My cats love me, and I love the rain.My cats, alas, do not transitively love the rain.
Yesterday, I rearranged my apartment a bit, to better use the excess of space I have. The back room is now a bedroom (which actually works better, because I like feeling like I'm outside when I sleep, and the back room has large windows on two of the walls), and the middle room is now a room for getting dressed and extra books, with a beanbag for lounging when I don't want to lounge in the front room. .. Hmm. Honestly, the middle room is now underused for its size, but I don't think there's any way, unless I ever get a significant other and they move in with me, that it will feel well-sized. The new arrangement does feel more spacious and better suited though.
One of the machines at work has been showing very odd behaviour -- the kernel occasionally complains about getting NMIs (dazed and confused message), and suggests the RAM is bad. After a long process of weaning people off of it (it was one of the main compute servers), I ran memtest86 on it, and found errors near the end of memory. I took out the SIMM that it should be on, boot back up, and rerun memtest86, and it looks like it's on the other chip. I shutdown, swap chips, and it's still there, near the end of the second-last bit of memory, no matter how big RAM is. Strange. My temporary solution is to put both DIMMs in and tell Linux to ignore the last bit of of memory. The box is undergoing a stress test, courtesy of a nice program called stress. If it can go for a week without that message showing up again, I'll return it to production. I have two theories as to what might be going on. First, it's possible that something is wrong with the cache or MMU. I don't know how the cache is organized on that particular system. Second, it's possible that there's some hardware that's doing IO mapping there (or, alternatively, some hardware device stealing from main memory, as AGP does). The second seems unlikely -- I'm actually unsure if PCs support memory mapped IO, and I imagine any device that would be using main memory in that way would probably encounter a lot of problems if the system is using it as main memory too. I would further hope that Linux would notice such things and not get in the way of such use. It's still, of course, possible that the memtest86 failure is a red herring -- perhaps memtest86 isn't as hardware-aware as Linux is, and Linux already knows not to use that RAM for anything. In that case, presumably the NMIs are unrelated to this failure. Hmm.
I found myself undoing the number of hacks I used to do to disable Unicode on my systems, largely because I'm starting to tire of needing to romanicise everything that should be written with Kana or Cyrillic.
There's something .. different .. about the idea of growing up with subway and bus systems, taking them year after year, from childhood through adulthood. There's a kind of familiarity with the completely public that somehow fascinates me, moreso by the way that they provide time for a certain kind of reflection.