DDraceNetwork - developer

ZombieToad

if you could provide proof for me being constantly racist that would be good

Screenshot your discord server chat at any moment of time

01:31

If you search the n-word on his discord server discord crashes

@Learath2 why did glibc decide to link nss the way it does?

08:16

https://stackoverflow.com/questions/3430400/linux-static-linking-is-dead

Linux static linking is dead?

In fact, -static gcc flag on Linux doesn't work now. Let me cite from the GNU libc FAQ: 2.22. Even statically linked programs need some shared libraries which is not acceptable for me. What ...

08:16

i was curious so i searched a bit xd

08:17

like why hardcode this

08:21

https://utcc.utoronto.ca/~cks/space/blog/programming/Go121LinuxStaticToolchain

08:22

@Jupstar ✪ https://github.com/pgosar/ChatGDB

GitHub - pgosar/ChatGDB: Harness the power of ChatGPT inside the GD...

Harness the power of ChatGPT inside the GDB debugger! - GitHub - pgosar/ChatGDB: Harness the power of ChatGPT inside the GDB debugger!

Ryozuki

@Learath2 why did glibc decide to link nss the way it does?

Well I think the reason was to have consistent behaviour across all programs on the same system, e.g. one program resolves a hostname one way and another a different way would be undesirable

NSS allows using different services (e.g. NIS, files, db, hesiod) by just changing one configuration file (/etc/nsswitch.conf) without relinking any programs.

08:25

what?

08:25

this allows u to change behavious

08:25

without the one the program was tested in

static linking was most epic in the past and it will be most epic in future

cuz static has more info so more opts

08:25

and perf is top 1 priority in this world

yes

Ryozuki

@Jupstar ✪ https://github.com/pgosar/ChatGDB

cool idea actually

Mh, I'm still unsure about that. Some libraries being truly shared is important for consistent behaviour

u mean like the graphics driver part that is in the kernel?

08:26

and needs communication?

08:27

basically all kernel stuff

That part can't be static anyway. No I was talking about things like curl e.g.

"That part can't be static anyway" depends xddd

08:27

we see us in 50 years

what should be made is a stable kernel interface for syscalls that doesnt rely on libc

when i released my own os

08:27

that even beats ryoos

Ryozuki

what should be made is a stable kernel interface for syscalls that doesnt rely on libc

would be interesting

You link to curl statically, it could use one source of trust and the system one another, which might be verybad(tm)

Jupstar ✪

"That part can't be static anyway" depends xddd

Some parts of it maybe, but as far as I'm aware, there is truly global state in most drivers

yep, question is if that implies it to be static linked, or if the global state simply lives

Ryozuki

what should be made is a stable kernel interface for syscalls that doesnt rely on libc

Hm, I think this is already there, you can do syscalls with just a couple instructions

e.g. lets say there is a global mutex that simply exists for this purpose

08:30

then all usages of lets say a global memory pool, would be safe to exist

08:31

well to some degrees there might still be a very small set of shared functions between stuff

Jupstar ✪

yep, question is if that implies it to be static linked, or if the global state simply lives

You could isolate such global state into a kernel part of the driver and the rest can be statically linked. But is that strictly better for performance? Now the code that could be certain about state needs to wait on a mutex

but if that interface is stable

Anyway, idk. It certainly is possible to split it. Just not sure if it strictly implies better performance, since dynamic linking is pretty cheap, only an indirect call away

Learath2

You could isolate such global state into a kernel part of the driver and the rest can be statically linked. But is that strictly better for performance? Now the code that could be certain about state needs to wait on a mutex

well it has to be minimized ofc, and mutex was just to illustrate the idea. but what i'd see is a kernel that is so micro that it really just prevents data races

Learath2

Anyway, idk. It certainly is possible to split it. Just not sure if it strictly implies better performance, since dynamic linking is pretty cheap, only an indirect call away

i dunno either, but i want to know