XIO

xWire

Posts from every feed QUEST has found across the web — blogs, news, podcasts, channels and more.

LWN.net comments

This feed contains the text of all comments posted to the LWN.net site.

I like it, and it could be even simpler in fact

It would be very nice to have this built into Python, and use actual Python syntax rather than a comment "hack": #!/usr/bin/env python3 """Print the latest version of each given PyPI package.""" # Haven't checked if a `script` stdlib already exists; bikeshed the name at your leisure. from script import script script( requires_python=">=3.11", dependencies=[ "requests>=2.34.2", ], # Specifying lock does 2 things: # # - Checks the hash of every installed wheel, raising an exception on mismatch. #…

Let's just improve lintian and standard deterministic upload machinery

I can't tell whether your questions are rhetorical or actual. I'm going to presume that they're meant as actual questions. without this LLM review process being run, would you have thought of that as a possible improvement? Yes of course. I've read historical discussions about what should happen to packages that enter the NEW queue, and these discussions include many ideas (from actual humans) that are substantially more sophisticated than the simple one that i proposed. Are you going to build…

Not a good decision

I'd add Gentoo to that list. :) "amd64" (i.e. 'stable') has been very stable for me for two decades. I feel like they actually jump through a lot of extra hoops to allow users to do things that some major distros just sweep away with unilateral decisions. For example, I'm still on X11+XFCE and OpenRC on my daily driver and still happy about it. And, my god, I still have a separate /usr and dracut handles creating a initramfs to handle it fine. Am I an 'ordinary' user? I don't know - but I…

Not a good decision

Tl;dr: Relying on Plasma 6+X11 for "anything serious" was already questionable given the upstream attitude -- but doing so on top of a distribution that has an explicit policy of tracking the bleeding edge of KDE releases was even more questionable. Yeah, relying on Fedora for stability is a mistake, and I say this as someone who has been using it on my everyday computer since Fedora Core 1. Fedora has always been aimed at getting the newest software into users' hands as quickly as practical;…

Not a good decision

> In a new version of Fedora, for example Fedora 45, released this week. So... you of course complained when every Fedora release for at last few years updated to a newer Plasma in the middle of its lifecycle? Look, I get that you're feeling inconvenienced, but at the end of the day, you were expecting something contrary to Fedora's (and upstream KDE's) explicit plan of record going back literally years now.

I switched to xfce

I knew the kde x11 removal was coming. A few weeks ago I spent about a week trying different desktop environment and I've settled on xfce. The hidpi support is good although requires changing the theme and window manager settings. It's best to tweak the font dpi and set the window theme to xhidpi or hidpi instead of using scaling under displays. I did give kde wayland a go but the blocker for me was that you can't remote into a headless login screen only a logged in session. Also the rdp…

"A mess" is right

IMHO the more correct behavior would be for Nix packages to strip this metadata and automatically convert it into a "proper" Nix container with the appropriate dependencies. But I don't know enough about the ins and outs of NixOS packaging to tell you whether that's feasible to automate. Probably it's at least easier than the old-school approach of writing import whatever at the top and hoping for the best.

hardware support

Is there a difference in support for old hardware between X11 and Wayland ? Well, X11 can be found on everything back to SPARCstations and SPARCbooks, so I'd say X11 does win this one, technically ;-) How important is that in practice is a different question.

Not a good decision

> At what point are users expected to take responsibility for their choices and divorced-from-reality expectations? In a new version of Fedora, for example Fedora 45, released this week. Software come and go all the time, maintainers job is too smooth that problem out and act as buffer between upstream and the users, that's one of the main reasons Linux distributions exists.

Not a good decision

> This might be correct in formal terms but it's up to maintainers to use their judgement, good taste and experience to bring the best possible solution for the user of the distribution At what point are users expected to take responsibility for their choices and divorced-from-reality expectations? > The sole valid reason for such drastic change would be huge security issues The "sole valid reason" is that the folks doing the actual work chose to do it that way. Which, I might add, is the same…

Not a good decision

> If removing X11 suppport during a stable release is ok, then any update goes, which means Fedora never can be trusted for anything serious any longer. > Let me quote TFA: >Gompa said that the X11 packages had been ""expressly deprecated"" and were expected to be removed as soon as > support was dropped by KDE upstream. ""Plasma X11 has also been unsupported since Fedora 40 when Plasma 6 was introduced."" This might be correct in formal terms but it's up to maintainers to use their judgement,…

pipx instead of uv as shebang line

> Missing from the article is also installing uv on hosts. I've found that be a pain, since it's not > included in most operating systems, and because of its fast update cadence, probably shouldn't be. Interesting. While editing the article, I did a 'dnf install uv' on Fedora 43 and got a version that worked for the examples in the article. But now I see that Debian and derivatives don't seem to have it. https://docs.astral.sh/uv/getting-started/installation/#s... It looks like it can be…

Not a good decision

> If removing X11 suppport during a stable release is ok, then any update goes, which means Fedora never can be trusted for anything serious any longer. Let me quote TFA: Gompa said that the X11 packages had been ""expressly deprecated"" and were expected to be removed as soon as support was dropped by KDE upstream. ""Plasma X11 has also been unsupported since Fedora 40 when Plasma 6 was introduced."" ~~~ Upstream KDE said X11 session was formally deprecated with the Plasma 6.0 release (Feb…

Not a good decision

IME, the only distributions you can really trust not to gratuitously break things in the middle of a release, are, in order of popularity, RHEL and its clones, Debian, Slackware, and Devuan.

pipx instead of uv as shebang line

The article touches it, but not in detail: you can use pipx instead of uv on the shebang line, which I think was available around pip 1.8: #!/usr/bin/env -S pipx run --path Missing from the article is also installing uv on hosts. I've found that be a pain, since it's not included in most operating systems, and because of its fast update cadence, probably shouldn't be. pipx, however, is included (e.g. the pipx package on Debian/Ubuntu). It will use the host OS's python interpreter (e.g. Python…

I like it, and it could be even simpler in fact

Also that way you can't specify versions except through further magic comments. Honestly the PEP is sufficient, the uv tool is really good (and I say this with deep loathing in my soul for OpenAI!) and the fact that Astral specifically went after the standards that already existed as implementing, rather than doing their own radical new thing, really made me feel better about it.

I like it, and it could be even simpler in fact

Instead of toml hidden as comment, one could just have external lib section in import list: import os import sys # external import request or simply have a list of libs in standard lib in solver tool, and install any unknown modules. No room for requirements on version, however that's a feature, not a problem, latest version of a module should always work, if found false, if module need to break things make it available under other name.

Thanks for the reporting

My guess is that the 'informed consent' is all the lawyers for Google would care about. If they had put on a sticker that this equipment would hack your other computers, be a proxy for criminal activity and whatever else.. even in 2 point font.. it would probably be a-ok. [I am joking.. it needs to be at least 4 point font.]

Why

Surely, but any device that cannot spare 150kiB is probably not going to be able to fit a driver for every device you could connect via USB. It isn't even likely to support such flexibility, since those systems tend to be embedded and have a very specific usecase, often not even having an exposed USB port in the first place. Not that I'm not in favor of efficiency for its own sake, but let's call it what it is if that's the case.

Let's just improve lintian and standard deterministic upload machinery

> You act as if building out and maintaining the Debian-specific LLM tooling comes at no cost (upfront or ongoing). Of course it has a cost. The question is: is it more or less than a dozen bespoke solutions? Adding rules to lintian is pretty cheap. But as soon as you start talking about maintaining a CI/CD infrastructure it becomes harder. Though of course, building such a bespoke solution is also cheaper with LLM assistance. In the end though, the choices are made by those doing the work. And…

Let's just improve lintian and standard deterministic upload machinery

> And then comes the question: building such an automated test also takes time and effort, and has to be maintained. Are you going to build it? If not, who then? Are the benefits of building such a bespoke solution greater than the costs? I mean, for every point in that list you could build a custom solution, but now you have solved them all in one go. You act as if building out and maintaining the Debian-specific LLM tooling comes at no cost (upfront or ongoing).