Notices by LEdoian (ledoian@pleroma.ledoian.cz)

@feld @lanodan I (my evil-ish part) sees this as an opportunity. CrowdStrike is bad, but corporates forget and people possibly too (“just a bunch of stuff didn't work for half a day 🤷”).

On the other hand, getting a lot of gamers around the world completely mad and vocal about it might actually be more memorable and push the ecosystem from this kinds of stuff.

@lanodan @nosleep yes. :-D

@lanodan I got the e-reader as a gift, so I didn't have much say about it. So I only use it for epubs and the subset of PDFs that fits the screen, and it is quite good at that 🤷

And I think there might be an alternative firmware (it's a random Pocketbook), but it is not a pressing issue for me atm :-)

@lanodan I don't know whether I have tried mcomix yet. I experimented with readers some time ago, zathura's cb plugin works good for me when on laptop, and on phone (PostmarketOS) I ended up using Okular, both of these were quite usable (but I don't have too sophisticated requirements).

My e-reader is too lame, both in screen estate (manga mostly does not fit in any reasonable way) and in firmware (no cbz support) :-/

@lanodan TIL mpv is a cbz reader!

@hj sounds relatable

@lanodan @lcamtuf I meant, if I were to implement switch in assembly, having a table determined by labels and indirectly jumping through it would make sense (and there are already examples of such tables).

I didn't expect that this would even work in C in any way (standard probably does not allow manipulating labels at all), though it is not surprising that GCC exposes this functionality (given many architectures support indirect jumps), TIL! I weren't aware that the original post was actually a valid gnu-c code…

@lanodan @lcamtuf Well, interrupt handler tables are basically this, right? And user-space wise, dynamic linker does iirc something similar. (I know those are calls, not jumps, but that should not matter much).

And in a quick glance, this looks quite readable too.

@lanodan Suspended my laptop several times accidentally just by putting a phone between the keyboard and the hinge. I guess the first time was confusing, but then I started doing that occasionally on purpose :-)

@lanodan @kirby Intuitively, I thought that this would be mostly needed for nontrivial immediate values in the instructions. But now I am not sure: prefixes, instruction "name", up to four arguments (incl. weird addressing modes of x86_64), it might add up.

A quick disassembly of /bin/df shows that 9-byte instructions are rather common:

I think this is some security guard or sth; the last 4 bytes look like an imm32 value. I am not sure whether "full" 15-byte instruction exists, though…

(Looked at some AVX instructions, they take ~7 bytes. Maybe some horror like VEXTRACTF128 might be long, it has 3 arguments, idk)

@lanodan @piggo Whenever I use manpages, I use man -a and quickly skip from GNU pages to the POSIX ones. (For section 1, that is; OTOH system APIs often read better from Linux man-pages than from POSIX.)

I quite hate how GNU manpages mostly give the incantation for the actual documentation (Full documentation […] available locally via: info '(coreutils) ls invocation'). And for stuff like make(1), GNU manpage is pretty much a complete joke (lists only flags for the command, not the language).

But still for common invocations, I find tldr is usually quicker than the manpage.

@lanodan @piggo tldr is priceless for the common magic incantations. I use it mostly for options to dig and sometimes rsync, but there are definitely many tools I only use rarely (join, paste come to mind)…

@hj @kaia @SuperDicq @magmaus3 @mischievoustomato I find GIMP's UI a bit worse. I wouldn't classify using either as pain, but in Inkscape I imagine the sets of SVG objects and usually am able to do what I need. I can do a lot in GIMP too, but some things require tricky operations (like actually drawing a circle).

@lanodan I wouldn't probably support the code removal so directly. There is something to learn from the issue and it's not like there would not be vulnerable code on GitHub anyway (you can get e.g. Linux kernel v5.12-rc1-dontuse if you want). Imho the issue is not "someone could see the bad code", but "somebody could get it by mistake", at which point a better way would be ~making sure people can only download the code interactively and with seeing a big warning for some time.

But GitHub needs to be on good terms with their shareholders (not community), so it would make sense they would remove it preemptively. Also, if Jia got enough rights over the xz repo, they might have killed it themself, no GitHub employee involved.

@lanodan @SuperDicq That is the reason I stick to vim and don't use nvim: :! shows the previous terminal, does not create a new one.

@lanodan @chjara @alexia @thelettuceman Doesn't matter which one, as long as the results can be used "complete[ly] automatic[ally] […] to tell computers and humans apart", no? :-)

Reminds me: https://www.commitstrip.com/en/2015/07/06/meanwhile-in-a-parallel-universe/

@kaia According to this article (Czech only, sorry), there should be an option for 50 more years. So we're safe :-)

The same article describes how Czechia has one lot permanently (if I read into it correctly) and that we are trying to exchange the two leased lots for another lot/s (there are too low bridges, so it is not very usable). The exchange could happen this year.

I didn't factcheck it, just remembered reading about this ~recently.

@Moon This actually happens to me quite often (EeePC T101MT). For me it seems to be related to hibernation and feels quite harmless, so looking into that is very low priority.

The only annoying thing is that this message gets to all my open terminal windows.

@lanodan Also it made me mad that firefox would "autocorrect" the domain names, like .vom → .com, or try automatically adding www prefix. I think it used to do that even if the changed site didn't exist, so it only made things worse…