@rk Memetics aside, I can think of no better use-case for "¿por qué no los dos?"

@oh_that_courtney @rk I try (if somewhat half-assedly) to keep it all...it's currently 6.1MB (287K lines, half of which are timestamps) dating back to 2016-09-30...I've got a commit to my bashrc from 2013-02-30 adding HISTFILE="$HOME/.bash_hist.zev" to avoid accidentally truncating it on the occasional run with --noprofile/--norc, which I figure I must have added after a big loss, but I no longer remember what might have destroyed whatever transpired in the 3.5 years in between the two...

@rk @oh_that_courtney Or perhaps "see my life flashing before my eyes" (and flashing likely in more ways than one, considering what it'd probably do to my terminal...)

@oh_that_courtney @rk Some years ago (2017 if the timestamps in my .bash_history are to be believed) while playing around extracting real URLs from twitter's t.co mandatory-shortener obnoxiousness I discovered that sending it a User-Agent header including a vertical tab would trigger a server-side error...and it looks like it still does:

$ curl -i -A $'beep\vboop' https://t.co
HTTP/2 520 
date: Mon, 25 Nov 2024 22:44:09 GMT
content-type: text/plain; charset=UTF-8
content-length: 15
x-frame-options: SAMEORIGIN
referrer-policy: same-origin
cache-control: private, max-age=0, no-store, no-cache, must-revalidate, post-check=0, pre-check=0
expires: Thu, 01 Jan 1970 00:00:01 GMT
server: cloudflare Pingora-Origin
cf-ray: 8e85295fbde62838-SEA

error code: 520

So who knows, perhaps they're using it for some in-band signaling roughly along those lines?

Seriously, a truly impressive mustering of biomass.

@regehr @pervognsen @zwarich Hah, another thing I hacked on in grad school! It would do online perf profiling to discover hot paths in the kernel (geared towards servers running relatively steady, homogeneous workloads), and then recompile a specialized kernel module containing all the code for the path(s) it decided to target with LTO doing hyper-aggressive inlining (including across speculatively devirtualized indirect calls between modules) to provide a single contiguous code path all the way from the syscall entry point down to all the relevant device drivers and then spliced it into the running system as a livepatch.

Had it been a few years or so later it might have seen better results due to the Spectre mitigations that later became necessary slowing down the "before" case more...

@regehr @4raylee It's not much, but in grad school I wrote a benchmarking program that JITted a small sequence of code in order to precisely control both the I-side and D-side cache footprints of the "think time" code executed between issuing syscalls...dusting off the cobwebs: https://gist.github.com/zevweiss/b9da17b661e35c4cabab2cc6e8330df8

@cliffle

If you represent byte sizes as floats I do _not_ want to use your: filesystem, hosting service, network appliance

Okay, that's understandable I suppose, but it makes perfect sense in a programming language interpreter, right?

https://github.com/php/php-src/commit/405a15043f89af7aafdf3975db84059093f0ecdc

re: Uspol

@rk I've had a bookmark of this one pinned to the home screen on my phone for the last year or two, revisiting it not infrequently. It only seems to get more and more vivid.

...and again a year later. It's a good tree!

@arj Even by amdgpu standards that does seem ponderous...did you find a culprit? Curiously, I just did a fetch to my local tree (the first since late May, apparently) and only got about 20% as much:

Receiving objects: 100% (145473/145473), 88.16 MiB | 5.32 MiB/s, done.

@rk I mean, the point is obviously to run as many instances of said text editor simultaneously as possible, right?

@rk Or if you've got sshd listening on a publicly-accessible port, /var/log/btmp. Mine's currently sitting at 726MB.

(I should probably set up some log rotation.)

@chalpin While it's now been over 7 years since I had any even tenuous, tangential involvement with ATE, I still get my daily reminder of it every time I open the dog food bag to get Bowie his dinner.

@kwf Another option is a skirt of sorts to fill in the gap between it and the floor to fend of incoming tennis balls and such (I ended up doing essentially that with our couch a few years ago).

@mattblaze I'm curious -- is that a lighthearted humorous observation of a pattern you see in your photos, or something you've done consciously & intentionally?

@benjojo yeah, that strikes me every time I'm there -- definitely one of the most off-putting aspects of the city, IMO.

@regehr If you're into this kind of heat nerdery and don't already have one, I highly recommend acquiring a thermal camera. Fancy new ones can be kind of pricey, but with a bit of craigslist-watching I snagged a decent-enough used one for <$100 recently.

All sorts of commonplace things are interesting to look at with it, like your "footprints" after walking across a floor barefoot. It's also useful for all sorts of miscellaneous troubleshooting (home, car, electronics...) and taking thoroughly eery photos of your pets.

@shenki Something vaguely along those lines, still sort of figuring it out though...

Decently pleasant view from the office window, with fall foliage colors already beginning to show.

@tj dc(1), or shell functions built thereupon:

$ type rpn rpnf
rpn is a function
rpn () 
{ 
    dc -e "${*//x/*} p"
}
rpnf is a function
rpnf () 
{ 
    rpn 10 k "$@"
}
$ rpnf 329447664 9672 811 x /
42.0000000000

@jk But is the third one a slightly larger lot, or just a worse deal?

@cks @0x0ddc0ffee @drscriptt

In theory I guess you could have the (soft) power switch wired to the BMC as basically a GPIO pin and then the BMC controlling the 'power switch' wired to the PCH or wherever it would go

On the platforms I've developed BMC firmware for that's pretty much exactly how it's arranged. Yes, it's more moving parts between the physical button and actually powering on the host, though often mitigated somewhat by a "GPIO pass-through" feature of popular BMC SoCs (Aspeed parts, at least), wherein you can configure the GPIO controller to basically just bridge two pins together in hardware so there's no software in between and it acts like a direct connection, which can be useful for times when the software that usually manages those GPIOs isn't able to do so for whatever reason (like during the BMC's boot sequence).

Some rocks just don't want to split even after extensive sledgehammering. Fortunately with the right blade a circular saw can make it a little easier.

Also, wow that's by a wide margin the fastest I've ever drained a 4Ah battery (and given that I guess the thermal self-protection shutoff shouldn't have been a surprise).

The verdict: very effective, would rent again (and probably will for the other half of the yard).

@kwf These pictures made me sort of curious about that tool and so I watched a brief video review of it, which happened to mention that the blades are reversible -- perhaps you're already aware, but might they still be intact enough to flip around and keep going?

@dannyjpalmer Oh wow, I hadn't thought about that game in a very long time! And I don't recall ever noticing the tagline on the paper, which is quite excellent.

@mxshift @danderson Can I ask what you used them for? "GHz-capable discrete logic gates" sounds like an interesting space to be in.