Hi

Am 10/09/2017 um 15:24 schrieb Johannes Löthberg via aur-general:

Quoting Alad Wenter via aur-general (2017-09-10 15:16:57)

...

Hello everyone,

My name is Alad Wenter, and I would like to apply for a position as Trusted User. I’m a student in Germany, majoring in Mathematics with a focus on Algebraic Topology. Many thanks to Johannes Löthberg who is sponsoring my TU application.

I’ve been using Linux since late 2013, as the EOL of Windows XP was approaching. My first steps were taken via Linux Mint, gradually moving to Debian in both Stable and Unstable variants.

I have started using Arch in middle 2014, after being convinced that “I should just install Arch” on a different forum. Since then I have become an ArchWiki Bureaucrat (since May 2016) and IRC Operator (since February 2017) for the #archlinux channel on freenode.

Amongst others, I helped complete the long-standing merge of the Beginners’ Guide [1], vastly extended the comparison of third-party AUR scripts on the “AUR Helper” article, [2] and improved coordination amongst users through a new #archlinux-wiki freenode channel. [3] Other activities include the Arch forums [4], Bug tracker [5], and submitting requests on the AUR.

[1] https://lists.archlinux.org/pipermail/arch-dev-public/2016-July/028140.html [2] https://wiki.archlinux.org/index.php/AUR_helpers#Comparison_table [3] https://wiki.archlinux.org/index.php/ArchWiki:IRC [4] https://bbs.archlinux.org/profile.php?id=81752 [5] https://bugs.archlinux.org/user/18684

For software development, I am familiar with the Bash and C++ languages, and have published a few projects related to Arch, such as aurutils. [6] Where possible, I try to coordinate with upstream authors if I encounter any issues with their projects (for example, [7])

[6] https://github.com/AladW/ [7] https://github.com/i3/i3/issues/2597

If I were to become a Trusted User, I would like to move the following AUR packages to community:

• datamash
• dmenu-extended
• eid-mw
• innoextract
• mit-scheme
• physlock
• polkit-explorer
• python-i3
• qpdfview
• wimlib

I have opened a repository with modifications from the original AUR packages to better suit them as candidates for [community]. [8]

[8] https://github.com/AladW/community

I would further adopt the following orphans in community:

• cabextract
• gpick
• sigil
• sshfs
• udevil
• usbview

Cheers,

Alad

I confirm my sponsorship and such. Let the discussion period begin!

Quoting Johannes Löthberg (2017-09-16 10:57:44)

Quoting Johannes Löthberg (2017-09-10 15:24:15)

...

I confirm my sponsorship and such. Let the discussion period begin!

Discussion period has ended and the vote is up!

https://aur.archlinux.org/tu/?id=96

Vote has closed, and you've been accepted into our ranks!

Yes: 35 No: 4 Abstain: 4 Total: 43

Am 10/09/2017 um 15:25 schrieb Alad Wenter via aur-general:

Sorry, there was a fluke with enigmail that prevented a corrected signature. This one should however work.

Alad

The issue was that I signed both the message and attachment (my public key) in enigmail, and the attachment was then removed by mailman. As such, the signature became invalid. Thanks to Florian to help figure this out.

Alad

On Sun, Sep 10, 2017 at 04:55:16PM +0000, Eli Schwartz wrote:

On 09/10/2017 09:16 AM, Alad Wenter via aur-general wrote:

...

Hello everyone,

My name is Alad Wenter, and I would like to apply for a position as Trusted User. I’m a student in Germany, majoring in Mathematics with a focus on Algebraic Topology. Many thanks to Johannes Löthberg who is sponsoring my TU application.

Somme comments on your AUR packages:

aurutils/aurutils-git:

• uses the ISC ("custom:ISC") license, which is not a common license (`pacman -Ql licenses`)and must therefore be installed separately.
• Thanks you kindly for upgrading the security of the *sums you SKIP from md5 to sha256 :p :p

Fixed the license array. The license was already installed by the Makefile.

bash-devel-git:

• You do not follow our repo bash package in defining, as per FS#50567, -DNON_INTERACTIVE_LOGIN_SHELLS
• Apparently this doesn't work with the system readline or something???
• system.bashrc is out of sync with the core/bash package
• Do you even pacman hooks? Get rid of bash.install

Fixed.

cottage:

• source tarball violates shared SRCDEST, v$pkgver.tar.gz clashes with other packages that have the same version.

Fixed.

cottage-git:

• Has an abomination of a pkgver(), you know how to do this properly -_- and upstream even has tags!!!
• clones #branch=master which is the default
• unconditionally defines options=('debug'), which is meant to be a makepkg.conf choice and uses a separate build target which just appends to CFLAGS the same way our DEBUG_CFLAGS do.
• again consider asking upstream to support DESTDIR.

Fixed. I agree about DESTDIR and filed a pull request upstream.

https://github.com/HarveyHunt/cottage/pull/9

datamash-git:

• pkgver() doesn't strip "v" from the start
• source is cloned over git://git.sv.gnu.org which redirects to git.savannah.gnu.org, also please clone over https:// as this verifies the server in addition to just encrypting the transport
• It seems bizarre that this makedepends on rsync and wget, the latter especially seems like some part of the build process attempts to download itself... I think it wants to bootstrap updated translations but this should still be done via source=().
• Do you even pacman hooks? Get rid of datamash.install

Fixed. The rsync usage for translation files seems hard-coded in the bootstrap script, so this is something to mention to upstream.

dpkg:

• git source at pinned commit should not re-clone itself to a new $pkgname-$pkgver every time you bump the pkgver
• autoreconf should be done in prepare()
• explicitly override options=(emptydirs) rather than depending on the user's choices in makepkg.conf

Fixed.

howm-x11:

• source tarball violates shared SRCDEST, v$pkgver.tar.gz clashes with other packages that have the same version.
• Makefile has install target.
• Those find commands are awkward, `mkdir -p ... && cp -a ...` is not a sin so please use it.
• So much whitespace separating variables...
• That install script depresses me, shouldn't users be expected to determine for themselves how to use this?

Fixed. The Makefile missed the -D argument to install, so I've filed a pull request upstream.

https://github.com/HarveyHunt/howm/pull/67

howm-x11-git:

• Basically just the same things mentioned for howm-x11 or cottage-git.

Fixed apart from the Makefile, which actually does not include the correct install target for debug builds...

https://github.com/HarveyHunt/howm/issues/68

jshon-git:

• pkgver should use tags via `git describe --tags`

Fixed.

kkrieger:

• Downloads an unversioned kkrieger-beta.zip -- I understand this is web.archive.org that will never be updated but it is still painful. :D
• kkrieger.jpg is a 404 not found, also you never even try to do anything with it...

Fixed.

lightspark-git:

• Do you even pacman hooks? Get rid of lightspark-git.install
• Depends on desktop-file-utils, not needed because of hooks, and curl, arguably not needed as it's a dependency of pacman among other things.

Fixed.

mc-git:

• autogen should really be done in prepare()

Fixed.

nvtv:

• Empty variables should be deleted.
• md5sums at the bottom of the PKGBUILD are weird, move this up with the other variables, like source=()
• You really need quilt for this?
• make prefix="$pkgdir/usr" install seems like it should be ./configure --prefix=/usr && make DESTDIR since this is after all autotools and if there is one decent thing about autotools it is the fact that DESTDIR can be essentially guaranteed to exist...

Fixed, though for now still using quilt. Since it's a debian patch package, using quilt seemed the correct tool and didn't require me to use an own ad-hoc method.

repoctl-git:

• pkgver() does not strip leading "v"
• Why does this depend on xz, which is a dependency of lots of core things e.g. libarchive, and unlikely to be explicitly needed vs. the many other compression formats makepkg/repo-add supports?
• Why !strip, does this package actually break if you try stripping it?

There's several upstream files that depend explicitely on xz:

https://github.com/cassava/repoctl/blob/1a140a3f928bfc69a2ec2e3a4bc9e033bee7...

It's however true that the dependency is already provided by pacman, so I've removed it. Also removed !strip and fixed pkgver.

thunar-git:

• autogen should be done in prepare()

Fixed.

vim-bracketed-paste:

• The source is a github master.zip, I'm not even sure how to properly express my disappointment. It should be a -git package since there are no releases.

Fixed to use pkgver. I've however requested to delete the package, due to related commits vim upstream and frequent issues with tmux.

xss:

• So much whitespace in the variables...
• Unversioned source xss-master.tar.gz
• None of it works because the upstream website is dead, everything redirects to https://sites.google.com/view/woozle/
• Author still exists at https://github.com/nealey, project has moved to https://github.com/9wm/xss (he is a member of that org)

Fixed. Regarding the source, I've asked for a relase upstream:

https://github.com/9wm/xss/issues/1

Some improvements regarding the Makefile should also be made.

...

I have opened a repository with modifications from the original AUR packages to better suit them as candidates for [community]. [8]

[8] https://github.com/AladW/community

physlock:

• source tarball violates shared SRCDEST, v$pkgver.tar.gz clashes with other packages that have the same version.

Fixed.

polkit-explorer:

• Since I see you already filed one pull request upstream for PEP 394, why not also contribute a decent setup.py and that desktop file...
• uses the ISC ("custom:ISC") license, which is not a common license (`pacman -Ql licenses`)and must therefore be installed separately.

Fixed the license array; the license file itself was already installed to /usr/share/licenses/polkit-explorer. I agree on contributing a setup.py.

python-i3-py:

• It is probably not important to point out in the description which programming language it uses, especially when the pkgname already includes that info.
• git source at pinned commit should not re-clone itself to a new $pkgname-$pkgver every time you bump the pkgver
• Python packages which are intended to install a command-line tool rather than a library should not be prefixed with python- and do not need to be installed for both Python 3 and Python 2.

Fixed the description and the git source. Regarding the command-line tool, the package is meant as a library, but the repository ships some examples which can be used as command-line tools. I've moved them to a separate "-examples" split package.

qpdfview:

• Does this really need desktop-file-utils and hicolor-icon-theme or was this a remnant of some pre-hooks install script?q

Fixed.

-- Eli Schwartz

Thank you very much for your elaborate review!

Alad

Am 10/09/2017 um 23:56 schrieb Eli Schwartz:

On 09/10/2017 04:08 PM, Alad Wenter via aur-general wrote:

...

...

aurutils/aurutils-git:

• uses the ISC ("custom:ISC") license, which is not a common license (`pacman -Ql licenses`)and must therefore be installed separately.
• Thanks you kindly for upgrading the security of the *sums you SKIP from md5 to sha256 :p :p

Fixed the license array. The license was already installed by the Makefile.

o_O such upstreams!

...

...

bash-devel-git: mc-git: nvtv: repoctl-git: vim-bracketed-paste:

Did you forget to push it?

Indeed, I did... fixed. Thanks for checking.

...

Fixed. I agree about DESTDIR and filed a pull request upstream.

https://github.com/HarveyHunt/cottage/pull/9

FWIW, usually this gets implemented as install: install -Dm755 progname $(DESTDIR)$(BINDIR)/progname

DESTDIR is limited to use in install since sometimes you will want to compile the location of e.g. $(PREFIX)/share/progname/ inside a binary and why move only some instances of $(DESTDIR) up top I guess.

Updated the pull request.

...

...

dpkg:

• git source at pinned commit should not re-clone itself to a new $pkgname-$pkgver every time you bump the pkgver
• autoreconf should be done in prepare()
• explicitly override options=(emptydirs) rather than depending on the user's choices in makepkg.conf

Fixed.

Broken, actually. options=(emptydirs) means keep empty dirs, and options=(!emptydirs) means don't keep them (delete them). The default is emptydirs, leaving them alone, but if someone modified their makepkg.conf to specify !emptydirs then the directory you created would be deleted during tidy_install. And now it will be deleted no matter what.

Not sure how I got confused there, especially when the initial commit I made had it right...

https://aur.archlinux.org/cgit/aur.git/commit/?h=dpkg&id=067d5267d7bd44b...

...

...

xss:

• So much whitespace in the variables...
• Unversioned source xss-master.tar.gz
• None of it works because the upstream website is dead, everything redirects to https://sites.google.com/view/woozle/
• Author still exists at https://github.com/nealey, project has moved to https://github.com/9wm/xss (he is a member of that org)

Fixed. Regarding the source, I've asked for a relase upstream:

https://github.com/9wm/xss/issues/1

Regarding the source, nothing says you cannot use "$_commit" in place of "$pkgver" and "master", at least until you have a release tag to use instead.

Updated, both for xss and vim-bracketed-paste.

...

Thank you very much for your elaborate review!

Happy to help, reviewing PKGBUILDs is always fun whether as part of a TU application or not. Cheers,

Alad

On Thu, Sep 14, 2017 at 01:01:58PM +0000, Evgeniy Alekseev wrote:

Hai,

On Sunday, 10 September 2017 19:55:16 MSK Eli Schwartz wrote:

...

python-i3-py:

• It is probably not important to point out in the description which programming language it uses, especially when the pkgname already includes that info.
• git source at pinned commit should not re-clone itself to a new $pkgname-$pkgver every time you bump the pkgver
• Python packages which are intended to install a command-line tool rather than a library should not be prefixed with python- and do not need to be installed for both Python 3 and Python 2.

BTW, it looks like this PKGBUILD works normally for this package, but in general calling setup.py with python3 interpreter assuming python2 build [1] is bad idea. Also it looks like the package has several licenses, e.g. winmenu.py header says WTFPL-v2 :)

Thanks, I've updated the build step and LICENSES. [2]

[2] https://github.com/AladW/community/commit/a48ba9a3ff917b596d89748a3a69e4f39e...

Note that I installed the license to /usr/share/licenses/python-i3-py, as python-i3-py is the common pkgbase. The PKGBUILD guidelines [3] however mention to use pkgname - and namcap issues a warning accordingly. Do these guidelines account for split packages?

python-i3-py-examples E: Missing custom license directory (usr/share/licenses/python-i3-py-examples)

[3] https://wiki.archlinux.org/index.php/PKGBUILD#license

1. https://github.com/AladW/community/blob/master/python-i3-py/PKGBUILD#L28.

-- Sincerely yours, Evgeniy Alekseev

Cheers,

Alad

On Thu, Sep 14, 2017 at 01:51:19PM +0000, Levente Polyak wrote:

On 09/14/2017 03:43 PM, Alad Wenter via aur-general wrote:

...

On Thu, Sep 14, 2017 at 01:01:58PM +0000, Evgeniy Alekseev wrote:

...

Hai,

On Sunday, 10 September 2017 19:55:16 MSK Eli Schwartz wrote:

...

python-i3-py:

• It is probably not important to point out in the description which programming language it uses, especially when the pkgname already includes that info.
• git source at pinned commit should not re-clone itself to a new $pkgname-$pkgver every time you bump the pkgver
• Python packages which are intended to install a command-line tool rather than a library should not be prefixed with python- and do not need to be installed for both Python 3 and Python 2.

BTW, it looks like this PKGBUILD works normally for this package, but in general calling setup.py with python3 interpreter assuming python2 build [1] is bad idea. Also it looks like the package has several licenses, e.g. winmenu.py header says WTFPL-v2 :)

Thanks, I've updated the build step and LICENSES. [2]

[2] https://github.com/AladW/community/commit/a48ba9a3ff917b596d89748a3a69e4f39e...

Note that I installed the license to /usr/share/licenses/python-i3-py, as python-i3-py is the common pkgbase. The PKGBUILD guidelines [3] however mention to use pkgname - and namcap issues a warning accordingly. Do these guidelines account for split packages?

python-i3-py-examples E: Missing custom license directory (usr/share/licenses/python-i3-py-examples)

If a particular package has some special license, it should indeed install those into the license directory for that particular pkgname instead of pkgbase. Note that using $pkgbase could potentially also yield into a scenario where py2 and py3 variant can't be installed at the same time as the license file would conflict.

All right, I've updated the PKGBUILD accordingly.

cheers, Levente

Cheers,

Alad

On Thu, Sep 14, 2017 at 01:36:20PM +0000, Levente Polyak wrote:

On 09/10/2017 03:16 PM, Alad Wenter via aur-general wrote:

...

Hello everyone,

My name is Alad Wenter, and I would like to apply for a position as Trusted User. I’m a student in Germany, majoring in Mathematics with a focus on Algebraic Topology. Many thanks to Johannes Löthberg who is sponsoring my TU application.

Hi Alad,

some technical feedback that I found after the current reworks mentioned here:

cottage:

• it ignores LDFLAGS and just uses CFLAGS wich results in a binary without our flags (like -znow)

Proposed upstream: [1]

[1] https://github.com/HarveyHunt/cottage/pull/11

cottage-git:

• same as cottage, LDFLAGS
• should provides and conflicts cottage as its a VCS pkgname

Fixed.

dpkg:

• would me more error prone and convenient to keep pkgver in sync when using a pkgver() function for pinned commits and f.e. do: git describe --always | sed 's/^v//;s/-/./g'

Fixed.

kkrieger:

• cd "$srcdir" doesn't do anything at all

Fixed.

nvtv:

• static mention of 0.4.7 version label in source array looks ugly and is less error prone when bumping a pkgver=

Fixed.

pkgelf-git:

• should provides and conflicts pkgelf as its a VCS pkgname

Fixed.

xss:

• packages doesn't even build line 18: cd: xss-master: No such file or directory and after that install: cannot stat 'README.md': No such file or directory did you test this at all?
• would me more error prone and convenient to keep pkgver in sync when using a pkgver() function for pinned commits and f.e. do: git describe --always | sed 's/^v//;s/-/./g' or just count the commits as prefix to pkv
• CFLAGS is completely purged and overwritten by the Makefile, it should be passed to be preserved or the Makefile be patched

Fixed. [2] I did test it, but forgot to remove the xss-master directory from the previous source (master.zip) before doing so...

[2] https://github.com/9wm/xss/pull/2#event-1249347915

cheers Levente

Cheers,

Alad