I kind of wish there was a version of Alpine that was libc based - just for desktop NVIDIA support really.
Got some complaints the #musl git server was sometimes giving errors from too much load. Load caused by abusive LLM scrapers hammering cgit. So I added frontend rules that apply a 10 byte/sec rate limit to client IPs hammering >5 requests per 10 sec for the duration of the hammering. Load average has plummeted.
Oh great, it doesn’t boot. Init fails to launch.
Is it because I used musl instead of glibc? I’ll never know because NO FUCKING ERRORS ARE OUTPUTTED.
Someone, please tell me how to try make a distro with musl… I’m really stupid.
Do I use another #Distro like alpine as a base or?
Thought that would probably make me a lot more productive on #musl, musl-cross-make, etc.:
When there's an open issue that needs any sort of testing cycle feedback, scheduling real-time chat with the reporter or patch submitter to go through it until it's resolved.
That's what just finally cleared some backlog on multiple longstanding musl-cross-make failures.
Heads up maintainers: there's a static analysis tool Svace, and perhaps some AI-slop tool built on top of it, producing very wrong analysis. So far #musl has gotten 6 patch submissions based on it and they're all wrong and mostly unmotivated.
In one case it claimed possible null pointer dereference when the code in question was already immediately conditional on the pointer being non-null.
FWIW this is what Samsung uses to find and "fix" bugs in their fork of Android...
Document I was searching for in case anyone is curious:
#musl #vulnerability https://www.openwall.com/lists/musl/2025/02/13/1
I guess #openwrt and #alpine system should be all upgraded now once fixes are available.
So, from my notes:
#voidlinux: symlink a service's name from /etc/sv/ to /var/service to enable a service then sv up & down will become available for that service.
#chimeralinux: dinitctl, kinda like systemctl
#devuan: update-rc.d <service> default to get it on the runlevels, then enable/disable to do fun things. Only do update-rc.d remove when the package is removed from the system.. Gotcha
@krishean that's not how #systemd works.
SystemD was created because #SysVinit was shit and noone fixed it or made something better.
#Wayland is the future as #Xorg is being #EoL'd.
For the shitty #GlibC we have alternatives like #bionic and espechally #musl!
@simontatham Pardon my straying from the joke, but FYI: #musl libc has a super compact representation for finding the rule for case mapping for a particular character. Under 5k for tables plus code. We don't have the car emoji mapping tho.
#musl mallocng just caught an interesting type of buffer overflow in mc: decrement of offset past start of buffer and overwrite of byte below allocated memory with '\0'.
@ireneista @marshray @gregly @rootfake @GavinChait @cwebber Reading bug reports for building software against #musl, I see examples again and again where the programmers have copied some Rube Goldberg machine of ifdefs or other hacks where there was a simple portable way to do what they needed.
@lavxnews point is: "reinventing the wheel" is rarely easy nor worth the effort.
That's why I use #Linux, #toybox and #musl for
OS/1337 ...
Where it always comes up: we run into code that's either not compiling with #musl or compiling but then failing to do the right thing at runtime, because it's doing some convoluted nonportable dance in place of a direct and simple thing that would have worked on any system.
And the bad thing it's doing is a pattern that's somehow become popular, probably from a few major pieces of sw doing it.
And I want to say it's "cargo culted", but that term perpetuates a racist myth..