Zig! π whatever you do will be dope ime sure!
Zig! π whatever you do will be dope ime sure!
been dark. libscrapli shenanegins still on going. we're getting there though!
zig 0.14.x -> 0.15.x was mostly easy, but some gremlins floating around still where release-safe causes some segfaults but release fast/small does not?
will continue to chip away... slowly π€
Damnit Mike
lol how did I know to expect this from you π
i used jujutsu all day instead of git and i *think* i get some of the value but... it defo sent me on a spiral that i did not enjoy. will try to stick with it a bit more to see if i can get it smooshed into my brain. anyone else using it and have feels?
todo...?
* better/cleaned up logging
* "trace" or "debug" mode maybe?
* outlier platforms (fortios, mikrotik)
* docs
* ???
* profit?
2/2
more slow and steady progress on libscrapli/scrapli/scrapligo things...
* examples for py/go (same ones, in both langs)
* knocked out some old netconf related todos
* ???
1/2
big payloads exposed the issue (ty srlinux for making it easy to load absurd configs in clab!)
fixed by ensuring we flush writes when reading which sounds silly but letes the acks through basically
2/2
follow up on the libssh2 proxy jump debugging fun: as expected... it was... user error!
turns out I wasnβt forwarding acks from the βouterβ session to the proxied connection, so the server just stopped sending mid-stream.
1/2
not the most successful weekend on the libscrapli frontβ¦
proxy jump stuff is *mostly* working w/ libssh2, cli is fine, but netconf just stalls out when receiving capabilities
feels real bad when you think youβre checking something off the list then realize... not quite!
got libscrapli + libssh2 proxy jump thing working this weekend
not quite as easy as -J w/ openssh, but it does the trick!
thinking its time for docs and examples... closign in on an rc i think!
c9s now uses overlayfs if in a privileged pod (almost everyone since have to to run NOSs generally)
result is container fs drops from 21g -> 13g for a srl pod. and maybe/allegedly sorta quicker?
if you are bored and using c9s, check out 0.3.0 and lemme know if it all works!?
3/3
anyway, culprit is basically vfs. vfs is default storage driver for docker in a container i guess, probably because it works anywhere. but, its bad from a storage perspective and allegedly is slow(er?).
2/3
finally got around to looking at vfs use in c9s this weekend...
tldr some folks were unhappy about storage ballooning on launcher pods a bit. i never cared cuz storage is cheap :P
1/3
just got done w/ some time off -- was the least amount of dev/tech related stuff ive done in.... years?
i did manage to make some progress porting some scrapli community things into the new unified zig/go/py yaml format though π
some folks apparently donβt like the name βclabernetesβ
to them i say: kick rocks πͺ¨
its fun, funny, and it literally describes what the thing is
go fork (it) yourself if hate the name π€·ββοΈ
dont forget to put carbon and nim and a zillion other cool ones in there somewhere π but yes, zig is super dope!
not sure if just cuz more go/zig lately or just "growing up" but meh re these docs now?
just stuff like go doc and then, much more importantly, "wordy words" explaining thought process behind things
like why these args, what is actually happening when you call XYZ, how did we get here etc.
2/2
i used to care a ton about "api" style docs -- like docs showing funcs and args and return types etc.
i think this is/was because i used to python a lot more and typing being a suggestion meant docs were (imo) even more important
1/2
this weekend was an accidental lesson in garbage collection!
scrapli(py) test kills a netconf session, lazy me didnt free it. test passes locally, but fails 100% of the time in ci
root cause: the killed session was trying to log something to a logger that has been gc'd π
spot the problem? computers man. they do exactly what ya tell 'em too eh?
tmate, strace, lsof, gdb and a lot of pain later... i was unlocking + signaling *before* setting the stop flag. thread kept running. join hung forever.
computers are hard
β€οΈ π thanks dude!!
have been told i am shit at linkedface (havent been told that about bsky, maybe bad at this too?)
guess you are supposed to do clickybaity stuff or like nicely format things or whatever
im just gunna (try to) keep posting nerd stuff im into. if you like that, neat, we can be friends!
if not... meh?
for the 1, mayyyyybe 2 people deeply invested in my epoll/kqueue side quest:
it was. as nobody could have foreseen.
user error π
so yeah, epoll/kqueue things are looking mighty nice for libscrapli... will play more this weekend
went on a kqueue/epoll sidequest for libscrapli
attempting to replace tight loop over non blocking reads (w/ sleeps ofc) with kqueue
maybe (probably?) user error but not as performant as i was hoping, gotta spend some more time poking about
also need to check out io_uring π«
if you do developy things, go do this plz
blog.jetbrains.com/team/2025/04...
I always enjoy looking at the results, more data more good. go do it
had seen but not used popeyecli then ran across it again somewhere recently. am re-intrigued
static analysis stuff is such low effort for varying amounts of reward
the payoff doesnt really have to be much though.
this one feels like a cool tool to keep in the toolbelt
You had me till windows π€£π
still a huge jetbrains stan, but recently pretty much completely cnverted to zed
bit of faffing about w/ extensions and getting settings where i want, but its quite snappy, simple, and generally nice to use.
bonus points for non m$-ness of it all π