@lorddimwit Please tell me you're not building a third-party implementation of AppleScript.

@arj I think I'm gonna go erase some flash chips just so I can reflash them again.

@tedu I haven't used one in any seriousness in some years now, though from the last time I did I recall some wistful sadness at the general feeling that macports was a dying project, losing out to (the IMO markedly inferior) homebrew.

From a skim of those comments it sounds like maybe macports has had some degree of resurgence though? So if in the future I am in fact ever compelled to use a mac again, perhaps at least macports will still be a viable option...

@robey @swetland I use insert heavily! (Shift-insert produces the equivalent of a middle-click to paste the current primary selection text, at least in my terminals.) I also occasionally use a bare insert to toggle overwrite-mode in emacs, but that's admittedly pretty rare.

@robey [Nods emphatically] Yeah, Framework's product announcements have really thrown into stark relief just how stagnant and boring the rest of the consumer-electronics industry has gotten.

@Obdurodon The slope of the computer-improvement line has really leveled off -- I'm typing this on a computer that's 10 years old, and in 2023 that's basically fine, and aside from some relatively minor performance differences not really noticeably "old". Running a 10-year-old computer in even, say, 2005 would have been laughable.

@tedu But are the files still named *.jpg? Apple seems to like doing that.

@benjojo @electronic_eel Entirely fair, though if you do ever change your mind and want to pursue that, I've done a number of ASRock OpenBMC ports (and maintain the meta-asrock layer in the OpenBMC tree) and would be happy to offer assistance!

@tedu What about "defun" -- does the negating prefix absolve it? (Okay, not strictly a keyword I suppose, but the moral equivalent.)

@lorddimwit

YEARS OF PROGRAMMING yet NO REAL-WORLD USE FOUND for applying an operation to anything more than a PAIR OF VALUES

My ?: operator would like a word!

@arj "Perfection is achieved, not when there is nothing more to add, but when there is nothing left to take away." They've finally done it!

@stephen0x2dfox Proprietary BMC firmware (especially from a particular three-lettered vendor, it seems) has been and likely will remain for as long as people care to poke at it a wellspring of easily-exploited vulnerabilities, yes.

Open-source alternatives are, I'd hazard an educated guess, likely to be substantially better in that regard, however. I'm hedging a bit here of course because there's still lots of non-memory-safe code involved (though some of us are working on getting Rust integrated), but (a) while there are still likely to be remotely-triggerable bugs, I'd guess they're probably somewhat less numerous/egregious, and (b) enabling some basic protections like ASLR and a non-executable stack at least raise the difficulty level of writing exploits a bit.

@leah @benjojo Indeed -- much like git's index, the exposed distinction between buffer content and file content (and explicit synchronization of the two) is while arguably an implementation detail that "should" (some think) be hidden from the user, also a very useful tool for users who understand it.

@erincandescent @benjojo You can also disable flow control with stty -ixon, reclaiming Ctrl-s/Ctrl-q for one's own purposes -- by default in bash, respectively forward-search-history (like Ctrl-r but in the opposite direction) and quoted-insert (handy for inserting literal tabs and such, but also available via Ctrl-v).

@arj Oog, maybe toss in technical writing too while we're at it!

@mxshift @fasterthanlime @sigmasternchen See also

@lorddimwit @todb Oh wow, is that EtherApe I spy at the very end (t=60s) of that clip? Judging by the release date of the film and when I was actively working on that project it's borderline, but maybe just maybe there's a little bit of my code in a movie! Guess that means I might have to watch it, despite the (predictably) terrible reviews...

@arj Should we add file formats and device register interfaces to that set as well? (Or maybe they're already included for a sufficiently general definition of "protocol"...)

@tedu Well, according to the coverage pie charts, 91.4% of that module is either fully or indirectly documented, so I think you just need to work on your reading comprehension?

@Obdurodon Wait, but doesn't "vanished" imply that it was actually here at some point?

@bascule @djm While perhaps technically a slightly different case, a trailing slash can also determine whether a symlink is dereferenced or not.

@monsieuricon @gregkh How about a tarball?

@monsieuricon I haven't noticed any other problems myself, but would whatever this is referring to have anything to do with me receiving 40 copies of this email from @gregkh?

@Obdurodon Ah yes...spend much time in Wisconsin and you'll doubtless (if you haven't already) hear grumbling about the FIBs and their driving.

@corbet In fairness, it sounds like (if I'm understanding both you and the linked post correctly) they'd acknowledge it as a compiler bug, but as a "non-security issue", no?

I wonder if @mattgodbolt might have some thoughts to share with them here though...

@tedu So here's an odd one I've noticed from time to time (hopefully not spoiling the streak)...

With my browser settings (firefox), control-click or middle-click opens a link in a new background tab, and adding shift to either does so in a new foreground tab.

With some links in honk however (e.g. convoy links and "@me" in the menu, but not "chatter" or "tags"), middle-click operations work normally, but the control-click ones instead navigate to the link in the current tab instead of opening a new one. Any idea what might be going on there?

@tedu Ooh, nice! I recently added some basic hotkeys to my version, which I've quite liked for desktop usage. (This is approximately the sum total of my knowledge of web development, and so may be full of embarrassing errors, but seems to work well enough for me.)

Attachment: honkhotkeys.patch a patch adding hotkey support to the honk web UI

@stephen0x2dfox @solene FWIW, git send-email can be used to send arbitrary (not necessarily git-generated) messages -- maybe slightly easier than learning a whole new full-featured client.

(Assuming google's not also mangling things server-side, but I think it's pretty widely-used with gmail so hopefully not.)

@tedu @eta Heh...I've only ever used the "retrieve this thing that didn't get sent to me" functionality of it, and had actually forgotten that that page had anything but that little form on it.

Or on second thought, maybe just JSON in general...possibly a competitor to C++ iostreams in the utilization-to-merit ratio category?

@tedu Indeed -- Jira is a major offender in the "going out of your way to break shit that works by default" category, and much like in performance profiling, death-by-a-thousand-cuts is a very real phenomenon.

@lorddimwit Earlier this year I opened the door to our (detached) garage to do some laundry and saw a black feline streak (the neighbor's) dart out...I think she'd been stuck in there over night after sneaking in unnoticed, but considering how often we go a lot longer than that without opening the garage, I was quite relieved it wasn't far, far worse.

@penryu @foone Thankfully in most browsers you can still "force" a proper control/right-click via shift-right-click, which bypasses javascript interception (in AFAIK a non-fuckuppable way).

It's irritating that it's needed as often as it is [glares at Discord], and I'm just waiting for the day when the browser makers decide to allow JS to intercept the event anyway, but at least for now it's a useful talisman against the garbage-flood of the web.

@tedu Most people's browser tabs probably aren't on the lean, mean, well-oiled machine that is #honk.

@lorddimwit Torx, Allen, Robertson...really basically anything but Philips (or more generally anything non-cruciform; pozidriv emphatically does not get a pass here).

@GeoffWozniak While I agree on the whole, in fairness I think that's less a macOS issue and more a CoW-filesystem issue...(I'm pretty sure btrfs can end up with the same problem, for example).

@regehr @munificent Hard agree -- the one semester I taught, despite it being an OS course I basically just treated it as an opportunity to pound "data structures are more important than algorithms" into their heads as many times as possible.

@Kazinator Neat, but seems to break when $PWD only contains a single / (status line doesn't show after e.g. cd /home or cd /).

@andy_warb Yes, sudo !! is the popular answer here, but I'll suggest the alternative of Ctrl-p Ctrl-a sudo<space> (alternately up-arrow home or the like) to go back a line in your history and then position the cursor at the beginning of it so you can just prepend the sudo part. Same number of keystrokes, but lets you see the whole command instead of semi-blindly executing things (personally, that aspect of history expansion scares me, so I turn it off in my bashrc).

@kaoudis @regehr

it's pretty hard to disguise an unintentional infinite loop

Am I misunderstanding your meaning here? Things getting stuck in unintentional infinite loops is IME while not the most ubiquitous bug category in the world, also not a particularly uncommon one -- I just recently hit one in mandoc for example, and after a brief glance at where it was happening in the source it was certainly well-disguised enough that the fix wasn't obvious to me.

(Sadly my efforts to report said bug ran up against the wall of a moderated mailing list that never accepted my request to post to it, but alas...)

@Obdurodon ...and every time anyone indicates they put any stock in it whatsoever I will continue referring them to this excellent demonstration from the late great James Randi.

@cks @jef True, I suppose it's nearly always going to be a sequence of hits in the local dentry (or whatever) cache; I guess it just seems like a bunch of pretty redundant syscalls -- the kind of thing that makes me go "ugh, so much wasted work, why isn't this written more efficiently" any time I look at something under strace.

@cks @jef In interactive mode maybe? In plenty of script contexts I could imagine that causing substantial performance regressions, especially if $PATH contains, say, a bunch of NFS directories followed by your basic local ones and you're executing things out of the latter in a loop.

Granted, I not infrequently run shell loops of non-trivial iteration counts in interactive mode too, so I might still want it even there.

@regehr I recall an xkcd about this...

Lately I've also found myself often thinking about the frightening potential of various other forms of stored energy, like big springs in car suspensions and garage doors.

@cliffle Same here -- only thing holding me back from preordering one myself is that I already have a Framework 13 (which I love!), and while I'm sure the 16 would be very cool, I've already got...[counts]...five (I think?) other laptops lying around, and I'm having a hard time justifying another.

@Obdurodon @TindrasGrove Heh, I've probably used those as makeshift bottle openers more times than I've actually hung clothes on them.

@benjojo What sort of sequence would lead to that path? It looks like mainline honk has a nil check there, but it leads to a fatal error (perhaps slightly better than a nil-deref crash, but not ideal)...should a similar patch go upstream?

(CC @tedu)

@tedu I recently played through Comet 64, which is basically just a sequence of fairly simple programming puzzles, except the machine is constrained and bizarre enough to make it tricky at times (at least if you want to get all the line-count/cycle-count optimization bonuses). How about one register per data type? (String included!)

@sjb @tedu I believe in Haskell this is referred to as "tying the knot".