Is there any rolling release distro that already follows the suggestion of only distributing tagged releases?
Is there any rolling release distro that already follows the suggestion of only distributing tagged releases?
https://gcc.gnu.org/legacy-ml/gcc-announce/2000/msg00003.htm...
It's not a "nightly" distro, it's a "let's take patches that are unfinished, release them into the wild onto unsuspecting users, and let the upstream developers deal with it" distro.
The software that actually has good release hygiene pales into comparison to the software that is actually more stable by just pulling from main.
It's just that some distributions make a huge mess of it, don't want to work with developers and actively push off all the work to the developers for patches that have not even shipped, not even been merged or most painfully: not even been reviewed yet.
Then don't do 15 different versions of your software. I ask again: Why do i need GTK1, 2, 3 and 4 on my system ? Why KDE 5 is not compatible with KDE4 ?
Of course this is a very one sided view. I've been on the receiving end of reckless upstream projects far too often not to understand why distributions which are expected to maintain compatibility with releases for years dislike a fast moving upstream for anything too important.
Manjaro Linux is accused of something different far less justifiable: forking upstream projects in all but name by shipping heavily patched packages without supporting them and even worse putting the support burden and blame on the upstream projects when things break (and oh boy do they break).
But also this should mean that the buck stops with the distribution, not with the developers.
Here's one example where I've done it: https://github.com/NixOS/nixpkgs/blob/350fd0044447ae8712392c... since new rust has already been merged and rbspy still has no release with the required patch, there are two options: a broken package or an unreleased patch.
It's a cost/benefit calculation for the maintainers.
Like the fun example of manjaro shipping a merge request of the SMS/chat application on the PinePhone that causes a database upgrade that can't be rolled back. This code was not intended to be merged, even less shipped.
I'm guessing that what you did fits within the article's guidelines so long as you could do it as a tagged release.
I understand the desire to get an early warning of breaking changes coming from elsewhere, but surely scraping together all work-in-progress is likely to raise a lot of false concerns?
I think the issue isn't an individual patch that hasn't been completed, such patches are unlikely to be checked-in anywhere.
It is patches that are part of unfinished work, or that have not been integration tested with the rest of the product, or are not stable/useful/safe without other work that has not yet been merged. Basically patches intended fr other project devs, not the general userbase yet.
But to answer the question directly, in case it is literally unfinished patches:
> What is the use case for shipping unfinished patches?
“Being the latest & greatest” willy waving in end-user aimed distributions.
I can see a use case for some dev communities, though mainly for proactive compatibility testing purposes, not for dev or non-test deployment environments.
Then make absolutely sure your users all realise this, and that they should come to you for support not the devs (unless they themselves are devs, have read the relevant documentation, all of it, and able to offer useful help with the matter).
> But also this should mean that the buck stops with the distribution, not with the developers.
The clarifying points further down the article, that is the thrust. The basic rule for releasing WiP code being “don’t” and the advanced rule being “don’t, unless you really know what you are doing and can support it yourself”.
Void Linux