telemetry_registry do not need to be a runtime dependency either. In fact in Elixir it has macro requirement only because module attributes aren't preserved by default. In Erlang you can expose that metadata without any dependency.
today, we're announcing our €3,8M ($4.5M) seed financing round, led by byFounders with participation from Bain Capital Crypto, Antler, Thomas Dohmke (former CEO of GitHub), Avery Pennarun (CEO of Tailscale) among other incredible angels.
read more on what's next: blog.tangled.org/seed
I may have solution in future (hopefully near).
Wouldn't it be easier to use existing VFS protocols like 9p instead of creating new one?
Do you have a link to that? I would be really interested in checking that out.
What is that?
Hello #ElixirLang
Does anyone look for highly skilled Elixir developer based in Europe for remote work or consulting services?
I am open to opportunities.
I call them "human sized"
Move already so you will not forget ;) you can do so and it should be free
Have you checked OVH?
Ugh, hardly tripped over deadlock in #Erlang that required me to use private API for now. Break my heart, but it is what it is.
A lot of people are using this protocol indirectly - WSL2 is using that to expose Windows FS to Linux in VM. But outside of the VMs it is IMHO not used enough, as it is IMHO great alternative to FUSE.
Currently sysfs is read only, but there are applications, processes, and system information exposed. The API for implementing own FS will be very much in flux, but this is MVP to show that it works. It is mountable by 9pfs FUSE system and v9pfs on Linux.
cc @arathunku.com as you seemed interested
After quite some time I present you `e9p` - 9p implementation in pure Erlang. Built in implementations provide Unix Filesystem proxy and sysfs - pseudo filesystem that exposes ERTS information in similar way to `procfs` and `sysfs`.
hex.pm/packages/e9p
#elixirlang #erlang #9p
For me it looks more like a screenshot from Death Stranding
Yes, but thing bigger. Not only FS for Erlang processes, but a library for implementing FS in Erlang for whatever you want.
It is possible to do it right now via eFUSE (github.com/mwri/erlang-...), but that requires NIF and OS that support FUSE.
I am not 100% sure. You would need to check, as I haven't dug deep enough to remember that.
Just a hint into what the goal is (it is not fully working yet, but that is what I am trying to achieve):
$ cat /erlproc/registered/logger_sup/state
{state,{local,logger_sup},
one_for_one,
{[default,logger_proxy,logger_handler_watcher],
…
All without any NIF
In theory that would be equivalent to `__block__(1, 2)`, but that will be rejected by the compiler/interpreter. Elixir is funky sometimes.
That is not 100% correct. `()` and `(1; 2)` (`;` can be replaced by newline there) is equivalent to "meta node" in AST called `__block__` (which is documented in `Kernel.SpecialForms` - hexdocs.pm/elixir/Kerne...).
It should work everywhere, as if OS will compute it independently, then it should just replace that value with proper one on their own.
It is awesome to see my code inspire others to do other awesome things.
BTW currently I am working on completely bonkers concept that can open completely new interesting scope for others, and I think it will be really nice for stuff like @nerves-project.org.
That one was tricky. And IIRC it was needed only on one of the supported OSes (do not remember which one) as another one was happily calculating that for you. It was "fun" testing it on other OSes (maybe one day someone will test it on BSD as well).
I have tested it on Linux and macOS (there are differences in how these handle user space ICMP sockets). It also supports IPv6 which I believe yours do not.
FYI I already have created wrapper over `socket` for ICMP - github.com/hauleth/gen_...
It is not on Hex as I do not want to push it without automated tests, but maybe I should do it anyway?
`:bar` as what you have inside parens is a code block. It is the same as
if true do
:foo
:bar
end
Just written in a funky way.
It is called Duckex (hex.pm/packages/duc...) and it uses Port instead of NIF (as other DuckDB client libraries do).
The advantages are obvious - bug in DuckDB (or loaded extension) will not cause BEAM VM to fail and it also helps with RAM handling (as OOM killer will handle subprocess instead of VM)
Heh, I couldn't attend CodeBEAM this year (unfortunately, maybe next year), but I have similar experiences.
So because NIFs are hard, and in general can be problematic when used with 3rd party, unsafe, code, myself and friend( who is not on Bsky) have created new @duckdb.org client for #elixirlang.
Why not use `image` library or if someone is really daring - using `wxImage` from OTP.
