Alexander Martin (alm) @alexbuzzbee@fosstodon.org

I would like to remind everyone that if I follow someone or boost their post, that does not mean I think they are a good person. It means I find their posts interesting, and that is it.

Are there any notable forks of MS .NET?

How do I tell if a Python asyncio StreamReader's underlying TCP connection has been closed at the other end?

Welcome to "Being the Operating System," an exercise in frustration with application programmers.

Python's threads are awful, because they have all the downsides of preemptive multitasking in a shared address space with no actual parallelism. Its async, however, is amazing.

(I've never worked with multiprocessing much, but a casual look at the documentation makes me think it's probably okay.)

Static linking but every symbol is in its own .o so link-time dead code elimination happens automatically.

Part of what I'm building here is a dynamic linker for Python, despite the fact that Python already has a good dynamic linker.

It turns out that Python pickles don't care about excess data at the end of the file, and it is well-known that ZIP files can be cleanly appended to other data. I have used this knowledge to create a system that stores pickled metadata about Python zipimport packages as a preamble to the ZIP file.

I am now gradually building an overly-complicated runtime environment based on this concept. The current problem is how to build the bundle that builds bundles.

POLL: What is your favorite command-line help system?

I downgraded to 5.4.13, my previous kernel, so that I can get through the installation.

Maybe CPU ucode patches aren't getting installed or something...

Something is definitely causing spurious protection faults, which is somewhat alarming (mostly it just causes things to segfault for no reason, but sometimes it causes a kernel oops and a lockup, what fun). I suspect it may have something to do with kernel 5.5.13, because it seemed to start after I upgraded to that version. I'll bump to 5.6 later and we'll see if it gets better.

I think this may be my fourth or fifth time reaching the emerge @world step of the Gentoo installation process, but none of those attempts have yet reached a graphical environment. The hope is that this one will finally get there now that I've nailed the biggest issue in the last attempt. And I still have said previous attempt as a template.

Emerging 1 of 263

Considering transforming the current LENS collection of repositories into a monorepo with multiple crates in it to ease building, versioning, and cross-crate coordination in general; anyone have thoughts?

I set TMPFS_XATTR but I still can't set them...

I still have not been able to test because UML is causing abnormal amounts of trouble with extended attributes.

Fun fact: If you try to cat /proc/sys/fs/binfmt_misc/register, you get "Invalid argument" (very much not a normal error message from cat), probably because it's write-only. I wonder why it's not "Operation not permitted" or something else more appropriate?

I managed to get it to use my real rootfs mounted read-only, which I guess works well enough? But it turns out I forgot to actually turn on binfmt_misc, and when I did I got ~3 compiler errors.

So I'm trying to test in user-mode Linux, and it compiles (yay) but I can't get UML to start (boo) because it can't find the rootfs (???). This isn't related to my changes; it's just because I've never used UML before.

Oh, file.path.dentry is exactly what I need.