Your twitter archive

this merges back into the main rp5700 tweet-set since that accursed platform has decided to cooperate with the NIC: https://twitter.com/nabijaczleweli/status/1249886190041522176

 Fri May 01 15:44:50 +0000 2020

the BCM5720 subset merges back here, following its refusal, convincing via forum, and decision to function, instead of breaking the platform: https://twitter.com/nabijaczleweli/status/1249847338568503303

 Fri May 01 15:47:02 +0000 2020

cat (splotchy, left) meets local menace (right)

 Sat May 02 00:01:31 +0000 2020

rp-pppoe code is indented with 8-wide tabs for even indents and 8-wide tabs + 4 spaces for odd ones. also labels are half-indented

how does old code always manage to be Like This; it's truly incredible

 Sun May 03 00:05:14 +0000 2020

first-version patchset prepared (https://github.com/nabijaczleweli/rp-pppoe/tree/patches/0002-pppd-bugfixes-3.13-v1), probatory mail posted, let's hope Dianne wants my patches at all (but, well, if she doesn't, then that's extremely understandable too)

 Sun May 03 23:47:43 +0000 2020

get it (and please test, if you can!) while it's hot!
rp-pppoe 4.14-BETA-1 is out with (most) pppd bugfixes from over the years: https://dianne.skoll.ca/pipermail/rp-pppoe/2020q2/000578.html

 Mon May 04 15:31:10 +0000 2020

a grand old time with graphviz (presented by: parsing RFC822 with awk)

 Tue May 05 03:41:34 +0000 2020

full-size; what a beauty

twitter doesn't want to show me a preview of this 18549x13860 image for some reason

 Tue May 05 03:44:06 +0000 2020

oh it's just black for some inexplicable reason; well, here's a screenshot because debugging twitter at 5:45am doesn't sound fun

 Tue May 05 03:46:08 +0000 2020

woken up by some goblin (right)

 Tue May 05 11:17:41 +0000 2020

and so, the sundance NIC I thought was a modem, for which I specifically picked this machine over the others, shall go

 Tue May 05 18:50:34 +0000 2020

love this massively overbuilt riser, such a delightful texture

 Tue May 05 18:58:16 +0000 2020

an unpopulated jumper between the B51 "AD[01]" pad and a squiggly on the other side?

 Tue May 05 19:14:25 +0000 2020

the blanks for the long slots have an additional pin that fits in the chassis so they don't fall out when not clamped by the swingy detent that holds them in

but this also means that they require a specific profile on the end since they're not screwed in, which, well

 Tue May 05 21:25:12 +0000 2020

this is the only of five blanks I had that fits, and it's still wobbly at best

it still falls out whenever you open the flap, and is very finnicky to get in properly; oh well

 Tue May 05 21:27:59 +0000 2020

twenty twenty: three gigabit NICs, a PS/2 mouse and keyboard

also the finish on the blank is completely different from that of the rest of the chassis, which is neat

 Tue May 05 21:33:46 +0000 2020

holy shit debbugs handles nested mails *so well*; I'm genuinely astounded

 Wed May 06 02:37:59 +0000 2020

gotta say, these do look incredible on a nice 4k display

 Wed May 06 19:36:51 +0000 2020

need this on a wall tbh

 Thu May 07 00:13:06 +0000 2020

yasssssssssssssssssssssssssssssssss

screenshots from 4k, since the only thing that managed to display these is feh

 Thu May 07 00:29:08 +0000 2020

had to turn it sideward to get any detail (and add a swap since that computer only has 8G of RAM and feh was getting OOM-killed, turns out swap files on NTFS work), even at that size, but fuck if it doesn't look good

 Thu May 07 00:32:44 +0000 2020

you vs the lass she tells you not to worry about, &c., &c.

 Thu May 07 00:33:44 +0000 2020

memory size changes are detected, but, Monitor Tracking doesn't seem to do anything?

 Thu May 07 02:19:14 +0000 2020

interesting for "BCM5755 [which] is a fully integrated 10/100/1000BASE-T Gigabit Ethernet media access control and physical layer transceiver solution" to say it's 4Gbps, wonder why that is?

 Thu May 07 02:31:27 +0000 2020

 Thu May 07 02:34:32 +0000 2020

pictured: me, in the very near future

 Thu May 07 03:07:41 +0000 2020

well fuck you too, twitter

 Thu May 07 12:12:36 +0000 2020

looks like it's because networkctl does this; not sure why there's no ethtool info for speed and duplex yet

 Thu May 07 15:55:05 +0000 2020

ethtool doesn't return the register data or the EEPROM from the card in question (eth2 here, 5755 PHY) either, but it does for the other two tg3-based interfaces (eth[01], 5720C PHY)

going by tg3.c, it looks like the 575X line is particularly Good™ so that's fun

 Thu May 07 16:10:23 +0000 2020

i'm a fucking moron, took me until reading this segment to realise it's unknown because there's no link negotiated :vvvvvvv

 Thu May 07 17:23:06 +0000 2020

well this is decidedly worse

 Thu May 07 21:11:14 +0000 2020

yassss

building with the right version of the patch works better :Р

 Thu May 07 21:59:01 +0000 2020

considering going absolutely feral and getting a dot-matrix for these

 Fri May 08 00:17:13 +0000 2020

PR: https://github.com/systemd/systemd/pull/15750/commits/7b4e10fc3d1c62152bb9d284d189f1f1f2747bfa

this bug was pleasant to solve, I really enjoyed it

 Fri May 08 01:55:05 +0000 2020

had a fun time today :v

 Fri May 08 19:08:46 +0000 2020

ever got hit in the head with the steel reinforcement of a garage door panel as two of them fall hinge around you? that was fun

 Fri May 08 19:10:38 +0000 2020

cameo

 Fri May 08 19:29:42 +0000 2020

👀

 Fri May 08 19:39:05 +0000 2020

heh, looks like i have a propensity for running into neomutt corner-cases

 Fri May 08 23:39:49 +0000 2020

by no means is this an indictment of any sort, I wouldn't dream of it; if anything, this speaks volumes to how much dumbass shit you can do with the thing before minor seams start showing

plus, all the neomutt team are absolutely lovely, so

 Fri May 08 23:53:42 +0000 2020

well i mean yeah but why in gourd year twenty twenty are /display names/ 7-bit-ascii-alnum and not, like, Unicode class L|Nd, at least (and is it the anglocentrism)

 Fri May 08 23:58:59 +0000 2020

was thinking of making a twit-to-mbox thing but (a) twitter is a pain in the dick, (b) TOS?, and (c) i use it for pictures mainly and those don't work satisfactorily on a teletype (also no results for "rfc822" on Crates and I don't understand it well enough to not fuck that up)

 Sat May 09 01:52:05 +0000 2020

вгэт, or i guess you could transcribe it as a single-syllable vget?

 Sat May 09 11:51:39 +0000 2020

two on average but they ears gradually convert between humanoid and catoid(?) the narrower their pupils are

 Sun May 10 01:21:44 +0000 2020

combination kitchen/server room when

 Sun May 10 22:14:17 +0000 2020

cross-built rp-pppoe 3.14-BETA-2 for mipsel, coaxed an UBNT EdgeRouter X, the router for the day, to use it instead of the stretch one, shockingly works fine so far?

 Mon May 11 14:36:54 +0000 2020

the more non-trivial-"here's some text" mail i send to my dad who uses outlook, the more outlook feels like babby's first MUA

 Mon May 11 15:06:57 +0000 2020

more than one inline attachment? it's not inline and has a filename now, which is "Unnamed Attachment 00245.txt" for some inexplicable reason

 Mon May 11 15:09:52 +0000 2020

when the attachment is a message/rfc822 it's "message.msg"? at least it shows the icon with the badge to indicate it's signed

and obviously not inline, despite the disposition

 Mon May 11 15:13:15 +0000 2020

wondered if dding a farbfeld into a framebuffer device would work, and it did!

this does mean a hilarious 2:1 pixel ratio with a 1280x1024x32 buffer

 Mon May 11 18:49:22 +0000 2020

with the first channel skipped?

 Mon May 11 18:50:20 +0000 2020

neat

 Mon May 11 18:53:04 +0000 2020

full-width, so interlaced?

 Mon May 11 20:06:37 +0000 2020

sat (secret cat)

 Tue May 12 13:24:29 +0000 2020

called out

 Wed May 13 15:03:35 +0000 2020

took some fucking about with a linux chroot in the hurd chroot in a d-i env for grub to agree with this setup but

 Wed May 13 16:22:22 +0000 2020

ah, yes; hopefully tune2fs will work without resorting to dump(8) :v

 Wed May 13 16:26:24 +0000 2020

let's goooooooooooo

 Wed May 13 16:48:35 +0000 2020

or not, seeing as I can't get it to translate /proc and therefore pipes and therefore no other part of the system works. good times

 Wed May 13 17:47:03 +0000 2020

!!!!!!!!!!!!!

 Wed May 13 21:02:06 +0000 2020

:/;/:/

 Wed May 13 21:05:15 +0000 2020

each previous time some server crashed to a db> prompt; progress!

 Wed May 13 21:07:10 +0000 2020

libp11-kit0, which provides http://libp11-kit.so.0, shows up in package content search for hurd, but isn't actually in the archive :v

end of the road for me i think

 Wed May 13 21:27:05 +0000 2020

i'd complain that this could've gone much easier if there were documentation or a non-unhelpful bugtracker, but the former is cited as a feature in a (the?) FAQ, and, well, GNU project

 Wed May 13 21:31:50 +0000 2020

haven't written any new code in a good while now, this feels weird

 Thu May 14 19:27:25 +0000 2020

grghblg I fucking missed writing manpages

 Thu May 14 21:37:14 +0000 2020

caught one

 Thu May 14 22:53:56 +0000 2020

oh hey I've done that before!

 Fri May 15 18:55:55 +0000 2020

for context, I was converting the machine from MBR to BIOS-on-GPT from a d-i shell and broke it somewhere the first time

that was also the day I learned that wget dislikes when you --body-file a blockdev, and that dd if=/dev/sda of=/tmp/sda works around that; a fun time overall!

 Fri May 15 19:34:58 +0000 2020

if you ever wondered (who wouldn't?), the linux repository, if viewed as a single format-patch --root mbox, is five gigabytes in size and has 849133 patches

and neomutt fucking *smokes* through it; I'm so fucking impressed

 Fri May 15 21:02:02 +0000 2020

what if you wanted to investigate a bug report but god said

 Sat May 16 00:00:48 +0000 2020

said bug report concerns a "brew doctor" interacting with one of my binaries

the gargler gives me (1) beers (2) beer recipes (3) a forum and (4) this "FAQ" page, where it's mentioned exactly once in a way that makes clear the reference is known to the reader

 Sat May 16 00:08:07 +0000 2020

there's a link to a "Formula Cookbook" at the bottom of the FAQ, it has a table full of nouns which make no sense but contradict the FAQ

please i would just like to read some documentation, ideally the moral equivalent to a manpage, for this "brew doctor", whoever that is

pleas

 Sat May 16 00:16:35 +0000 2020

>Why does Homebrew prefer I install to /usr/local?
>Why does Homebrew say sudo is bad?
>Homebrew refuses to work using sudo.
>[To] run Homebrew in a multi-user environment, consider creating a separate user account especially for use of Homebrew.

i .. but.... wha ??how

 Sat May 16 00:28:13 +0000 2020

im SO confused WHY is a beer software so adamant to write to /usr, WHY does it want to allow everyone to write to /usr/local, WHY does it hate sudo in particular and not any other elevation package, HOW would it react to being suid, also didn't sudo do finer-grained capabilities?

 Sat May 16 01:42:20 +0000 2020

first take on a parser took 27 real/17 user/10 sys in debug and 12/4/8 in release (in tmpfs); acceptable, i guess

 Sat May 16 12:41:35 +0000 2020

oh hey, it's not every day you find a bug in a util-linux manpage (here fdisk.8 from 2.35.1)

 Sat May 16 16:18:31 +0000 2020

King Iso/IEC 14882:2017

 Sat May 16 23:47:37 +0000 2020

started seriously considering implementing GPT for Hurd since it just enumerated the protective MBR and that's how i know i need to sleep before i do something i regret

 Sun May 17 01:55:59 +0000 2020

woke up after four hours in a cold sweat, probably haunted by the ghost of rms, gonna have to at least look at it i think

 Sun May 17 09:43:35 +0000 2020

oh cool, didn't know apt could pull from two archives at once and display the progress so neatly

 Sun May 17 10:58:08 +0000 2020

turns out most of the references died in a rename in 2016? https://bugs.debian.org/960839

 Sun May 17 13:00:30 +0000 2020

ugh "licensing as public domain" i hate writing things

 Sun May 17 13:02:23 +0000 2020

patched parser library to avoid one allocation per header (https://github.com/staktrace/mailparse/issues/61), and gained 0.58/0.26/0.32 mean across four runs; neat

 Sun May 17 14:17:11 +0000 2020

debian maintainer (understandably) didn't want to upstream this, a friend graciously sent it instead: https://lore.kernel.org/util-linux/DBBPR08MB55480F8E7F62EFE9323E29FA9BBB0@DBBPR08MB5548.eurprd08.prod.outlook.com/T/#u

 Sun May 17 15:47:06 +0000 2020

the SPL has allowed us to use ZFS on a plethora of kernels but it doesn't remotely support the x32 ABI so it's impossible to say if it's good or not

 Sun May 17 18:46:31 +0000 2020

currently trying to build zfs for x32 to avoid implementing gpt for hurd to to avoid dealing with a sub-par mail API for feembox to avoid looking through k&r c for rp-pppoe aaAa

 Sun May 17 18:48:48 +0000 2020

hehe zfs go segvvvvv

 Sun May 17 19:15:05 +0000 2020

yo ngl this fuckin slaps

 Sun May 17 19:27:12 +0000 2020

holy shit i think im gonna have to implement a logical equivalent to dpkg-divert in cargo-update jesus fucking christ

 Sun May 17 20:21:59 +0000 2020

i386 doesn't really interest me, but zfsutils-linux does support it; x32 (https://wiki.debian.org/X32Port) was reported broken in 2012, and again in 2016 (https://github.com/openzfs/zfs/issues/844); a cursory glance at packages.d.o reveals that zfs-fuse is, unsurprisingly, similarly broken there

 Sun May 17 22:33:26 +0000 2020

has she gone and fucking done it again?

 Mon May 18 00:26:36 +0000 2020

the numbers DON"T LIE, lETS FUCKING GO

 Mon May 18 00:56:28 +0000 2020

still a faster scrub than ceph on hft hardware

 Mon May 18 01:12:28 +0000 2020

commit messages in the streets, patch names in the sheets

 Mon May 18 01:25:11 +0000 2020

oh i think i just got accidentally moderately high on a maft of nail varnish, it'd explain a lot

 Mon May 18 02:08:56 +0000 2020

accepted into stable/v2.35, should be in 2.35.2: https://git.kernel.org/pub/scm/utils/util-linux/util-linux.git/commit/?id=5fdb0866964f0911224ea37b6e42d7ff763138b8

 Mon May 18 12:10:06 +0000 2020

the chad POSSE vs the virgin HURD

 Mon May 18 18:26:53 +0000 2020

hm

 Mon May 18 18:32:34 +0000 2020

busybox 1.27.2 (2019-11-20 10:51:38 UTC), mipsel, an old stretch from a vendor, a perfect combination to segfault on "ls --color=" lol

 Mon May 18 18:39:14 +0000 2020

the busybox is also from the vendor, which explains a lot, tried with the current stretch busybox from debian and it works fine 🙄

 Mon May 18 18:50:15 +0000 2020

hm, the current version in stretch is 1.22.0, the vendor's 1.27.2, and the sid one (the static variant of which also works) is 1.30.1

according to strace, vendor gets to ioctl(0, TIOCGWINSZ, pointer), others continue though ioctl(1, TCGETS, {B38400 opost isig icanon echo ...}) ?

 Mon May 18 18:57:48 +0000 2020

can't get gdb on this machine (well, it's a router the size of my hand, so), and it looks fixed so no Cool Fun Bugs here :/

 Mon May 18 18:59:29 +0000 2020

llvm dragon purring

 Tue May 19 11:32:45 +0000 2020

passed out: 20%
woke up: 60%
i am in a clang hell, Core 2 Duos were not meant for this

 Tue May 19 11:34:26 +0000 2020

id use the clang from the archive but the build failed on the buildds for some reason so there's no llvm-toolchain-9 artifacts in for x32 and there was no other computer with the right kernel i could use last night so this poor "E6400 @ 2.13GHz" has gone at it for 12 hours so far

 Tue May 19 11:42:31 +0000 2020

fondly remembering when a friend had a clang patch i was optimising and the full build took like 11-15 minutes https://twitter.com/nabijaczleweli/status/1190808171700465664

 Tue May 19 11:45:52 +0000 2020

well, that's the difference between a core 2 duo+DDR2-666 on spinny vs two xeon E5645s+DDR3-1600 in tmpfs i guess

 Tue May 19 11:54:13 +0000 2020

/bin/ld: relocatable linking with relocations from format elf64-x86-64 (.../llvm-toolchain-9-9.0.1/build-llvm/lib/clang/9.0.1/lib/linux/libclang_rt.fuzzer_no_main-x86_64.a(FuzzerCrossOver.cpp.o)) to format elf32-x86-64 (fuzzer_no_main.o) is not supported

ah yes, normal day

 Tue May 19 14:17:41 +0000 2020

the good news is that this is the same error as on x32-do-01; the bad news is that for some inexplicable reason this build has managed to somehow create a 64-bit object?

 Tue May 19 14:34:10 +0000 2020

gave up on building the package after realising RUN_TESTS in d/rules is only ever assigned to :v

 Tue May 19 15:47:47 +0000 2020

im not even russian tho?

 Tue May 19 17:13:07 +0000 2020

building clean upstream got to linking and swiftly OOMed, but at least it got to linking; added another 7GB of swap and maybe it'll work?

 Tue May 19 18:57:31 +0000 2020

well duh, what other countries are in eurasia?

 Tue May 19 19:57:18 +0000 2020

7 megabytes of available memory, 1.8 of 8 gigabytes of swap free; it took on the order of two seconds to get a prompt back after cat exited lol

 Tue May 19 20:49:46 +0000 2020

not many (0?) sensible consumer boards that allow a two-socket/six-core/two-HT config with three DDR3-1600 16GB channels per NUMA node tbh

I got those numbers on an R720, my desktop is still a second-gen i7 with 16G of DDR3-1333 since I can't convince the board to clock it up

 Tue May 19 22:19:18 +0000 2020

plus, having the CPUs elsewhere means you don't get the signature PowerEdge takeoff you have to bitbang the BMC to fix, and other people can use the processing power as well, I'd recommend scaling that way over an uberdesktop tbh

 Tue May 19 22:21:24 +0000 2020

fucking 2+ hours of CPU time for both of these (vast majority spent in the kernel, waiting for pages, im pretty sure it's been more than two hours real-time) and it's just gonna "error: too big"? a cruel joke

 Tue May 19 23:29:03 +0000 2020

:(

 Wed May 20 01:15:39 +0000 2020

i built this again, in a different build tree; it failed with the same error; thing is, all the objects have the same 32-bit elf x86-64 (SYSV or GNU/Linux) format as all other objects in this tree; what

 Wed May 20 11:49:28 +0000 2020

e x p r e s s i v e

 Wed May 20 13:22:06 +0000 2020

running the link with half the archives first oomed (??) then failed with the same error on a different archive; shoot me

 Wed May 20 13:23:05 +0000 2020

in a flash of brilliance i decided to link this on a normal-size system and it's working better and better after copying the 10 gigs of objects and most of the system libraries :v

 Wed May 20 13:53:23 +0000 2020

linker script?? more like "mistake"

 Wed May 20 14:01:25 +0000 2020

knowing which symbols are in which so: passé
grep -R symbol /lib: the shit

 Wed May 20 14:06:06 +0000 2020

WHY is this out of thread https://twitter.com/nabijaczleweli/status/1263107564847407109

 Wed May 20 14:07:51 +0000 2020

these are the secret good symbols libc writers are hiding from us

 Wed May 20 14:14:12 +0000 2020

okay this does explain why it failed so badly on the poor old 4GB machine

 Wed May 20 14:19:56 +0000 2020

the gold linker didn't get past the secret supersymbols, the bfd variant fuCKINGWOKRED

 Wed May 20 14:39:32 +0000 2020

with each interpreter we get closer to running this piece of shit

 Wed May 20 14:49:12 +0000 2020

nice

 Wed May 20 15:37:03 +0000 2020

always segfaults in after/during this statement; Argv0 (here: Data) is usually an empty string, but it behaves identically if it's "./clang-10"; avoiding this return by setting Data=0 continues into another call into the same driver function, which calls this eventually :v

 Wed May 20 16:56:45 +0000 2020

made an x32 chroot on a different computer (8G RAM), linker gets to just under four allocated gigabytes and (expectedly) gives up; joy

 Wed May 20 17:07:29 +0000 2020

in an unprecedented stroke of not being a complete moron, i thought to install libstdc++6-dbgsym and have symbols for.. the allocator? really? oh no

 Wed May 20 17:11:28 +0000 2020

this is the most complete backtrace so far, and it ends with "Cannot access memory at address 0xfa0cd256" for some reason; ive run out of ideas

 Wed May 20 18:34:35 +0000 2020

across a few dozen runs, the RIP in the crash message always ends in 266, hm

 Wed May 20 18:39:40 +0000 2020

cat lost a wire under my bed and now she's frantically searching and trying to get under there to find it and i cant see it either and im fucking heartbroken dude

 Wed May 20 19:59:35 +0000 2020

buster,sid, amd64: nMigen Yosys 0.9.post4248.dev7 (git sha1 aee43936)

sid, x32,i386:

 Thu May 21 12:23:29 +0000 2020

linkER PLEASE what THE FUCK does "out of file descriptors and couldn't close any" fiuckING mEAN

 Thu May 21 13:55:48 +0000 2020

what

 Thu May 21 13:58:52 +0000 2020

96 gigabytes of ram btw

 Thu May 21 14:03:05 +0000 2020

man, fuck this; never test your patches on more than one setup; never attempt to make things work on non-mainstream stacks

 Thu May 21 14:08:07 +0000 2020

turns out openzfs uses a Fun autohell setup, I'm genuinely not sure where I'm supposed to put things lol

 Thu May 21 19:59:15 +0000 2020

PARTIAL SUCCESS; I REPEAT: PARTIAL SUCCESS

 Thu May 21 20:31:54 +0000 2020

hmm i should become a tcl dipshit of enormous proportion, that feels very on-brand for me; also, called out again

 Thu May 21 20:36:10 +0000 2020

posted to the debian-amd64@ list about this earlier today and up to now my mail has not shown up in the archive; hm

 Thu May 21 21:06:37 +0000 2020

It is!
Upstream release: https://lore.kernel.org/util-linux/20200520143307.m46d5u3vdmtrkhd6@ws.net.home/T, also in sid: https://tracker.debian.org/news/1146400/accepted-util-linux-2352-1-source-into-unstable/

or, as the friend whom i roped into this succinctly put it,

 Thu May 21 22:49:53 +0000 2020

something you know (how to use your muscles) and something you have (muscles): door handles are 2FA

 Thu May 21 23:07:04 +0000 2020

[obligatory rap segment in a pop song voice] bitches be hating on me / call that duplication of labour

 Thu May 21 23:14:02 +0000 2020

was barely in a state to report this (http://bugs.debian.org/961097), so forgot to post it here, but: htop on x32 segfaults when you ^Z and fg it back, but only on a VT and only with TERM=linux somehow??

 Fri May 22 02:07:49 +0000 2020

yass bitch https://lists.debian.org/debian-amd64/2020/05/msg00000.html

 Fri May 22 12:33:11 +0000 2020

very rare sub-1k issue in 10k+-issue repo being fixed at incredible hihg speed

 Fri May 22 16:24:55 +0000 2020

anyway, PR in https://github.com/openzfs/zfs/pull/10357, patches in https://github.com/nabijaczleweli/zfs-linux.deb-x32 (apply cleanly on 0.8.4, also ones with a (trivial) merge into current upstream master) – if you have an x32 machine it'd be great if you could try them!

 Fri May 22 17:05:06 +0000 2020

uhoh just found a gpm .so in the backtrace

 Fri May 22 21:53:55 +0000 2020

oh boy gpm code is a fucking adventure, dae #ifdef SIGTSTP /* true if BSD system */

 Fri May 22 23:08:33 +0000 2020

printf(3) debugging? no thanks; this is a write(2) debugging household

 Sat May 23 00:00:41 +0000 2020

ported the patch from https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=326709#10; it didn't work, of course, but it'd be hilarious if a patch from uhhh Mon, 5 Sep 2005 23:43:12 +0930 fixed an issue in a.d. today

 Sat May 23 00:09:49 +0000 2020

undefing SIGTSTP does fix it though so urghhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhh

 Sat May 23 00:15:05 +0000 2020

porting zfs was cool and fun because upstream is very active and debian is relatively alive as well, as long as this was broken htop this was fine since that's probably a type or alignment; this? i hate this. upstream is dead, last list post in Aug 2019, debian bugs are rotten

 Sat May 23 00:22:15 +0000 2020

(that patch I ported? from 2005? it's in the top "Outstanding bugs -- Important bugs; Patch Available (3 bugs)" section; the top bug in there is #178723, from 2003, with no posts beside the submission)

 Sat May 23 00:26:24 +0000 2020

((clearly nobody gives a god-forsaken /shit/ about this (and, well, can they really be blamed, in anno domini twenty fucking twenty?), does anyone beside me even use the damn thing? (popcon recent says 101 https://qa.debian.org/popcon.php?package=gpm; all Py2 packages below 100 got yote by default)))

 Sat May 23 00:36:25 +0000 2020

(((at the end of the day, why should I? hell – how? this is an undocumented mess of "hasn't broken yet" (there's comments where it has, usually from 199[89], usually about XEmacs) the current maintainer has taken over the code but, well, none of the intent, and he says as much)))

 Sat May 23 00:45:11 +0000 2020

im just so, so tired

 Sat May 23 00:46:15 +0000 2020

woke up, tested on an amd64 system, it crashed even harder, in that it didn't leave the crash message, immediately put to sleep again

 Sat May 23 10:32:03 +0000 2020

what a fucking trainwreck https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=961097#22

 Sat May 23 11:16:29 +0000 2020

how's your day going?

 Sat May 23 16:06:58 +0000 2020

well, fuck

 Sat May 23 17:15:10 +0000 2020

i was not aware this was possible how and why is gdb spinning so hard

 Sat May 23 17:17:20 +0000 2020

jumping over the procmask()/sigaction()/kill(SIGTSTP) stanza (https://sources.debian.org/src/gpm/1.20.7-6/src/lib/liblow.c/#L173-L175) seems to work (which is to say "not segfault")

 Sat May 23 19:07:26 +0000 2020

the x32 ABI is a trans icon

 Sat May 23 19:42:15 +0000 2020

...which works in the debugger but means that the program cannot be job controlled away in a normal shell :v

 Sat May 23 20:11:14 +0000 2020

my usual go-to for this is ":)" but i cannot make one that's sarcastic enough and my eyes hurt

 Sat May 23 21:53:48 +0000 2020

omgomgomgomg libgpm is getting unmapped i think

 Sat May 23 23:13:01 +0000 2020

note the address after the handler restoration (in this case it's ncurses' handle_SIGTSTP) and kill(2) call, that kill() will return to (0xf5e09f5b) later "Cannot be accessed", which is congruent with it being unmapped

 Sat May 23 23:34:04 +0000 2020

LD_PRELOADing something that overrides dlclose(3) does, in fact, fix this; h/t @Guriwesu for the suggestion lol

 Sat May 23 23:37:25 +0000 2020

interesting! I'll try tomorrow, building boost for two niche platforms on a core 2 duo sounds like an incredible time; although I haven't bootstrapped Rust (yet?), so unsure how comparatively relevant my results will be with whitequark's without wasmtime

 Sun May 24 00:42:01 +0000 2020

just wrote the *nastiest* fucking piece of the worst unreadable C89 you'd find in something like gpm (heh, fitting I guess)

 Sun May 24 15:59:38 +0000 2020

the options were writing the same thing in Rust, correct Rust with image-rs for direct output to normal formats, or, well.. this, but this works well enough to convert the 1280x1024xRGBA /dev/fb0 dumps i have to farbfeld, so meh

 Sun May 24 16:03:04 +0000 2020

yassssssssssss

 Sun May 24 16:29:49 +0000 2020

interesting, fbset says the buffer is rgba8, but it was bgra8; i liked the yellow colour scheme more tbh

 Sun May 24 16:52:12 +0000 2020

at what point does a bug report become a blog post and is it bad form

 Sun May 24 18:23:17 +0000 2020

turns out it's much larger than a blog post at 580L and 33794C, i was worried the screenshots were gonna be huge compared to the text, and, well, they're smaller lol

 Sun May 24 22:12:07 +0000 2020

https://bugs.debian.org/961097#35

 Sun May 24 22:24:25 +0000 2020

there's an obvious "I've peaked, nothing i ever write will top this" bit but it's probably shit and i've been staring at it for twelve hours so i can't see :v

 Sun May 24 22:31:40 +0000 2020

shockingly painless for a C++ thing that uses Boost tbh

 Mon May 25 00:43:25 +0000 2020

aaaaand we're back to reality with a gcc build that doesn't work and refuses to work and the documentation tells you to look at a config.log in this specific case which is to say it tells you to fuck yourself because there's at least a dozen of those in the build tree

 Mon May 25 02:22:31 +0000 2020

≈19% speedup over amd64 in the ecp5_evn example on a C2D E6400

building gcc for the arch soc_ecp5_evn requires defeated me (for now), no cpus from this millennium i can test on (yet)

 Mon May 25 04:12:19 +0000 2020

configure: error: cannot compute suffix of object files: cannot compile
configure: error: cannot compute suffix of object files: cannot compile
configure: error: cannot compute suffix of object files: cannot compile
configure: error: cannot compute suffix of object files: cannot

 Mon May 25 17:11:18 +0000 2020

dickhead

 Mon May 25 17:33:17 +0000 2020

soc_ecp5_evn: ≈22% on E6400, ≈29% on i5-3450, and ≈28% on i7-8565U

assuming everything else being equal, that'd mean an equivalent user time of 11.706/11.115/11.186 in whitequark's screenshot, respectively

 Mon May 25 20:29:56 +0000 2020

If you're interested in measuring x32 speedups in compute-intensive workloads from https://twitter.com/nabijaczleweli/status/1265017275658182657 for yourself, you can try the tarball (and instructions) from https://lfs.nabijaczleweli.xyz/0002-nextpnr-x32-comparison/; if you do, please report back!

 Tue May 26 02:29:48 +0000 2020

a friend had a go on one of his HFT machines with a 5 GHz i9-9980XE, HT off, and got ≈29% without a scheduler and ≈23% with, so that's probably a vague limit for this workload

surprised it goes even remotely near a 20-30% gain, was expecting like ten; thanks for tagging me in!

 Tue May 26 15:25:02 +0000 2020

bitches really out here being transphobic like okay bitch i hate me too, you ain't special??

 Wed May 27 02:09:52 +0000 2020

remembered that ELFs have a shebang
wild shit

 Wed May 27 04:16:32 +0000 2020

👁️

 Wed May 27 13:36:29 +0000 2020

 Wed May 27 13:37:46 +0000 2020

RΠ-ΠΠΠoE 3.14 оут, баыбеее https://dianne.skoll.ca/pipermail/rp-pppoe/2020q2/000587.html

 Wed May 27 14:14:42 +0000 2020

built mach, booted my mach, told by mach to piss off

 Wed May 27 15:37:48 +0000 2020

well, okay, it had a point, i didn't include any drivers; a build with drivers on does actually work like the debian one (well, "work")

 Wed May 27 15:46:07 +0000 2020

omogmogm gogmgo

 Wed May 27 18:37:47 +0000 2020

hehe printk go scrolllllllllll

 Wed May 27 20:39:40 +0000 2020

it'd be great if the efi spec actually said this, but it doesn't, afaict, so that's cool

 Wed May 27 21:34:21 +0000 2020

turns out my clown ass got the EFI 1.10 spec instead of the UEFI spec, but 2.8A says this which is still pretty bad

 Wed May 27 22:55:55 +0000 2020

https://github.com/nabijaczleweli/gnumach/commit/8169a06dc8547a4509a0ec7f15ef3c6ad8887582?diff=unified

 Thu May 28 01:31:43 +0000 2020

that was supposed to have a witty tagline but it doesn't because (a) brain bad and (b) twitter bad; my apologies

 Thu May 28 01:49:26 +0000 2020

oh, thanks!

 Thu May 28 03:05:09 +0000 2020

i need a shower

 Thu May 28 03:06:14 +0000 2020

hmm yeah i was wondering if it's as bad as it looks or if i was missing something; ace pun btw

 Thu May 28 04:41:29 +0000 2020

name a better morning, i'll wait

 Thu May 28 13:56:44 +0000 2020

yass bitch https://github.com/openzfs/zfs/issues/844#event-3383893277

 Thu May 28 17:43:47 +0000 2020

pointer incidents: 1
pointer crimes.. well let's not count

 Thu May 28 18:34:44 +0000 2020

why does mach think my HDD is 44136353 blocks shorter than it actually is?? what on earth

 Thu May 28 20:20:11 +0000 2020

nice

 Thu May 28 21:48:19 +0000 2020

https://github.com/nabijaczleweli/gnumach/commit/72983ec49ac6afbecb805476581fca595e11c436

 Thu May 28 22:08:55 +0000 2020

hurd even mounts its rootfs!

 Thu May 28 22:10:00 +0000 2020

hehe C go *********************

 Thu May 28 22:18:14 +0000 2020

what if you wanted to type but a cat said

 Fri May 29 15:15:19 +0000 2020

moderate multi-drive success

 Fri May 29 15:48:59 +0000 2020

turns out you can just give qemu mach as -kernel and it boots well enough, so that's nice

 Fri May 29 15:53:00 +0000 2020

that diffstat? love to see it

 Fri May 29 16:02:15 +0000 2020

ace

 Fri May 29 17:51:56 +0000 2020

wrapped up the README and now im sad for some inexplicable reason :(

 Fri May 29 19:54:23 +0000 2020

It even showed up in the archive on the first try; outstanding https://lists.debian.org/debian-hurd/2020/05/msg00014.html

 Sat May 30 12:24:09 +0000 2020

looking at the FSF assignment bullshit and regretting writing this patch

 Sat May 30 14:25:13 +0000 2020

"send us /all/ your personal details and we may consider allowing you to fix our software" what the fuck is this
and for all this stupid fucking red tape there's no GDPR notice anywhere, which you'd think would be relevant when they require you to send identifying information lol

 Sat May 30 14:32:52 +0000 2020

but, as we all know, nothing exists past the oceans, north of the canadian border, or south of the mason-dixon line, and everyone should want their "full legal name", "postal address", birth year, and country of citizenship plastered everywhere; what kinda goddamn scam is this?

 Sat May 30 14:39:32 +0000 2020

"your full legal name (in ASCII characters) as the subject"

................ this has to be a joke, right?

this margin is too narrow to describe how most of these alienate a large chunk of populace, but FUCKING "in ASCII characters" ARE YOU JOKING

 Sat May 30 14:46:32 +0000 2020

the funny thing is that gives me an easy out since, even if I wanted to use my "full legal name" (which, lmao), i cannot, so that's pretty ace

 Sat May 30 14:50:43 +0000 2020

it's true!

 Sat May 30 14:55:56 +0000 2020

looks like there's a reason the components of The GNU System™ that aren't used in linux are in such a sorry fucking state lol

you can just write "we don't accept contributions" instead of funding a legal fuck you to anyone that wants to contribute, it's fine; really

 Sat May 30 15:14:10 +0000 2020

ngl this whole thing got me rethinking ever using any licence that isn't 0BSD for anything i ever write again

 Sat May 30 16:53:53 +0000 2020

also all GNU projects, but, well, that was already decided the first time I looked on the bugtracker for the hurd, what with its classic GNU hospitability

 Sat May 30 16:56:23 +0000 2020

pistachio is cute

 Sat May 30 18:34:25 +0000 2020

hmm

 Sat May 30 21:06:05 +0000 2020