debian-curiosa@ is great – some days you get manufacturer spam, others you get a flurry of dunks on some absolutely incomprehensibly wilding fucking take
bladee server
timeline is rest-less.......if i needed to allocate memory, i would simply end my program slightly later. easy enough, I'd think..
thinking of getting into NFTs (buying gold)
Replying to @bobbybobson4888
oh how i wish that were true
just reduced ZED start-up time from "over a second at 100% CPU" to "instant" by making the closefrom() not close(2) every single fd from 3 to 1024^2 😶
Replying to @nabijaczleweli
oh jesus oh piss i didn't grep for users before-hand, the fucking ZEDLET pre-exec runs this. this was usually like 1-1.2s. i literally made ZEDLETs a thousand times faster. aaaaaaaaa
Replying to @the6p4c
hm, like, i logically understood what you said, but I that hasn't prepared me to actually experience it.. what, dare I say, the fuck?
Replying to @the6p4c
also, is the text in the first installer screen also shot to hell on real Windows or is it just a Wine Thing?
Replying to @the6p4c
i clicked through the installer, but i cant read so it doesn't apply to me anyway, i'm pretty sure
im going to commit an atrocity :) https://github.com/netdata/netdata/pull/10901
Replying to @nabijaczleweli
i usually try to not be too snarky or annoying, i really do, but (a) this shit has been going on for years and netdata still hasn't managed to retain a single competent programmer after multiple rounds of millions of dollars of VC funding and (b) come the FUCK on (but -u clean!!)
Replying to @nabijaczleweli
anyway, yeah https://github.com/openzfs/zfs/pull/11834
hours spent attempting to bastardise resume-from-zvol today: 4
Replying to @nabijaczleweli
well, the resume part would work fine, I guess, if it got a correct image
but the suspend stage manages to write all the pages, but doesn't seem to actually flush the header in time, even with 0-sized ARC, 0 txg time-out and synchronous ZVOLs, so the magic's wrong, so
Replying to @marcinskarbek
i mean, so long as the image is actually flushed to disk and you import the pool read-only in the initrd, I don't (naively, mayhap) really foresee a problem tbh, or am I missing something obvious?
Replying to @marcinskarbek
(granted, it isn't, so what do I know, but still)
Replying to @marcinskarbek
i'm pretty sure the thing I'm hitting is just normal ZFS transaction grouping + sync behaviour (i.e. mark_swapfiles() (and all others) submit bios with REQ_SYNC, but ZFS only guarantees that those go into the ZIL, and the txg they end up in is not flushed by the time linux S4s)
Replying to @marcinskarbek
I don't think the ARC is going to have much of an impact here, so long as we can guarantee the txgs with the swap image hit the disk (but that'd require rebuilding kernel to add a sleep, which) – all hibernation hinges on the disk state not changing during bardo anyway, right?
[this kills the crab voice] this appeals to the gentoo user
if i, as a packet, didn't want to get filtered, i would simply not route through Berkeley
Replying to @nabijaczleweli
angle/devil/&c.
lord forgive me, but I'm going back to the old me (arguing about the GNU project on the OpenZFS bug-tracker)
bootloaders
Replying to @frameslip and @dazabani
oh god, i was gonna gag that mms is just e-mail but smaller but i didnt expect it to be the actual literal truth. wild shit
if the state is "closed" or "can't open", and the aux state is "corrupt data"
getting a tinder account and cat-fishing redhat employees so they review my systemd PRs
Replying to @nabijaczleweli
please ill do anything (okay, most things (okay, some things)), my pr has been stuck in lennart-reviewed-the-C-bits-and-someone-needs-to-review-the-shell-bits hell for 22 days now and i kinda wanted it to still maybe land in bullseye in some form but that's not gonna happen :v
Replying to @nabijaczleweli
THAT IS RIGHT I DID IT GODDAMMIT YES NUBMER FIVE ARE YOU KIDDING ME
Replying to @nabijaczleweli
i think this is the first ever resume-from-zvol, hehe https://paste.sr.ht/~nabijaczleweli/be724dc43c1178c58169b48871ecc3ed1570d291
Replying to @nabijaczleweli
the fix for resumption is so sophisticated, not even a wise-man could understand it (it's just a long sleep after swap_writer_finish() commits the last bio, so the txg with the header gets flushed before committing S4)
Replying to @marcinskarbek
uhoh, i checked private instead of public because visibilities were in reverse order; should be good now?
Replying to @nabijaczleweli
i was Not In The Mood™ for eye-parsing the resume swap header so i reimported R/W to resume, so of course it got corrupted, but just the swap dataset, and only the header, so
Replying to @nabijaczleweli
fun fact: you can resume off zram (but not loops), apparently, which means that
zpool import -No readonly=on
zfs get -Hpo value [swap dset] > /sys/block/zram0/disksize
dd if=/dev/zvol/[swap dset] of=/dev/zram0
cp /sys/block/zram0/dev /sys/power/resume
just works, no corruption
Replying to @nabijaczleweli
(uhoh, that shoulda bin "-Hpo value volsize [swap dset]")
Replying to @marcinskarbek
and immediately banned by Twitch for [redacted] because "zpool import -aN -o readonly" is valid and just means "zpool import -aN" :ж
Replying to @nabijaczleweli
here's a kernel, initrd, and instructions if you wanna replicate my results and have a spare ZFS-on-root image https://lfs.nabijaczleweli.xyz/0007-resume-from-zvol
🅱️ruh
Replying to @marcinskarbek
https://github.com/nabijaczleweli/zfs/commit/f9016b62ad614137730c6c9aa59b9f2c560bd1f3 :)
Replying to @the6p4c
if i encountered code like this i would simply CC accounting
fun fact: five golden american dollars and two months' shipping is all you need to become an enterprise in the eyes of the Dell corporation
Replying to @nabijaczleweli
No Push
Replying to @nabijaczleweli
im told that the SD card slot is for config and log transport, which means that it won't even let me run code on the third separate computer in my computer, cringe
Replying to @nabijaczleweli
if mister Dell, who I assume still green-lights all products, wasn't a coward, he'd let me run my Java applets on the fucked-up computer plugged directly into the BMC
normalise ending commit explanations with ", but, y'know"
😳
Replying to @nabijaczleweli
it wouldn't be a Wednesday if (a) this function didn't leak a FILE* and (b) one caller of it didn't exhibit a vestigial tail that can be traced back to 1999 Solaris' getextmntent(3C)
one day i will pay for writing PRs like this. today is not that day
Replying to @nabijaczleweli
i spoke too soon, i have to generate abigail ABIs and they hardcode full paths to headers for some insane reason so im on my twentieth run-around through a multi-minute autogen/configure/make storeabi/sed on buster+clang in hopes of pleasing the fedora+gcc CI, aaA
Replying to @nabijaczleweli
he's ill, please send him your most mischievous energies
if i were free(3) i would simply quietly accept and ignore strings from .rodata
bruh what the fuck is a GPL-only symbol, just resolve the damn name
Replying to @nabijaczleweli
linux sends notifications on power state transitions, including hibernation, but subscribing to them will land you in an american court, normal day
Replying to @nabijaczleweli
oh apparently its "it's GPL or it's proprietary". normal licensing stance
Replying to @nabijaczleweli
im going to alt-f4 the fsf
Replying to @nabijaczleweli
the CDDL is a free, also known as "Proprietary", licence
Replying to @marcinskarbek
the gcc multilib triple is x86_64-redhat-linux for some insane reason, i couldn't break the fedora system more if i tried
sid kernel updated to 5.10.26, exactly three net subsystem changes, but nothing's going out of my bond for some reason (but there's DHCP broadcasts going /in/), great way to be spending 5am
Replying to @nabijaczleweli
apparently someone has already hit this (no kidding!) – https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=986619 – but those one and a half hours are still wasted, because you can't search debbugs when your router has decided that bond devices don't exist actually
Replying to @nabijaczleweli
reported by a bot, posted by human who clearly trusts that bot and doesn't understand the code at hand, applied by bot, released, hell only raised two days later, reverted another two. how
Replying to @nabijaczleweli
bruh what's the actual point of spending multiple hour-long passes testing the tiniest tweak patches when you can you can just post what-ever bullshit to netdev and it'll get applied automatically and likely hit distributions anyway
lol /proc/sys/kernel/spl/hostid was broken since linux 5.7-rc1 (exactly a year ago) and nobody appears to have noticed before me, yesterday
super sucker cum jetting machine
any1 have some respect for the silence which serves as the foundation of creativity? i'm fresh out
kinda fucked up how dup2(f, f) doesn't clear cloexec on f tbh, i should start a moral panic about this
if i were an ABI generator and/or conformance checker i would simply not store the full path to the source and system headers in an XML file with sequentially-numbered attributes, thereby generating a diffstat smaller than -5k +5k for a field removal in an opaque struct
🅱️ruh
Replying to @nabijaczleweli
oh, fixed in 2.0.0 apparently
Replying to @nabijaczleweli
according to these posts, getting split-/etc and/or split-/usr hierarchies on dracut took only like six hours, which is time definitely well-spent https://github.com/openzfs/zfs/pull/11898
pacman was the introduction to vore many of us needed but didn't understand
kernel taint 😩
Replying to @nabijaczleweli
allow me to rephrase: kernel taint 🥺
we've done a.out, we're doing ELF, it's time to look to the future
buy my albums RIGHT NOW to support my endeavours to make a PE-based Debian
if i were a C compiler i would simply provide a atomic compare-and-swap intrinsic that isn't absolutely fucking useless
Replying to @nabijaczleweli
i have a "so I'm skipping the dishes, but I'm taking the garbage out twice a day" gag for this in my head but cannot verbalise it
Replying to @nabijaczleweli
multiprocessing was a mistake, we shoulda stuck to one context per physical memory region
Replying to @nabijaczleweli
actually im fucking stupid, the intrinsic's fine, but this description is criminal
Replying to @nabijaczleweli
a kingdom for a /shred/ of pseudocode, like for __atomic_xxx_fetch and friends, so it'd take less than a half-hour to find that old *ptr lands in *exp
Replying to @nabijaczleweli
github dont reset the draft PR body after submitting because it was "too long" challenge, fuck sake
Replying to @nabijaczleweli
f to pthread-mutex-based atomics, hopefully https://github.com/openzfs/zfs/pull/11904
bruh autoconf just fucking SIGABRTed the QEMU it was in, the hell
Replying to @nabijaczleweli
granted, s390x qemu isn't.. great, historically, but come on
Replying to @nabijaczleweli
works on sid qemu, as it so often does, but linux panics on boot after upgrading to current buster, so
calling mister m, who i assume owns the bank, to sell me a mortgage so i can buy a single long compute
each day i read a new file in the openzfs distribution and each day i am absolutely smashed by the contents
Replying to @nabijaczleweli
one day the thumbnail will make sense and highlight the point of the post probably! but not today
sod becoming the joker, me n the girls are becoming the antichrist
anno domini 2020 and still no fopenat(3) :(
Replying to @nabijaczleweli
why the fuck does posix require seekable files to be rewound at exit :(
Replying to @nabijaczleweli
only took 14 hours of staring at the screen, but here's zfs-mount-generator that takes 0.017s instead of 0.674s
Replying to @nabijaczleweli
a shocking improvement even on a Core 2 Duo, which, in fact, has fewer CPUs than this
inane clown pussy
Replying to @nabijaczleweli
f to a real one
Replying to @nabijaczleweli
as of this PR, I have removed roughly 1500 lines more code than I've added, heh
uhoh, fucked moment, left eye noticeably out of focus today
Replying to @nabijaczleweli
https://github.com/openzfs/zfs/pull/11917
collectible cum games
are mtg cards like nfts or summat, because jesus piss
Replying to @nabijaczleweli
searched "power9" on ebay, as you do, first result is a Rare One Copy Proper Name Magic The Gathering Card, it's a literal card, you wouldn't buy it for the art, apparently you cant even play it in most games, current bid 2900$, what da hell
Replying to @nabijaczleweli
yeah bro just drop tens of thousands of dollars to own a set of 9 jpegs and rotate them on a table its for a tournament bro
hello, how are you, please help me, i am under the splotchy
absolutely fucked how itanium died this year
Replying to @marcinskarbek
yeah, and going by debian-ia64@ it was also Very Not Good as of recent, but its still a cool Fucked Up Architecture and i kinda wanted one for that reason alone tbh
Replying to @nabijaczleweli
1 x 1.6GHz Intel Itanium single core processor!
3 x 2GB Memory Kits (6GB total)!
1 x 300GB Ultra320 SCSI 10k rpm Disk Drive!
what a beastly config
Replying to @nabijaczleweli
i should not be allowed near financial instruments, because holy jesus piss
Replying to @nabijaczleweli
sucks so so bad that low-end itanium isn't usable even as a gag
Replying to @nabijaczleweli
(yes i am also an incessant core 2 duo enjoyer, but an entire business desktop perfectly suitable for gigabit routing (so long as you crank the optimiser on rp-pppoe on x32) can be had for twenty five dollars, not twenty five hundred)
wish lamarckism was real and i could learn species traits from my cats
ability to set github prs to author+collaborators+atted people when
Replying to @nabijaczleweli
i was gonna go on but it's the SAME FUCKING DIPSHIT AGAIN
Replying to @nabijaczleweli
tell me you can't fucking read without telling me you can't read
Replying to @nabijaczleweli
notably, this fucker has opened /no/ issues, posted /no/ PRs, opened *one* "discussion" that consists of more moronic FUD about a consistency PR and ends with "no one reads the documentation." when calmly explained to by @marcinskarbek
bro. bro just shut the fuck up
Replying to @nabijaczleweli
this user has comments on 67 issues/prs, and every single one so far they got either dunked on out-right or ignored because the comments made no sense or were confusing to the reporter. why continue.
Replying to @nabijaczleweli
about to have a CoC Bad Moment because i can't go in on the guy but he seems to ignore the CoC to be annoying just fine
Replying to @nabijaczleweli
ah, no, i fucked the search, three issues, and they speak for themselves (the virtue signalling one is about removing "slave" and "blacklist")
Replying to @eng1nqueer
they all are :)
bruh what does this /mean/
Replying to @nabijaczleweli
"this URL can be loaded by a browser, therefore [????????], therefore there's a vulnerability somewhere?" i dont understand
Replying to @nabijaczleweli
am i stealing bank passwords now??
Replying to @nabijaczleweli
hey @Zekv4n mind explaining why you're scare-spamming random iframeable sites with requests of money for bogus non-existent "vulnerabilities"?
Replying to @jordiplam
oh, I know, but if the guy looked at the site for three seconds instead of spamming everything he scraped he'd maybe realise that it's bullshit for this site (and likely 99% of others that got this)
Replying to @nabijaczleweli
i usually try not to be too scathing, but I also don't usually end my mails with
> I hope to receive reward for the responsible disclosure of the vulnerability
>
> waiting for your response
>
>
> Best Regards
> White HaT
Replying to @nabijaczleweli
cant believe it took me this long to write to a CoC board (hehe), but here we are
a cat is just a small mysterious guy who loves food and long walks
Replying to @nabijaczleweli
final timing works out to 52x faster in physical time and 19x faster in CPU time on a realistic, if large, test dataset (1.195s -> 0.023s and 1.2189 CPU·s -> 0.06394 CPU·s)
Replying to @nabijaczleweli
a fully-populated 8-blade enclosure for only 700 euros!
it's full-rack 10U, and 195 kilograms
it also has six(!) 2k4W PSUs, for a theoretical peak of 14k4W.
the contract for my /entire house/ is 12kW
Replying to @nabijaczleweli
i will not road-trip to Kleinaitingen. i will not road-trip to Kleinaitingen. i will not road-trip to Kleinaitingen. i will not road-trip to Kleinaitingen. i will not road-trip to Kleinaitingen. i will not road-trip to Kleinaitingen. i will not road-trip to Kleinaitingen. i will
Replying to @nabijaczleweli
published 2002
Replying to @nabijaczleweli
published 2001, and by god someone /sure/ was
suing DHL for having a fucking MINUTE-LONG unskippable spiel about "fraudulent messages illegally using the DHL brand" followed by two short option listings that don't work half the time
Replying to @nabijaczleweli
im not even calling goddamn customer support, this is CUSTOMS
Replying to @nabijaczleweli
i would have avoided calling them but they're apparently too fucking stupid to reply to an e-mail and have sent the reply to my customs form to an "- " address.
Replying to @nabijaczleweli
HOLY FUCKING SHIT AFTER SIX FUCKING CALLS TO GET TO THE RIGHT LEAF THEY HAVE THE AUDACITY TO TELL ME THAT "DHL CUSTOMS IS ONLY OPEN 8-16 MONDAY-FRIDAY". COULDN'T YOU HAVE PUT THAT IN THE FUCKING SMS YOU FUCKWITS
why happen :(
Replying to @nabijaczleweli
(as seen in https://twitter.com/nabijaczleweli/status/1287024745272291333)
Replying to @nabijaczleweli
logging off, got epically owned!
Replying to @nabijaczleweli
:/
Replying to @nabijaczleweli
not great
Replying to @thecoshman
it's not Browsers, it's just chromium: https://github.com/thecoshman/http/issues/132#issuecomment-826390403
no way to disable writes by maintainers by default. shocker!!
Replying to @nabijaczleweli
alternatively "stop writing over my HEAD and comments challenge 2021"
Replying to @nabijaczleweli
bonus points for rewriting my comments(!?) in a way that indicates a complete lack of understanding of the material
Replying to @nabijaczleweli
another day, another guy who clearly doesn't understand that type-setting is not an art of shitting as many words as possible onto each line
bro what the fuck is a "stackless coroutine", just say you store your locals in statics, its fine
Replying to @nabijaczleweli
apparently this just means they dont have a call stack, so wrapping them in {auto ret = stackless(); coro_exit(ret);} sounds like a functional stackfulness exercise to me, but what do i know
Replying to @marcinskarbek, @9600 and @databasescaling
🥺
touch grass? more like touch ass,
Replying to @lethalbit
ah, itanium
imagine a world where you could do L2ARC on LTO-6. idyllic
Replying to @nabijaczleweli
i swear i was going somewhere with this
switches are just ASIC Ethernet miners
men will literally write two separate paragraphs and a broken PR with an essay in it then say they're "looking into a different fix" instead of just testing your PR in five minutes all told
Replying to @nabijaczleweli
i guess what this comes down to is that men don't have respect for the silence which serves as the foundation of creativity and it shows
Replying to @nabijaczleweli
every time i call DHL i manage to exchange one (1) mail. ive exchanged 4 mails to date and will be exchanging another one soon. we really do live in a future
Replying to @nabijaczleweli
fun fact: there's no way to make eBay indicate that "$" is USD on the order page, unless you change language to non-english
sometimes a family is just a barista from LA and a devil who robs gods
Replying to @nabijaczleweli
okay so i did just read kill six billion daemons in one sitting (well, got up to make tea before the last book); wish i had something profound to say, but it's apparently the dead of night now and meisterstücke require no comments
considering becoming even more annoying and vocabstituting "core" for "memory" to sound like an old UNIX manual