Package Details: crowdsec 1.6.1-1

Git Clone URL: https://aur.archlinux.org/crowdsec.git (read-only, click to copy)
Package Base: crowdsec
Description: The open-source and collaborative security suite
Upstream URL: https://www.crowdsec.net
Licenses: MIT
Submitter: solveit
Maintainer: mhdi
Last Packager: mhdi
Votes: 20
Popularity: 0.98
First Submitted: 2021-12-01 20:49 (UTC)
Last Updated: 2024-04-17 12:40 (UTC)

Latest Comments

1 2 3 4 5 6 Next › Last »

Exentio commented on 2024-04-08 20:17 (UTC) (edited on 2024-04-08 20:18 (UTC) by Exentio)

@xiota that makes sense, I saw that the wizard script basically uninstalls and reinstalls crowdsec by deleting all the binaries but maintaining all the config files including keys and such, which is probably why running the older commits that ran the scripts worked and the current ones don't. I won't go into details because I admittedly didn't read the whole script nor do I know exactly how pacman tracks files.

Would the new PKGBUILD address the issue when upgrading from the newer versions (1.6.0-0 and over) that never used the previous scripts? If it does, I suggest implementing a message that discourages upgrading and suggests a complete reinstall for the older versions, thus addressing the fact that people often don't read AUR comments; otherwise, addressing the issue of untracked files should be prioritized. I wish I could help but this is where both my knowledge and free time come short

xiota commented on 2024-04-08 20:05 (UTC) (edited on 2024-04-08 20:05 (UTC) by xiota)

@Exentio The error you reported last week, about missing libabsl file, is probably from old files left over from poor packaging of previous versions of this package. pacman could not track files because they were installed by script. The following command should show most of the troublesome files, which should be deleted:

find /usr/local -type f,l

Exentio commented on 2024-04-08 18:29 (UTC)

@xiota I have no doubts that the package was already a mess, I've read most of the comments and as I said before the necessary process to upgrade shouldn't be this complex in the first place.
This doesn't change that after changing the scripts the situation got worse because it lacks necessary steps.
I'm not "complaining", I'm respectfully doing my part in explaining that things need to be rolled back to at least have the package in an almost working state, I don't think you can blame the user for not even having the necessary scripts to follow official instructions.
Maintainers need user feedback too, they're not omniscient and they're allowed to make mistakes especially when it's for community projects like the AUR, and if they want to do their part they have to take that feedback and correct their mistakes.
What I'm doing is providing user feedback to help the community as a whole, despite lacking the mental energy to do so, otherwise I could've stayed quiet and spun a Docker container. It's not that deep.

xiota commented on 2024-04-08 17:59 (UTC)

@Exentio Most of your complaints are user responsibility. The package was a mess before @mhdi adopted it.

Exentio commented on 2024-04-08 17:37 (UTC) (edited on 2024-04-08 17:37 (UTC) by Exentio)

@mhdi I don't have a test system to test the changes and I can't mess with my server due to its importance, crowdsec protects my public-facing sites and services, but I don't think much changed as the install_collection function alone probably won't do much, the --upgrade argument for wizard.sh doesn't even call it, it instead it runs update_full, but this shouldn't be done indiscriminately: what's needed is a check of the version and running the upgrade commands as needed. Again, check the previous commits.

You also shouldn't suppose that people will always read the AUR comments and follow your suggestion to uninstall and reinstall instead of upgrading, unless you meant that it could fix my install, but after the process explained before my crowdsec install is running just fine.

I appreciate that you're willing to maintain the package, but I think you should get some help from more competent people if the reason why it's important to run crowdsec's scripts still isn't clear to you. I'm admittedly not skilled enough to write proper install scripts, I barely managed to fix my install when I noticed that things were being handled differently than what was instructed by the official docs. I don't like how communities are walled behind Discord these days, but I think crowdsec's server might be able to help you if you want to do a good job at maintaining the package.

I also think that a properly sustainable upgrade process shouldn't require such scripts to be run every time, but it is what it is.

mhdi commented on 2024-04-08 15:59 (UTC) (edited on 2024-04-08 15:59 (UTC) by mhdi)

@Exentio I've updated the PKGBUILD and made some improvements. Could you please review it and let me know if there are any issues?

I kindly request removing the previous package and install the new one instead of updating it.

Exentio commented on 2024-04-08 12:34 (UTC) (edited on 2024-04-08 13:01 (UTC) by Exentio)

@mhdi I find it very hard to believe that the few lines you added were enough since I had to restore the files myself and run the scripts in order to fix the program. I have the wizard.sh file before my eyes in this exact moment - it does way more steps than the current crowdsec.install file in this repo does.
I don't think the previous maintainer wrote all those lines with all the necessary steps for fun, and neither did the crowdsec devs. Check the wizard.sh file yourself and you'll find yourself having to agree with me; when you do, please restore the install scripts so that the program will be able to run again, it'll be as simple as rolling back most of your changes to previous commits and doing very minimal tweaks.
If Arch's packaging guidelines discourage using /usr/local then the files should be moved somewhere else, not removed entirely.

mhdi commented on 2024-04-08 11:32 (UTC) (edited on 2024-04-08 11:50 (UTC) by mhdi)

@Exentio Thank you for your comment. The previous version of the PKGBUILD incorrectly installed files to /usr/local, which is not in accordance with Arch packaging guidelines. In the new PKGBUILD, I've followed the official Debian and Fedora packaging methods, which can be found in the Crowdsec GitHub repository.

Regarding the wizard.sh script you mentioned, it was necessary in the previous PKGBUILD but is not required in the new one. Everything has been integrated directly into the PKGBUILD.