ChanServ changed the topic of #kisslinux to: Unnofficial KISS Linux community channel | https://kisscommunity.org | post logs or else | "An idiot admires complexity, a genius admires simplicity." -- Terry A. Davis
claudia02 has joined #kisslinux
<claudia02> there is this ubikbsd link which is I think obsd related.
<claudia02> It was accessible without account. maybe its gone now. dunno
<illiliti> i don't understand why they can't just use poll/select
<ryoshu> in wayland?
<ryoshu> upstream
<illiliti> yep
claudia02 has quit [Quit: zzz]
illiliti has quit [Quit: leaving]
progenyx has quit [Quit: progenyx]
<tleydxdy[m]> hmm, when I'm trying to make the kernel the build script seems to want diff -I
<tleydxdy[m]> a few other errors, like for find
kiedtl|litterbox has joined #kisslinux
mov_eax has joined #kisslinux
mov_eax has quit [Quit: WeeChat 3.0.1]
Uks2 has quit [Quit: Byee]
Uks2 has joined #kisslinux
<testuser[m]> Hi
<testuser[m]> tleydxdy: it doesn't matter
<tleydxdy[m]> hmm, it only built a tiny part of the kernel
<tleydxdy[m]> guess my config is wrong somewhere
<tleydxdy[m]> perhaps I should use my current one from arch as the base
<tleydxdy[m]> noticed the use of underscore
<tleydxdy[m]> didn't know that was a thing in the shell
GalaxyNova has joined #kisslinux
Erus_Iluvatar has quit [Quit: Erus_Iluvatar]
Erus_Iluvatar has joined #kisslinux
GalaxyNova has quit [Quit: zzz]
claudia02 has joined #kisslinux
<claudia02> illiliti: heres some wip work from 3-4 ago https://github.com/jasperla/openbsd-wip/tree/master/x11/wayland
claudia02 has quit [Client Quit]
<ryoshu> 3-4 years ago
progenyx has joined #kisslinux
mahmutov has quit [Ping timeout: 252 seconds]
gtms has quit [Remote host closed the connection]
mahmutov has joined #kisslinux
illiliti has joined #kisslinux
Torro has joined #kisslinux
Uks2 has quit [Ping timeout: 265 seconds]
Uks2 has joined #kisslinux
progenyx has quit [Quit: progenyx]
Uks2 has quit [Ping timeout: 252 seconds]
Uks2 has joined #kisslinux
mahmutov has quit [Ping timeout: 260 seconds]
mahmutov has joined #kisslinux
elenapan has joined #kisslinux
<tleydxdy[m]> ree, compiling the kernel still gives __always_inline error, but I've already patched swab.h
<testuser[m]> just find the file with the error and add the same linux/stddef.h include
<tleydxdy[m]> yeah, it's still complainingabout swab
elenapan has quit [Quit: WeeChat 3.2.1]
Torro has quit [Quit: bye]
GalaxyNova has joined #kisslinux
<tleydxdy[m]> ah, wth
<tleydxdy[m]> it's using the headers from the system
<tleydxdy[m]> rather than the unpacked directory
<GalaxyNova> wierd
<tleydxdy[m]> lol samba.org is down
<tleydxdy[m]> is there someway to do split package?
<tleydxdy[m]> or are the source file downloaded automatically deduped
mahmutov has quit [Read error: Connection reset by peer]
mahmutov has joined #kisslinux
GalaxyNova has quit [Quit: zzz]
<acheam> tleydxdy[m]: sources are not deduped between pacakges
<tleydxdy[m]> hmm
<tleydxdy[m]> would make sense to dedup using the checksum for example
<tleydxdy[m]> ah, I guess that would be a problem if say the checksum was outdated
<tleydxdy[m]> still fighting the initramfs, fun
<acheam> i guess it could, but that might be slow
<acheam> would require storing the checksum of every downloaded file and grepping for it each time you download a package
<acheam> for a not too common use case
<rio6> checksums are already stored in package db
<rio6> oh hm nvm
<rio6> the files need them too
<tleydxdy[m]> well, it would just be a symlink
<tleydxdy[m]> so the files are stored in sources with their name as the checksum
<tleydxdy[m]> and the current way source work are converted to symlinks
<tleydxdy[m]> but yeah, now I think about it it would has some problem
<tleydxdy[m]> better to have split package or provides
<tleydxdy[m]> s/name/checksum/, s/checksum/file name/
<cotangent> <tleydxdy[m]> better to have split package or provides
<rio6> I do want a "provides" for kiss, I think the current way of doing it is to symblink alternative package to the package it provides, but it's not the greatest
ctb_ has joined #kisslinux
ctb_ has quit [Client Quit]
Guest3114 has joined #kisslinux
testing_catgirl has joined #kisslinux
testing_catgirl has quit [Client Quit]
testing_catgirl has joined #kisslinux
testing_catgirl has quit [Client Quit]
Guest3114 has quit [Quit: Client closed]
mahmutov has quit [Ping timeout: 260 seconds]
Guest4475 has joined #kisslinux
<Guest4475> Are there any RSS feed readers that have been packaged for kiss already?
GalaxyNova has joined #kisslinux
<tleydxdy[m]> rio6: I tried the symlink method, but it doesn't quite work right (or I'm misunderstanding something)
<tleydxdy[m]> basically it tries to install the package twice
<tleydxdy[m]> since some package depends on the alternative directly
GalaxyNova has quit [Quit: Client closed]
<rio6> yeah I don't like that method either
<illiliti> does openssl still depend on perl?
<tleydxdy[m]> I think the fix would be relatively simple
<tleydxdy[m]> just dereference the symlinks when reading the dependency
<tleydxdy[m]> rather than using them as is
Guest4475 has quit [Quit: Client closed]
<tleydxdy[m]> so I'm reading through the sources of kiss
<tleydxdy[m]> seems like circular dependencies are not handled
<tleydxdy[m]> what's the intended way to do it?
<illiliti> they are handled
<illiliti> it's a very hacky and ugly way but it works
<tleydxdy[m]> well, die is not really handling it
<tleydxdy[m]> I guess the right solution would be to ditch the software since it's badly designed
<tleydxdy[m]> * the software (that has circular dependency) since it's
<illiliti> die prevents segfault due to stack overflow caused by circular dep
<tleydxdy[m]> okay, by handling it I mean some way to make the package install correctly
<illiliti> circular dep is bug not a feature
<illiliti> that's why kiss just dies if it sees circular dep
<illiliti> this is the only *proper* way to handle circular deps imo
<tleydxdy[m]> > I guess the right solution would be to ditch the software (that has circular dependency) since it's badly designed
<tleydxdy[m]> so this
<illiliti> yeah
<illiliti> see also: freetype-harfbuzz
<tleydxdy[m]> ah ic
<tleydxdy[m]> so zfs probably need the same thing
<illiliti> what's the problem with zfs?
<rio6> all these packages have readme now, nice
<ryoshu> for what events we get device change in udev?
<ryoshu> network link up/down?
<illiliti> ryoshu: display brightness change
<ryoshu> hmm, is there a comletr
<ryoshu> complrtr
<ryoshu> complete list somewhere
<illiliti> idk
<ryoshu> (sorry screen is wet and misclicks)
<tleydxdy[m]> zfs packages has circular depencies, but I guess it would make sense to jusr package as one
<illiliti> ryoshu: all good
<illiliti> grep change action in linux source
<illiliti> https://grep.app/search?q=KOBJ_CHANGE&filter[repo][0]=torvalds/linux
<ryoshu> thanks
<ryoshu> i am working on enumeration, is there a need to store match patterns in udev_list_entry? i used a dedicated data structure
<ryoshu> list, but i plan to switch to rbtree
<illiliti> API hardcodes udev_list_entry
<ryoshu> in future, for now the goal is to get it operational
<illiliti> you must comply
<ryoshu> in what call hardcodes?
<ryoshu> i plan to return to the code in 0.5h
<illiliti> udev_enumerate_get_list_entry
<ryoshu> On success, udev_enumerate_get_list_entry() returns a pointer to the first entry in the list of found devices.
<ryoshu> but i am thinking about match_subsystem etc
dilyn has joined #kisslinux
<ryoshu> match_parent
<ryoshu> match_property
<illiliti> the whole match thing is the mess
<dilyn> tleydxdy[m]: wdym zfs has a circular dep?
<illiliti> horrible mess
<illiliti> i think callback is better than these hardcoded match functions
<illiliti> anyway, you can copy the logic from libudev-zero code
<illiliti> ryoshu: also, don't bother with sysattrs
<illiliti> same for match_parent
<ryoshu> it uses udev_list_entry and stores value as key, keeping right hand value null
<ryoshu> kind of surprising design if this is mandated by api
<illiliti> yeah, the api is shit
<illiliti> libudev-devd stores devnodes instead of syspaths
<illiliti> i think you should do this too
<ryoshu> i only keep /dev/nodeX
<ryoshu> major minor unit (same as unit?)
<ryoshu> i need to submit from the kernel subsystem and input discrimination data, like touchpad or mice
<ryoshu> /sys is trimmeed fully
<illiliti> good
<ryoshu> also parent, but the kernel version is i think buggy
<ryoshu> but as no api tests it, we get what we have
<ryoshu> is refcnt design necessary for the api? are there some wins with it?
<ryoshu> except being fancy
<illiliti> it's useful if you pass context between many libraries. so you don't need to manually track when you need free context
<illiliti> it's also useful keeping parent device of udev_device
<ryoshu> but the library is generally not thread safe, am i right?
<ryoshu> by design
<illiliti> yes
<ryoshu> is this a problem ever?
<ryoshu> in practice
<illiliti> no
<ryoshu> okay, thank you
<illiliti> np
<illiliti> refcnt is basically a primitive implementation of garbage colletion
<ryoshu> i know
<ryoshu> as of today my dream api would be to create udev context, create a tree for all devices
<ryoshu> export api to traverse the tree
<ryoshu> then add api to monitor updates in the tree
<ryoshu> feedbacking what device changed how
<ryoshu> so no creation of udev_device on demand
<tleydxdy[m]> <dilyn> "tleydxdy: wdym zfs has a..." <- just they way it's usually packaged elsewhere (e.g. arch)
<ryoshu> which i find fragile
<dilyn> ew
<ryoshu> does it make sense this api idea?
<dilyn> zfs only has like five dependencies :v
<dilyn> zlib, linux-headers, libblkid, libuuid, libtirpc
<ryoshu> the tree would be read only and a single ref on udev
<ryoshu> perhaps ref on enumetation