liblinux VS oil

> libc isn't really getting in the way here.

For the standard set of system calls, the libc is pretty great. For Linux-specific features, it could take years for glibc to gain support. Perhaps it's gotten better since then, perhaps it still takes years. I don't know.

Years ago I read about the tale of the getrandom system call and the quest to get glibc to support it:

https://lwn.net/Articles/711013/

A kernel hacker wrote in an email:

> maybe the kernel developers should support a libinux.a library that would allow us to bypass glibc when they are being non-helpful

That made a lot of sense to me. I took that concept and kind of ran with it. Started a liblinux project, essentially a libc with nothing but the thinnest possible system call wrappers. Researched quite a bit about glibc's attitude towards Linux to justify it:

https://github.com/matheusmoreira/liblinux#why

Eventually I discovered Linux was already doing the same thing with their own nolibc.h file which they were already using in their own tools. It was a single file back then, by now it's become a sprawling directory full of code:

https://github.com/torvalds/linux/tree/master/tools/include/...

Even asked Greg Kroah-Hartman on reddit about it once:

https://old.reddit.com/r/linux/comments/fx5e4v/im_greg_kroah...

Since the kernel was developing their own awesome headers, I decided to drop liblinux and start lone instead. :)

It gives you access to 100% of Linux's system calls. It eliminates a lot of global state. It gets rid of a lot of legacy libc crap.

Years ago I wrote a fairly referenced rationale in my liblinux project:

https://github.com/matheusmoreira/liblinux/blob/master/READM...

> Now, do I think it would make total sense for syscall wrappers and NSS to be split into their own libs (or dbus interfaces maybe) with stable ABIs to enable other libc's, absolutely!

I worked on this a few years ago: liblinux.

https://github.com/matheusmoreira/liblinux

I'm not developing it anymore though because I found out the Linux kernel itself has a superior nolibc library:

https://github.com/torvalds/linux/tree/master/tools/include/...

It used to be a single header but it looks like they've recently organized it into a proper project!

I wonder if it will become some kind of official kernel library at some point. I asked Greg Kroah-Hartman about this and he mentioned there was once a klibc:

https://old.reddit.com/r/linux/comments/fx5e4v/im_greg_kroah...

> This is something the BSD's got absolutely right.

BSDs, every other operating system really, force us to use the bundled C libraries and the C ABI. I think Linux's approach is better. It has a language-agnostic system call binary interface: it's just a simple calling convention and the system call instruction.

The right place for system call support is the compiler. We should have system_call keywords that cause it to emit code in the aforementioned calling convention. With this single keyword, it's possible to do program literally anything on Linux. Wrappers for every specific system call should be part of every language's standard library with language-specific types and semantics.

I'm not using this stuff professionally, it's just my own home lab's virtual machines with little services implemented as freestanding C programs. Not doing anything fancy right now, much of it was just to see if I could do it.

I've seen other people commenting here on HN saying they're using the same approach so it's defenitely not my invention.

I published some of my work in the form of a liblinux that I use to make system calls:

https://github.com/matheusmoreira/liblinux

I'm not developing it anymore though because I found out the kernel itself has a nolibc library:

https://github.com/torvalds/linux/tree/master/tools/include/...

It used to be a single header but it looks like they've organized it into a proper project.

That's awesome. I didn't know about rwildcard until now. Is it part of GMSL? I searched for rwildcard on gmsl.sourceforge.io but didn't find it.

I think my function is needlessly complicated compared to rwildcard. Here's my code:

https://github.com/matheusmoreira/liblinux/blob/modular-buil...

https://github.com/matheusmoreira/liblinux/blob/modular-buil...

The file? and directory? functions were inspired by GMSL.

I wrote a general recursion function. It takes a function to apply to lists and a function to compute whether an element is a base case.

The recursive file system traversal function applies a directory globbing function to the list of paths and has file? as base case.

The find function filters out any items not matching a given predicate function. It was my intention to provide predicates like C_file? and header_file? but I stopped developing that project before that happened.

I think rwildcard is probably simpler and more efficient!

will prevent almost all of the "silent footguns".

YSH has strict:all and then a bunch of NEW features.

There's been good feedback recently, which has led to many concrete changes. So your experience can definitely influence the language! https://github.com/oilshell/oil/wiki/Where-To-Send-Feedback

Regarding writing tools/tests/benchmarks in bash+Python, vs. writing tools in your main language:

I think we might eventually concede that something Debian-like is the “standard development environment” (at least for server side stuff, i.e. not iOS apps)

In this case, bash+Python is a non-issue. It works extremely reliably. That’s actually why I use it! Everything else seems to break, or it’s really slow (node.js is a very common alternative).

- Microsoft conceded this back in ~2017, by building Linux into their kernel with WSL, and providing Ubuntu on top

Yes bash + Python is a disaster on Windows (I have scars from it), but Microsoft agrees that the right place to solve that is in Windows :-)

- Every CI system runs Debian/Ubuntu

- Every hosting provider runs Debian/Ubuntu

- Every online dev env like gitpod.io provides Debian/Ubuntu

This is somewhat related to remote dev envs: https://lobste.rs/s/ucirlx/lapdev_self_hosted_remote_dev

One vision for https://www.oilshell.org/ is that the CI environment is the dev environment is the hosting environment.

Everything is just an equal node in a distributed system. BUT it’s more git like, in that you explicitly sync and work “locally”, wherever that is. You don’t have the network chatter and flakiness of “the cloud”.

Oils has a very large set of monotonically increasing properties too - https://www.oilshell.org/release/0.21.0/quality.html

All that is bash+Python that is run on every commit, and it’s extremely good at catching bugs and perf regressions.

I’m skeptical that any project has that level of quality automation written in pure Rust or Zig. More likely it’s a bunch of cloud services with YAML.

Also a bunch of “hard-coded” toolchains that you can’t script with bespoke code. Like some shell commands in your package.json, which is just a worse way of writing a shell script.

Our quality process is all self-hosted, in the repo, and runs on both Github Actions and sourcehut - https://www.oilshell.org/release/0.21.0/pub/metrics.wwz/line...

bash and Python runs perfectly on Github Actions and sourcehut, with zero change. Containers also do.

(Although we need to unify the CI and release, because the release runs on 2 different real hardware machines, while CI is cloud only.)

Also, a main point Oils is that bash now has another highly compatible, spec-driven implementation – OSH. Having 2 independent implementations is something newer languages don’t have.

(copy of lobste.rs comment)

in your bashrc to enable it. I've used it for probably ~18 years now.

It also works with https://www.oilshell.org/ since we use GNU readline. Just 'set -o vi' in ~/.config/oils/oshrc

No other shell does that.

But I didn't know it was called MULTIOS until now. (I guess that's read "mult I/O's"? I have a hard time not reading it was multi-OS :) )

It seems a bit niche to be honest, but it's possible to support in Oils.

---

Oils also uses Unix domain sockets already for the headless shell protocol

https://github.com/oilshell/oil/wiki/Headless-Mode

We could do something like dgsh, but so far I haven't seen a lot of uptake / demand. Every time it's mentioned, somebody kinda wants it, and then it kinda peters out again ... still possible though.

I think flat files work fine for a lot of use cases, and once you add streaming, you also want monitoring, more control over backpressure/queue sizes, etc.

which works well. You don't have to clean when rebuilding variants. IMO this is 100% essential for writing C++ these days. You need a bunch of test binaries, and all tests should be run with ASAN and UBSAN.

---

I wrote a mini-bazel on top of Ninja with these features:

https://www.oilshell.org/blog/2022/10/garbage-collector.html...

So it's ~1700 lines, but for that you get the build macros like asdl_library() generating C++ and Python (the same as proto_library(), a schema language that generates code)

And it also correctly finds dependencies of code generators. So if you change a .py file that is imported by another .py file that is used to generated a C++ header, everything will work. That was one of the trickier bits, with Ninja implicit dependencies.

I also use the Bazel-target syntax like //core/process

This build file example mixes low level Ninja n.rule() and n.build() with high level r.cc_library() and so forth. I find this layering really does make it scale better for bigger projects

https://github.com/oilshell/oil/blob/master/asdl/NINJA_subgr...

Some more description - https://lobste.rs/s/qnb7xt/ninja_is_enough_build_system#c_tu...

This is sort of a category error...

re2c is a lexer generator, and YAML and Python are recursive/nested formats.

You can definitely use re2c to lex them, but it's not the whole solution.

I use it for everything possible in https://www.oilshell.org, and it's amazing. It really reduces the amount of fiddly C code you need to parse languages, and it drops in anywhere.

SEEKING VOLUNTEERS - https://www.oilshell.org/ - https://github.com/oilshell/oil/

I'm looking for people to help fill out the "standard library" for Oils/YSH. We're implementing a shell for Python and JavaScript programmers who avoid shell!

On the surface, this is writing some very simple functions in typed Python. But I've realized that the hardest parts are specifying, TESTING, and documenting what the functions do.

---

The most recent release announcement also asks for help - https://www.oilshell.org/blog/2024/01/release-0.19.0.html (long)

If you find all those details interesting (if maybe overwhelming), you might have a mind for language design, and could be a good person to help.

Surveying what Python and JavaScript do is very helpful, e.g. for the recent Str.replace() function, which is nontrivial (takes a regex or string, replacement template or string)

But there are also very simple methods to get started, like Dict.values() and List.indexOf(). Other people have already contributed code. Examples:

https://github.com/oilshell/oil/commit/58d847008427dba2e60fe...

https://github.com/oilshell/oil/commit/8f38ee36d01162593e935...

This can also be useful to tell if you'll have fun working on the project - https://github.com/oilshell/oil/wiki/Where-Contributors-Have...

More on #help-wanted on Zulip (requires login) - https://oilshell.zulipchat.com/#narrow/stream/417617-help-wa...

Please send a message on Github or Zulip! Or e-mail me andy at oilshell dot org.

This is true, but then the corrolary is that new PRs need to come with this higher and rigorous level of test coverage.

And then that becomes a bit of a barrier to contribution -- that's a harness

I often write entirely new test harnesses for features, e.g. for https://www.oilshell.org, many of them linked here . All of these run in the CI - https://www.oilshell.org/release/latest/quality.html

The good thing is that it definitely helps me accept PRs faster. Current contributors are good at this kind of exhaustive testing, but many PRs aren't