<illiliti>
you have to deal with them because POSIX doesn't have `local` feature
<illiliti>
also, note that `,`, `:`, ` ` characters are banned
<illiliti>
if you pass them to input, the output may be screwed
<illiliti>
this is limitation that can't be resolved
chaffity has joined #kisslinux
<schillingklaus>
can't the bad characters be escaped?
gtms has joined #kisslinux
<illiliti>
probably
geekthattweaks has quit [Quit: Connection closed for inactivity]
strajder has quit [Quit: leaving]
<chaffity>
I'm using a namespace to to run glibc programs (MongoDB for work), and it works brilliantly, but I'm having an issue taking this one step further by using overlay mounts to simplify the flow... Has anybody had any success with overlay mounts?
<testuser[m]1>
wdym overlay mounts ?[27;2;13~
<chaffity>
To clarify, I'm creating an overlay, but I just get "Out of memory" from mount()
<chaffity>
testuser[m]1 mounting a union filesystem
<chaffity>
For proprietary glibc stuff I have a separate /usr directory that I'm currently mounting on top of my regular /usr directory
<testuser[m]1>
have you looked into bubblewrap ? i think it might satisfy your case
<chaffity>
Hrm, user namespaces is another thing I actually had issues with... Running "unshare -r" reports an error, and I'm not sure if that's a BusyBox bug, or another issue entirely
<chaffity>
Thanks for the link though, will check this out more
<testuser[m]1>
maybe it's an issue in your kernel config ?
<testuser[m]1>
CONFIG_USER_NS=Y
<chaffity>
You hero
<chaffity>
That's one less thing, and bubblewrap (whilst a lot bigger than my lowly single file "glibc" environment jobbie) seems to be along the lines of something else I was pondering on a while back, thanks!
<midfavila>
bugs, right. I believe I encountered one relating to GCC. using busybox tar results in a failure, claiming that the liblto plugin would be an unsafe symlink
<midfavila>
and using libarchive results in such an amazing series of segfaults that it crashes X, or in the case of a tty, boots you to login
<midfavila>
fwiw, when using bbox tar, KISS reports that said plugin is missing from the manifest but fails sanely
<testuser[m]1>
This is during installation of the gcc tarball ?
<testuser[m]1>
I guess uploading it somewhere would hell
<testuser[m]1>
Help
<midfavila>
yeah, specifically extraction
<midfavila>
can't upload it rn, gotta build it again on my laptop
<midfavila>
which is going to take like three hours
<midfavila>
:|
rio6 has quit [Quit: WeeChat 3.0.1]
<testuser[m]1>
oof
<midfavila-laptop>
the 800mhz blues
rio6 has joined #kisslinux
<midfavila>
right, that's just started.
banshie has quit [Ping timeout: 244 seconds]
mahmutov_ has joined #kisslinux
chaffity has joined #kisslinux
<chaffity>
testuser[m]1 thanks to you pointing out my missing "CONFIG_USER_NS=Y" I now have DRM content streaming in Firefox... KISS can now be my absolute main driver for work and personal use \ (*o*) /
<dilyn>
I was just looking at that meson replacement last week, glad to see somebody did something!
<dilyn>
I'm gonna have to check it out...
<acheam>
muon requires meson to build
<acheam>
which is....
<acheam>
annoying
<acheam>
idk if it can build itself
<acheam>
its especially silly because upstream used a makefile, and they rewrote it for muon
<omanom>
>muon requires meson to build
<omanom>
its become the very thing it sought to destroy!
<dilyn>
lol yeah at least the other one had a configure script
<dilyn>
oh jeez i can't read...
carbsbot has joined #kisslinux
carbsbot has quit [Quit: carbsbot]
<acheam>
dilyn: i was surprised to see make as an implicit dep
<acheam>
are you at all interested in differentiating between make implementations, or will "make" always refer to gnu make?
<acheam>
lots of packages have "make make"
<dilyn>
do a lot of packages have make make?
<dilyn>
the reason for marking toolchain related things as implicit is so that nobody has to worry about what toolchain any user has. unless you specifically need something (like lld requiring llvm, or ccls requiring clang, etc)
smartin has quit [Quit: smartin]
cem has quit [Quit: connection reset by purr]
Uks2 has quit [Ping timeout: 244 seconds]
Uks2 has joined #kisslinux
cem has joined #kisslinux
cem has quit [Client Quit]
cem has joined #kisslinux
carbsbot has joined #kisslinux
an3223 has joined #kisslinux
midfavila-laptop has joined #kisslinux
mahmutov has joined #kisslinux
gtms has quit [Remote host closed the connection]
midfavila has quit [Remote host closed the connection]
carbsbot has quit [Quit: carbsbot]
<zenomat>
90% of the used k&r offers are more expensive than a new copy...
<zenomat>
the cheapest one is 22 euro, but i'd have to import it from the usa...
GalaxyNova has joined #kisslinux
<GalaxyNova>
I don't think packages should imply that Git or Busybox is installed
<GalaxyNova>
assume*
<rio6>
if it doesn't need busybox specifically (ie, using coreutils, toybox, etc still works), I don't think busybox should be a dependency
<GalaxyNova>
i guess
<GalaxyNova>
but generally i believe it would be better to list dependencies explicitly
<GalaxyNova>
and not assume anything about the user's system
<GalaxyNova>
rio6: If a package were to use only POSIX stuff I could see that
<omanom>
busybox is there, and if it isn't that means either a broken system or the /user/ explicitly decided to use an alternative. meaning the user should know they might need to adjust packages to support their alternative choice.
<omanom>
i think this also satisfies one of the guidestones: "The user is smart, the package manager is dumb. The package manager is written under the assumption that the user has some kind of functioning brain in their skull"
<midfavila-laptop>
big agree
<midfavila-laptop>
GalaxyNova, you can rest assured that outside of sed -i, everything in community and core should be posix
<midfavila-laptop>
at least in my experience, I've never encountered otherwise
<midfavila-laptop>
(ignoring, of course, packages with explicit build deps)
<midfavila-laptop>
also, zenomat, you can always acquire a copy from Library Genesis
<GalaxyNova>
What I'm worried about is that things like Gkiss will be harder to maintain because all the packages will start excluding binutils or gcc from dependencies
<GalaxyNova>
or GNU make
<dilyn>
It's implicit precisely BECAUSE the user can change them, and we want to allow them to change them without having to fork every package
an3223 has quit [Quit: leaving]
<dilyn>
for instance, see wyverkiss. the only reason it can leverage submodules is because we don't explicitly state 'binutils required'
<dilyn>
at least, leverage them to the degree that it does
<dilyn>
most of the packages get to be symlinks
<midfavila-laptop>
i'm still not sure what the problem here is...
<midfavila-laptop>
implicit dependencies aren't a problem. by establishing a base system and only working with things in that base, portability is ensured. if the user deviates from that base in some critical way (e.g removing their build toolchain), that's on them if they end up wearing egg on their face
<dilyn>
having a functioning toolchain be an implicit dependency on a source based distro is an obvious thing to me
<dilyn>
the only one i take issue with is git, and it's not even really a dependency
<midfavila-laptop>
meh, you don't need to have a build toolchain as long as there are binary packages available
<midfavila-laptop>
i still intend to experiment with that...
<dilyn>
you can not have git just fine -- it's just implicit so we don't need it for git sources
<dilyn>
i mean, sure. but the user who does that is smart enough to do it. and in the case of using binaries... you aren't specifying toolchain components
<midfavila-laptop>
well, yeah. i was just saying that KISS can be used as a binary distribution without much, if any difficulty
* midfavila-laptop
shrugs
<dilyn>
i mean, yes
an3223 has joined #kisslinux
an3223 has quit [Quit: leaving]
an3223 has joined #kisslinux
an3223 has quit [Remote host closed the connection]
an3223 has joined #kisslinux
<acheam>
i've released chorizo 1.0.0 for those who are interested