing was printing out times in "Irish Summer Time" - IST - and something else was parsing that output and interpreting IST as "India Standard Time". Ireland/India time difference: 6.5 hours. Doesn't explain where this 3.5 is coming from, though, or why.

Virgin Media connection bounced at least three four times today, including twice while I was writing the above just now. I do hope I'm not going to have to navigate their technical support again.

June 13
Needlessly tedious task: flipping a secondary DNS server to primary on macOS. The zone files appear to be some sort of binary thing, so after trying the obvious approach of stopping the server, editing the named.conf, and starting the server again, I copied the zone files from the primary and then poked at it until it worked. Not sure what the problem was, and the Server app seemed inclined to hold onto "primary" status past the point at which the underlying server had already flipped. Anyway, done. I've two more things to remove from the old Mac Mini and then the new Mac Mini will have fully replaced it. (the old Mac Mini is a 2009-ish refurb, the new Mac Mini is a 2011-ish refurb, so none of this stuff is new new).

Oh, REALLY annoying. Checkbox to allow zone transfers. Click it. Save it. Reopen the zone. Box is unchecked. Are there any errors logged anywhere to tell you why? No. Edit the named.conf file manually. Reload the server. Check the zone. Box is unchecked again. I mean, to be fair this is a temporary measure while I move everything off the old server, but shouldn't this stuff either Just Work, or give you a hint as to why it doesn't?

June 12
While trying to add some code to cater for poorly-constructed RSS feeds back in May, I accidentally broke the collection of feeds which were poorly-constructed in a different way (no, this code has no tests, it's an overgrown multi-year hack that I just sort of periodically kick until it breaks, then continue kicking until it works again). One of the affected feeds was BBC News, and that meant that for a couple of weeks - until I noticed the problem - I was getting none of their news stories. Looking through what they're publishing since I fixed it, I realise this is maybe not such a bad thing, because there's a streak of... I dunno, condescenion, maybe inanity, maybe something else in the headlines, never mind the stories. It doesn't help that the feed I'm getting is a blend of actual news headlines and "magazine" pieces, nor does the fact that - as previously mentioned - they have a tendency to rewrite certain stories' headlines repeatedly.

I might just switch it off again, but deliberately this time. The Guardian suffered this fate previously, as did The Irish Times. That would, however, leave me with RTÉ as my only news source, and that's not an idea that thrills me.

June 11
Well, that's annoying: as far as I can tell, Pick (one of Sky's six billion TV channels) put two ST:TNG episodes out of order, which the Virgin digibox then took to mean that the season had ended, so it didn't record tonight's episode. Which means that we can record Unification, Part 1 but we've lost Unification, Part 2. On the plus side, the Horror channel is zipping through Season 1 at a rate of knots so it should eventually catch us up to the missing episode.

This means, btw, that we have yet to watch an entire season of ST:TNG without the digibox screwing us over in some way.

Oh yeah. The old digibox was supposed to be collected yesterday. It's still in the front room. Such a great sense of customer care.

June 3
Virgin Media, bless them, decided we didn't actually need TV or broadband at a time suspiciously close to 15:30, almost as if something scheduled had occurred. We have a fallback cellular broadband gadget but it's a bit wimpy and the bandwidth isn't great so it's not exactly work-from-home material. Virgin's site helpfully informed us that there was a fault in our area and that things would be fixed "as soon as possible" by which they meant "20:00". As it turns out, various lights came back on at 18:00. The company I work for tends to respond to this level of service interruption with a public statement - sanitised and approved, of course - of what went wrong, even if it's due to pilot error, and sometimes there's a followup with more detail. My ISP? Even if I asked for a summary I wouldn't get it. The TV's working, what more do I expect?

Speaking of De Telly, we've been tracking our way through Star Trek: The Next Generation, mainly at the suggestion of Mrs. Waider after we'd watched Picard and because there are not one but two channels showing reruns of the series. We had a minor upset when we swapped out the Virgin kit a few weeks ago - a few episodes on the DVR are still on the DVR which is now defunct and awaiting collection, because once again Virgin don't much care for their customers beyond squeezing a monthly sub out of them - but we've seen most of seasons 3 and 4, we've made inroads on season 1, and last night watched the first episode of season 5. I was aware of this the first time around when I was in college, but didn't generally have access to the appropriate TV channels to watch it and in any case I've never been a huge Trekkie, but I have to say I'm generally enjoying it. Aside from the occasional awful clanger like the episode where Geordie creates a holodeck representation of a woman he's interested in, imbues her with his impression of her personality (which of course is sympathetic to him) and then the woman in question encounters the holodeck version and a small amount of fireworks ensue. But really, what happens? Does Geordie understand that he's being creepy and wrong? No, he makes an impassioned speech about how he's just "reaching out" to this woman, who then apologises to him. Cringeworthy, and thankfully the script generally aims a bit higher.

previous month | current month| next month