Hi,
I do have the impression that the number of aur comments about pacakges not building unless foo is installed has gone up. Very often foo turns out to be part of base-devel or multilib-devel.
Answering that is easy, by linking to main AUR wiki page that clearly states base-devel is assumed to be installed when building aur packages. That part appears to be overlooked often though and results in unnecessary clutter in aur comments.
Perhaps we could put base-devel group in makedepends of aur pacakages ? Lib32-* / multilib packages could list both multilib-devel and base-devel.
Lone_Wolf
On Sat, Sep 16, 2017 at 02:19:09PM +0200, LoneVVolf wrote:
Hi,
I do have the impression that the number of aur comments about pacakges not building unless foo is installed has gone up. Very often foo turns out to be part of base-devel or multilib-devel.
Answering that is easy, by linking to main AUR wiki page that clearly states base-devel is assumed to be installed when building aur packages. That part appears to be overlooked often though and results in unnecessary clutter in aur comments.
Perhaps we could put base-devel group in makedepends of aur pacakages ? Lib32-* / multilib packages could list both multilib-devel and base-devel.
Lone_Wolf
I can see where you're coming from.
I have to be stubborn. Arch assumes some sort of sensible working brain of its users. Users building software that haven't installed things for building software… it makes me concerned for the community.
David
On 09/16/2017 07:52 PM, David Phillips wrote:
On Sat, Sep 16, 2017 at 02:19:09PM +0200, LoneVVolf wrote:
Hi,
I do have the impression that the number of aur comments about pacakges not building unless foo is installed has gone up. Very often foo turns out to be part of base-devel or multilib-devel.
Answering that is easy, by linking to main AUR wiki page that clearly states base-devel is assumed to be installed when building aur packages. That part appears to be overlooked often though and results in unnecessary clutter in aur comments.
Perhaps we could put base-devel group in makedepends of aur pacakages ? Lib32-* / multilib packages could list both multilib-devel and base-devel.
Lone_Wolf
I can see where you're coming from.
I have to be stubborn. Arch assumes some sort of sensible working brain of its users. Users building software that haven't installed things for building software… it makes me concerned for the community.
David
At the risk of coming off as a little BOFH-y, I've been in this situation before (where some users CBA to read[0]) and it makes me want to write something that will:
1.) check the comments sections for new comments on my packages, 2.) parse those comments for pasted error messages that would be caused by not-installed (base|multilib)-devel 3.) follow to the comment author's AUR profile, and grab contact info from their email address/PGP key FPR/IRC nick/whichever it finds, and 4.) emails/PMs them a link to [0].
Now, I don't see a Terms of Service for the AUR, but if there's demand for this from other AUR package maintainers and nobody sees any obvious ethical flaws with this, I'd be glad to make this my next project and share with other maintainers.
We can automate everything else, why shouldn't we automate RTFMs?
[0] https://wiki.archlinux.org/index.php/Makepkg#Usage
On Sat, 16 Sep 2017 20:19:56 -0400, brent s. wrote:
We can automate everything else, why shouldn't we automate RTFMs?
+1 but the robot should not only respond to stupid users, but also react on idiotic maintainers after checking PKGBUILDs and scripts of the tarballs.
All jesting aside, pin a comment, if there should be a highly recurrent issue for some users.
On 09/16/2017 08:36 PM, Ralf Mardorf wrote:
On Sat, 16 Sep 2017 20:19:56 -0400, brent s. wrote:
We can automate everything else, why shouldn't we automate RTFMs?
+1 but the robot should not only respond to stupid users, but also react on idiotic maintainers after checking PKGBUILDs and scripts of the tarballs.
well, i'm not too worried about that because I don't get emailed every time a maintainer uploads a bad PKGBUILD, but I do every time someone comments on one of my packages with a "broken pls fix: "make: command not found".
;)
All jesting aside, pin a comment, if there should be a highly recurrent issue for some users.
and yet even then, they still somehow manage to not read it, which is why I'm tempted by a more proactive (or rather, proactively reactive) and direct approach.
On Sat, 16 Sep 2017 20:49:42 -0400, brent s. wrote:
On 09/16/2017 08:36 PM, Ralf Mardorf wrote: well, i'm not too worried about that because I don't get emailed every time a maintainer uploads a bad PKGBUILD
Hi,
as a user I always need to check PKGBUILDs and other content of the tarballs. Sometimes the content is terrible and it's time wasting even if I should ignore the issues completely instead of fixing them and writing the maintainer.
On Sat, 16 Sep 2017 20:19:56 -0400, brent s. wrote: All jesting aside, pin a comment, if there should be a highly recurrent issue for some users.
and yet even then, they still somehow manage to not read it, which is why I'm tempted by a more proactive (or rather, proactively reactive) and direct approach.
The community could help the AUR maintainers.
I don't maintain a package, but e.g. a few days ago a maintainer of a package pinned one of my explanations. Some users didn't follow the latest news regarding the Perl library path change and I explained that they should follow the advices of the latest news and when rebuilding packages using an AUR helper, to ensure that the build dependency chain is followed in the correct order.
Apart from not following the latest news/announcements the most recurring issue is, that users don't import public keys, while the output they post by a comment says that they need to import the public key. Another user could help the maintainer by following the comments and paste how to import a key from the clipboard. I'm doing this all the times.
The issue you mentioned isn't a common issue.
Résumé:
You are not forced to maintain AUR packages.
I'm not forced to use any tarball from AUR at all.
We as a community could help each other. Once we explained how to solve the issue to somebody who didn't read official announcements, FAQ, pinned comments and even does not copy the output to the Google search widget, before coping it to the AUR comment widget, this persons might change this behaviour and will help other inexperienced users, too.
If we are patient we waste less time, than with writing hate mails about "help vampires", "rude jerks" and "clueless morons" and even changing the policy or using a robot for auto-responses won't change the real world. We always need to temper the wind to the shorn lamb, since there always will be a shorn lamb, whatever we do. Being patient does cost us less energy, than being annoyed.
Regards, Ralf
On 09/16/17 at 08:19pm, brent s. wrote:
On 09/16/2017 07:52 PM, David Phillips wrote:
On Sat, Sep 16, 2017 at 02:19:09PM +0200, LoneVVolf wrote:
Hi,
I do have the impression that the number of aur comments about pacakges not building unless foo is installed has gone up. Very often foo turns out to be part of base-devel or multilib-devel.
Answering that is easy, by linking to main AUR wiki page that clearly states base-devel is assumed to be installed when building aur packages. That part appears to be overlooked often though and results in unnecessary clutter in aur comments.
Perhaps we could put base-devel group in makedepends of aur pacakages ? Lib32-* / multilib packages could list both multilib-devel and base-devel.
Lone_Wolf
I can see where you're coming from.
I have to be stubborn. Arch assumes some sort of sensible working brain of its users. Users building software that haven't installed things for building software… it makes me concerned for the community.
David
At the risk of coming off as a little BOFH-y, I've been in this situation before (where some users CBA to read[0]) and it makes me want to write something that will:
1.) check the comments sections for new comments on my packages, 2.) parse those comments for pasted error messages that would be caused by not-installed (base|multilib)-devel 3.) follow to the comment author's AUR profile, and grab contact info from their email address/PGP key FPR/IRC nick/whichever it finds, and 4.) emails/PMs them a link to [0].
Now, I don't see a Terms of Service for the AUR, but if there's demand for this from other AUR package maintainers and nobody sees any obvious ethical flaws with this, I'd be glad to make this my next project and share with other maintainers.
We can automate everything else, why shouldn't we automate RTFMs?
I sincerely don't believe that this will have the right affect, it will be treated as spam.
Maybe we should start advocating to use extra-$arch-build for building AUR packages, since that contains base-devel :-). This is probably also what Debian does I guess or does every package implictly depend on 'build-essential'.
Em setembro 17, 2017 5:57 Jelle van der Waa escreveu:
I sincerely don't believe that this will have the right affect, it will be treated as spam.
Maybe we should start advocating to use extra-$arch-build for building AUR packages, since that contains base-devel :-). This is probably also what Debian does I guess or does every package implictly depend on 'build-essential'.
+1 for using devtools and extra-x86_64-build for AUR packages too. I have been doing that for a while now, and it does catch a lot of issues.
Regards, Giancarlo Razzolini
On 09/18/2017 12:06 AM, Giancarlo Razzolini wrote:
Em setembro 17, 2017 5:57 Jelle van der Waa escreveu:
I sincerely don't believe that this will have the right affect, it will be treated as spam.
Maybe we should start advocating to use extra-$arch-build for building AUR packages, since that contains base-devel :-). This is probably also what Debian does I guess or does every package implictly depend on 'build-essential'.
+1 for using devtools and extra-x86_64-build for AUR packages too. I have been doing that for a while now, and it does catch a lot of issues.
Regards, Giancarlo Razzolini
I also agree on that. Building in a clean chroot has multiple advantages: * Mostly equal builds for everybody (until they are reproducible) * Less errors, as only, but all makedepends must be installed in the chroot * Less trashed archlinux installations because of makedependency installation * Devtools is even simpler to handle if you ask me
I also do not understand why AUR helpers dont use devtools. It would give much better package quality.
Maybe we should advice to use devtools more in the wiki page. Most people dont even know that is a single command to type.
Pro tip: Include other AUR pkg deps like this: extra-x86_64-build -- -I aurdep.pkg.tar.xz
~Nico
On Mon, Sep 18, 2017 at 02:01:18PM +0000, NicoHood wrote:
On 09/18/2017 12:06 AM, Giancarlo Razzolini wrote:
Em setembro 17, 2017 5:57 Jelle van der Waa escreveu:
I sincerely don't believe that this will have the right affect, it will be treated as spam.
Maybe we should start advocating to use extra-$arch-build for building AUR packages, since that contains base-devel :-). This is probably also what Debian does I guess or does every package implictly depend on 'build-essential'.
+1 for using devtools and extra-x86_64-build for AUR packages too. I have been doing that for a while now, and it does catch a lot of issues.
Regards, Giancarlo Razzolini
I also agree on that. Building in a clean chroot has multiple advantages:
- Mostly equal builds for everybody (until they are reproducible)
- Less errors, as only, but all makedepends must be installed in the chroot
- Less trashed archlinux installations because of makedependency
installation
- Devtools is even simpler to handle if you ask me
I also do not understand why AUR helpers dont use devtools. It would give much better package quality.
Mine does... </shameless_plug>
https://github.com/AladW/aurutils/blob/master/bin/aurbuild_chroot
Maybe we should advice to use devtools more in the wiki page. Most people dont even know that is a single command to type.
Pro tip: Include other AUR pkg deps like this: extra-x86_64-build -- -I aurdep.pkg.tar.xz
~Nico
On 09/18/2017 02:59 PM, Alad Wenter via aur-general wrote:
On Mon, Sep 18, 2017 at 02:01:18PM +0000, NicoHood wrote:
On 09/18/2017 12:06 AM, Giancarlo Razzolini wrote:
Em setembro 17, 2017 5:57 Jelle van der Waa escreveu:
I sincerely don't believe that this will have the right affect, it will be treated as spam.
Maybe we should start advocating to use extra-$arch-build for building AUR packages, since that contains base-devel :-). This is probably also what Debian does I guess or does every package implictly depend on 'build-essential'.
+1 for using devtools and extra-x86_64-build for AUR packages too. I have been doing that for a while now, and it does catch a lot of issues.
Regards, Giancarlo Razzolini
I also agree on that. Building in a clean chroot has multiple advantages:
- Mostly equal builds for everybody (until they are reproducible)
- Less errors, as only, but all makedepends must be installed in the chroot
- Less trashed archlinux installations because of makedependency
installation
- Devtools is even simpler to handle if you ask me
I also do not understand why AUR helpers dont use devtools. It would give much better package quality.
Mine does... </shameless_plug>
So does yaourt, simply append MAKEPKG="extra-i686-build -- " to /etc/yaourtrc and it will use that as the makepkg command for building packages.
pacaur, however, does not support custom makepkg/pacman commands. :( It's probably useless bloat and feature-creep or something.
On 09/16/2017 08:19 AM, LoneVVolf wrote:
Hi,
I do have the impression that the number of aur comments about pacakges not building unless foo is installed has gone up. Very often foo turns out to be part of base-devel or multilib-devel.
Answering that is easy, by linking to main AUR wiki page that clearly states base-devel is assumed to be installed when building aur packages. That part appears to be overlooked often though and results in unnecessary clutter in aur comments.
Perhaps we could put base-devel group in makedepends of aur pacakages ? Lib32-* / multilib packages could list both multilib-devel and base-devel.
I'm not entirely sure why you think this suggestion will ever be accepted. One of the defining elements of Arch Linux is the belief in RTFM, and especially the belief that people who build AUR packages without first installing multilib-devel and/or base-devel are a pox on the community and obvious help vampires.
By definition the only thing the Arch Linux community as a consensus will be okay with as a solution, is a better way of ignoring such time-wasting rude jerks.
I guess I would be okay with adding those as runtime dependencies to the AUR helpers most commonly used by clueless morons.
aur-general@lists.archlinux.org