09:23 <sortie> today in osdev: my bison port has stopped working for unknown reasons, just messed a bit around with sysconf, pathconf, unistd.h, limits.h

09:24 <sortie> /bin/bison: m4 subprocess failed: Bad file descriptor

09:29 <GeDaMo> Does m4 work on its own?

09:31 <sortie> It used to :)

09:31 <sortie> It appears that m4 is not even managing to get executed

09:31 <GeDaMo> Could it be a file permissions thing?

09:31 <sortie> I already listed the things that changed above :)

09:32 <sortie> I'm going to guess towards gnulib bullshit and some interpretation of sysconf limits or something that blows up in my face

09:41 <sortie> god it's happening inside an insanely complex if condition that's longer than a page in my editor

09:46 <sortie> yep.. it broke because I changed sysconf(_SC_OPEN_MAX) from 0x10000 to -1 (unlimited)

09:47 <sortie> https://cgit.git.savannah.gnu.org/cgit/gnulib.git/tree/lib/spawn_faction_adddup2.c#n40 ← dammit this is even a bug in the latest gnulib

09:55 <Ermine> TIL CONFIG_RELR

10:07 <sortie> god.. it's not just gnulib.. it's _so much_ that's broken with no file descriptor limits. SO MUCH software iterating from 3 to sysconf(_SC_OPEN_MAX) to e.g. close all file descriptors, since closefrom(2) didn't exist portably. Other stuff too. It might be feasible to fix but it sure it is a lot of work and who knows how upstream will react

10:08 <Ermine> RELR is not news though

10:21 <heat> > no file descriptor limits

10:21 <heat> found the problem

10:59 <sortie> heat: I mean, I'm just trying to have sysconf and friends describe the current situation, without changing it yet

10:59 <sortie> The whole limits thing is a longer discussion, you saw the meta issue :)

11:17 <zid> I mean, shouldn't the limit be INT_MAX regardless

11:17 <zid> given int fd

11:29 <heat> -1 means "limit is indeterminate"

11:29 <ZetItUp> sqrt(-1);

11:29 <heat> which no one handles because there's a valid reason for fd limits that exist _everywhere_

11:31 <zid> I mean, looping 0 to 2 billion sounds bad too

11:31 <zid> so it's fucked all round

16:58 <therealsdf> wsg

16:58 <therealsdf> i'm new to this lol how do y'all be making operating systems?

17:00 <zid> With a compiler, generally

17:01 <EmanueleDavalli> Sometimes an assembler too

17:05 <kof673> like that

17:07 <therealsdf> yea lol imma read up on the wiki fr

17:15 <kof673> https://wiki.osdev.org/Richard_Theseus that has other meanings, but ...one approach is just replace every piece until you have your own dictatorship set up

17:15 <bslsk05> ​wiki.osdev.org: Just a moment...

17:15 <kof673> generically, that doesn't require cloning per se, just gathering all the pieces

17:19 <kof673> so just need to write a manifesto to plan your coup /s

17:42 <heat> imma read up on the wiki frfr on god bro

17:43 <heat> chat i cant believe bro wrote a slab allocator 💀

17:47 <CompanionCube> huh?

17:47 <kof673> i think it was a wiki-directed comment "oh god"

17:49 <kof673> as in: "i'll just read the wiki" .... "oh, god"

18:45 <heat> i already have it

18:46 <Ermine> oh

18:46 <Ermine> late to the party

18:47 <heat> yeah

18:47 <heat> well the kernel doesn't have it, but musl and glibc do

18:47 <heat> but i don't relocate the kernel anyway

20:28 <therealsdf> hate how hp sets a password on my bios on my computer that my (dad) paid his money for

20:36 <Ermine> the what

20:37 <heat> the linuxkernel

21:16 <SarahMalik> they left

21:19 <heat> openrc really is unfun

21:19 <heat> sadly systemd is probably far off

21:28 <geist> really? thought openrc would be fairly straightforward

21:28 <geist> or maybe building it up from scratch is unfun?

21:33 <heat> yeah i mean, it's just not fancy at all

21:34 <geist> does it make you feel bad for using something simple?

21:34 <heat> and the init scripts are just super unreadable (IMO)

21:34 <geist> not enough python or rust or C+=?

21:34 <heat> and that's the big problem

21:34 <geist> yah okay, that would check out

21:34 <heat> they try to do more shit than what my system needs, requires all sorts of UNIX tooling (particularly ifconfig, route)

21:34 <heat> all sorts of "IF UNAME IS LINUX DO THIS, ELSE THAT"

21:35 <SarahMalik> openrc is not fancy and it doesn't have to be

21:49 <geist> ah i see, if it has autoconfig style compatibility that might be annoying as heck

22:14 <Ermine> do you look at gentoo scripts

22:14 <Ermine> because i don't remember seeing this shit in alpine scripts

22:15 <heat> i looked at whatever is shipping with upstream openrc

22:18 <Ermine> i believe openrc upstream stuff is tailored to gentoo

22:18 <Ermine> so alpine basically ships its own scripts

22:21 <heat> it's all delightfully quaint at the very least

22:21 <heat> manually add ttys and stuff

22:23 <Ermine> actually, are they trying to support bsds?

22:25 <heat> probably

22:25 <heat> hence the if uname = Linux stuff

22:26 <geist> yah maybe the gentoo stuff has had that all stripped out

22:26 <geist> heat: have you asked A1 for help?

22:26 <heat> who

22:26 <geist> hey A1, make me some openrc scripts for my os kthx plox

22:28 <geist> you need to add more A1 to your diet

22:30 <Ermine> geez

22:31 <geist> (the joke doesn't really work on non americans, but A1 is a well known steak sauce and one of the current govt administrators gave this long speech about how we need to add more A1 to our schools, instead of AI)

22:31 <geist> A1 the company went all in for a minute and posted adds on twitter and whatnot like 'yeah make sure your kids get more A1!'

22:32 <geist> the speech was hilarious and sad because she clearly had no fucking clue what she was talking about and hadn't even proof read the script

22:33 <geist> anyway i always locally name everything A1 and whatnot, just in case i can get the nomenclature to stick

22:33 <geist> it's my little piece of being a hard core rebel

22:34 <geist> actually writing a chrome extension that replaces 'AI' with 'A1 Steak Sauce' would be kinda fun

22:34 <geist> i used to run one that replaced 'Millenials' with 'snake people'

22:34 <geist> and all the usual variants of it

22:37 <geist> there, found an extension that does it

23:06 <kof673> this A1 fellow seems like a type of rubber duck

23:19 <chiselfuse> https://gabi.xinuos.com/elf/07-pheader.html#:~:text=during%20dynamic%20linking.-,An%20executable,-or%20shared%20object

23:19 <bslsk05> ​gabi.xinuos.com: 7. Program Loading — ELF Object File Format 4.3 DRAFT documentation

23:19 <chiselfuse> Why does this this calculation include (twice) truncation to the nearest multiple of the maximum page size instead of immediately just getting the difference between the memory address and the p_vaddr? I read somewhere that the memory address and p_vaddr ought to always be congruent module the page size which would mean this truncation step doesn't affect the result.