ChanServ changed the topic of #kisslinux to: Unnofficial KISS Linux community channel | https://kisscommunity.bvnf.space | post logs or else | song of the day https://vid.puffyan.us/H7PvgY65OxA
<illiliti> but it can't be that simple right?
<illiliti> what i'm missing?
rohan has joined #kisslinux
ioraff has joined #kisslinux
rohan has quit [Ping timeout: 268 seconds]
<ioraff> illiliti: how would the package manager know what provides /usr/lib/libLLVM.so?
<illiliti> grep /usr/lib/libLLVM.so /var/db/kiss/installed/*/manifest
<illiliti> basically kiss-owns
<ioraff> but if a package that provides it isn't installed, the pm won't know what dependency to resolve
<illiliti> then it's up to user which package to install to satisfy dependency
phinxy has quit [Quit: WeeChat 3.5-dev]
phinxy has joined #kisslinux
<illiliti> we could also include manifests into repo to automate that
<illiliti> or leave a comment to suggest which dependency to install
<illiliti> /usr/lib/libLLVM.so make # llvm, llvm-fat
<midfavila> if the manifests were included, that could work
<midfavila> otherwise I fail to see how it would be an improvement over the current system
<ioraff> manifests won't be constant until the package manager sandboxes builds
<ioraff> even so, a manifest could be different under a different version of a package in the build environment
<illiliti> fair, let's proceed to another option
<illiliti> what about symlinks? has anyone tried to implement provides using them?
<ioraff> what do you mean?
<illiliti> pkg -> llvm(symlink) -> llvm-fat
<illiliti> pkg depends file: llvm make
<illiliti> after build is complete, we could change llvm to llvm-fat in depends file
chomwitt has quit [Ping timeout: 268 seconds]
<illiliti> so /var/db/kiss/installed/pkg/depends will have: llvm-fat make
<illiliti> or just keep: llvm make
<illiliti> i didn't measure which one is better
<illiliti> we would need a new tool to control symlinks though
<illiliti> actually provides is possible already via abusing KISS_PATH and repos
<illiliti> but it isn't great
<ioraff> I feel a provides file is still the best option. seems the most flexible and least opaque
<illiliti> yeah
<illiliti> i'm just looking at other options to make sure we didn't miss anything
fitrh has joined #kisslinux
<illiliti> how kiss supposed to choose "provided" dependency if there are many of them?
<illiliti> if prompt - i don't like it
<illiliti> priority? i suspect that would make logic more complicated
<ioraff> I had only imagined a prompt. priority wouldn't let the pm be agnostic about package names, and has the possibility of packages claiming the same priority.
<illiliti> prompt sucks
<illiliti> i'd rather bail out and let user decide which dependency to install
<ioraff> that's functionally equivalent to a prompt. x package needs y -> pm lists all the packages in KISS_PATH that provide y -> prompt for which to install or exit and have the user manually install one.
<illiliti> yeah but without input parsing nonsense
<illiliti> if we could make it a simple Y/N, then ok
<illiliti> we also have KISS_PROMPT. we should respect it somehow
<ioraff> it'd just be parsing a number. I wonder how every other distro that has a similar function determines what package is the default.
<illiliti> priority, as far as i'm aware
<illiliti> no other way
<illiliti> if multiple packages have same priority then abort()
<illiliti> any other options?
ella-0 has joined #kisslinux
ella-0_ has quit [Read error: Connection reset by peer]
fitrh has quit [Remote host closed the connection]
fitrh has joined #kisslinux
<testuser[m]1> Hi
<virutalmachineus> hi
<ioraff> hi
fitrh has quit [Remote host closed the connection]
<testuser[m]1> illiliti: how is priority decided
rohan has joined #kisslinux
rohan has quit [Ping timeout: 244 seconds]
rohan has joined #kisslinux
rohan has quit [Ping timeout: 260 seconds]
rohan has joined #kisslinux
<testuser[m]1> Also we should prompt to rebuild every package that depends on the provider that has been changed
<testuser[m]1> wait no
<testuser[m]1> yea
<testuser[m]1> no
<ioraff> no
<virutalmachineus> but why not!!!
<virutalmachineus> my CPU is idle everyday, it needs a workout once in a while.
<virutalmachineus> My CPU needs to be in tip top shape when I need to compile chromium.
<virutalmachineus> getting the CPU warmup can increase performance by 0.00843%
<wael[m]> is there a tool that can automatically fill in the commit message for a kiss package update/new ?
<wael[m]> eg kiss commit <package> and it automatically adds the package and does the commit message on if its a new or a update
<testuser[m]1> cm() { git add $1; git commit -m "$1: $(awk '{print $1}' $1/version)"; }
<wael[m]> fair enough
<testuser[m]1> what should be done about unrelated packages tainting other packages due to auto enabled features?
<testuser[m]1> eg freetype-harfbuzz just got built with brotli support cuz i have android fools installed
<ioraff> sandboxed builds
<testuser[m]1> that came to my mind too but it's unportable
<testuser[m]1> it could be portable with enough ifdefs
<testuser[m]1> also SUID vs user namespaces vs landlock vs ...
<testuser[m]1> enumerating all possible build options like genpoo doesnt seem feasible / reliable
<ioraff> fall back to unprivileged builds. and probably landlock, but it'd require a relatively new kernel
<testuser[m]1> it can have all backends ig
<ioraff> I guess the number of people using KISS on another kernel is about 0.
<wael[m]> tbf having sandboxed builds can introduce the possibility of having make dependencies not being installed on the host system
<wael[m]> that as well can introduce the possibility of having *-libs *-devel packages
<ioraff> good, then the depends files will actually be accurate
<ioraff> don't see how it'd require split packages
<wael[m]> it can save some space on the host system by having development/library packages at build time
<wael[m]> an example of this is xbps-src, unfortunately its bash only and probably isnt portable
<ioraff> both are still required at build time
<testuser[m]1> split packages dont make sense on any source based distro
<wael[m]> true, gentoo doesnt split packages
<wael[m]> probably does but not for splitting devel libs and stuff
<ioraff> the only advantage you'd have is not needing headers to be installed outside of build time
ioraff has quit [Quit: ioraff]
chomwitt has joined #kisslinux
sam_sepi0l has joined #kisslinux
fitrh has joined #kisslinux
fitrh has quit [Remote host closed the connection]
fitrh has joined #kisslinux
<illiliti> we can reserve second field for priority
<illiliti> i.e: <name> <uint>
<illiliti> package with higher number has higher priority
<illiliti> based on that we reorder "provided" dependencies
<illiliti> if user has set KISS_PROMPT, package with highest priority will be default
midfavila has quit [Quit: Leaving.]
<illiliti> KISS_PROMPT=0 to be exact
midfavila has joined #kisslinux
<testuser[m]1> Where is the priority field going to be set and how would it make sense for independent repos
<illiliti> in provides file
<illiliti> what is independent repo
<testuser[m]1> Repo doesn't have knowledge of other repo
<testuser[m]1> llvm-fat provides file would have `llvm 2`
<testuser[m]1> llvm-superfat would have .
<midfavila> >mfw github requires JS to access release tarballs now
<illiliti> soydevs don't have anything better to do than reworking ui it seems
<midfavila> fortunately you can access releases in plaintext using https://github.com/$AUTHOR/releases/expanded_assets/$RELTAG
<midfavila> well, not plaintext, but minimally formatted HTML
<illiliti> llvm-superfat would have: llvm 1 ?
<illiliti> llvm-sane would have: llvm 3
<illiliti> what problem are you implying?
<testuser[m]1> I mean if u have llvm in main repo, llvm-fat in some zig repo, llvm-superfat in some $LANG repo and each repo is maintained by different authors
<testuser[m]1> how would they set propriety
<testuser[m]1> priority
<testuser[m]1> Also what does a priority greater than 2 mean
<testuser[m]1> It's either sane package or modified package
midfavila has quit [Remote host closed the connection]
midfavila has joined #kisslinux
<illiliti> well, some coordination would be required
<illiliti> and IMHO must not have same priority
<illiliti> packages*
<illiliti> values of priority have no special meaning
<illiliti> so if modified package has 4, i.e highest priority, then it will be default1
<illiliti> same if package has 400 priority
<illiliti> it doesn't matter
<illiliti> we could reserve 0 for something special, but i don't know for what
<testuser[m]1> What about just making automatic decision based on repo priority
<testuser[m]1> Or better just error out if kiss prompt is set and print msg to `kiss select llvm llvm-fat` or something
<illiliti> yeah, that's my first thought
<testuser[m]1> erroring out is appropriate according to kiss ideas stuff should be explicit
<illiliti> > What about just making automatic decision based on repo priority
<illiliti> that's how people implement provides right now
<illiliti> KISS_PATH="./modified:./orig"
<illiliti> kiss b pkg will take llvm from modified
<illiliti> reorder KISS_PATH and you have provides system
<illiliti> but as i said it isn't that great
<illiliti> there are some limitations
<testuser[m]1> What do u think about the erroring out idea
<testuser[m]1> For me erroring > decision on repo priority > priority system
<illiliti> it's fine
Beni has joined #kisslinux
midfavila has quit [Remote host closed the connection]
midfavila has joined #kisslinux
fitrh has quit [Remote host closed the connection]
an3223_ has quit [Remote host closed the connection]
an3223_ has joined #kisslinux
<testuser[m]1> Does it look ok
<testuser[m]1> I'll make new release with rpath fix and this fix before major changes like b3 or destdir stuff
an3223_ has quit [Ping timeout: 258 seconds]
<illiliti> looks good
an3223_ has joined #kisslinux
<illiliti> should we unbundle zip/unzip from firefox as well?
<testuser[m]1> I added it to issue already
<testuser[m]1> i think we should
<testuser[m]1> Btw yasm doesn't seem to be needed at all
<testuser[m]1> illiliti: what about gcc mpfr mpc gmp
<illiliti> hard to say
<illiliti> probably should be unbundled too
<illiliti> libwebp, ttf-croscore must use https
<testuser[m]1> illiliti: are u in college
<illiliti> no
<testuser[m]1> job?
<illiliti> yep
<testuser[m]1> Oh
<illiliti> i wouldn't call it job tho
<illiliti> typical underpaid shithole is better name
<testuser[m]1> oof
sam_sepi0l has quit [Quit: Textual IRC Client: www.textualapp.com]
Beni has quit [Quit: Client closed]
<virutalmachineus> how much are you talking about?
<virutalmachineus> 2.50?
chomwitt has quit [Ping timeout: 268 seconds]
<illiliti> usd per hour? about that yeah
<illiliti> i know people who survive on 250$ per month
<illiliti> most of them are pensioners
<illiliti> half goes on bills
<illiliti> some goes on medicine and stuff
<illiliti> some goes on food
<illiliti> and you have 0
ioraff has joined #kisslinux
<illiliti> should we enable new-dtags by default for binutils?
<ioraff> do you mean add that flag when building binutils or set that flag when the linker is bfd/gold?
<illiliti> i mean add configure option
<illiliti> --enable-new-dtags
<illiliti> also --enable-default-hash-style=gnu
<illiliti> for DT_GNU_HASH
<illiliti> it seems to be better than DT_HASH, but non-standard
rohan has quit [Ping timeout: 252 seconds]
rohan has joined #kisslinux