* Doug McMillan (Sat, 13 Dec 2014 23:32:05 -0600):

To have a safer more reliable place I made a sourceforge project. The xen files are herehttps://sourceforge.net/projects/xenforarch/files/build/xen/ The package I am asking for the critique on ishttps://sourceforge.net/projects/xenforarch/files/build/xen/xen-git-4.5_rc3....

First of all, you should name this package xen-git, as you're using the most recent git version of that package. Second, this is an old way to pull git sources (i.e., manually). Since pacman 4.1 you can let the package builder pull git sources. See the wiki [1] and an example PKGBUILD [2]. Please modify your PKGBUILD accordingly. Third, don't use md5sums to check file integrity; to avoid collisions, it is recommended that you use sha256sums. You can set this in /etc/makepkg.conf (and then you can use updpkgsums to generate them). You should enclose every instance of $pkgdir within quotes, to allow for spaces in directory names. Minor detail, but why are you cd'ing to "$srcdir/$_gitname" and then cd'ing to ../../? That's superfluous. Just cd to "$srcdir" and you'll be able to install files included in the source array. The same is true for that cd at the end of 'make install' block. Please rework the PKGBUILD and post it at Github gist or Pastebin, so it's easy for all of us to have a look at it. Just another note: please don't top-post *and* remove lines you're not referring to. And you know you should create source tarballs with mkaurball[3] nowadays, don't you? Otherwise you won't be able to upload your package to the AUR. Regards, Marcel [1] https://wiki.archlinux.org/index.php/VCS_PKGBUILD_Guidelines [2] https://aur.archlinux.org/packages/gt/gtk-engine-murrine-git/PKGBUILD [3] this will be changed in the (near?) future, when pacman supports the generation of .SRCINFO files

* Robert Mackanics [2014-12-16 03:42:51 -0500]:

On Monday, December 15, 2014 16:54:04 Marcel Korpel wrote:

...

Third, don't use md5sums to check file integrity; to avoid collisions, it is recommended that you use sha256sums. You can set this in /etc/makepkg.conf (and then you can use updpkgsums to generate them).

Should we have the makepkg.conf in the pacman package changed to sha256? Seems like a good idea that shouldn't bite anybody.

I submitted a patch and it was declined: https://lists.archlinux.org/pipermail/pacman-dev/2014-June/019081.html https://lists.archlinux.org/pipermail/pacman-dev/2014-June/019083.html https://lists.archlinux.org/pipermail/pacman-dev/2014-June/019084.html Florian -- http://www.the-compiler.org | me@the-compiler.org (Mail/XMPP) GPG: 916E B0C8 FD55 A072 | http://the-compiler.org/pubkey.asc I love long mails! | http://email.is-not-s.ms/

* Pablo Lezaeta Reyes [2014-12-16 13:50:58 -0300]:

2014-12-16 6:05 GMT-03:00 Florian Bruhin :

...

...

On Monday, December 15, 2014 16:54:04 Marcel Korpel wrote:

...

Third, don't use md5sums to check file integrity; to avoid collisions, it is recommended that you use sha256sums. You can set this in /etc/makepkg.conf (and then you can use updpkgsums to generate

* Robert Mackanics [2014-12-16 03:42:51 -0500]: them).

...

Should we have the makepkg.conf in the pacman package changed to sha256?

Seems

...

like a good idea that shouldn't bite anybody.

I submitted a patch and it was declined:

https://lists.archlinux.org/pipermail/pacman-dev/2014-June/019081.html https://lists.archlinux.org/pipermail/pacman-dev/2014-June/019083.html https://lists.archlinux.org/pipermail/pacman-dev/2014-June/019084.html

Florian

-- http://www.the-compiler.org | me@the-compiler.org (Mail/XMPP) GPG: 916E B0C8 FD55 A072 | http://the-compiler.org/pubkey.asc I love long mails! | http://email.is-not-s.ms/

As far I underestand it was because "md5 is the most used", so I now add a commented line in my makepkgs that explain that "I use md5 cause pacman -g give that".

Just set INTEGRITY_CHECK=(sha256) in your /etc/makepkg.conf and makepkg -g (not pacman -g, by the way) will give you that. I still think that should be the default, but whatever.

I thing maybe if a big number of user submit the patch ask in forums and/or add the bug (carelessly if is regected) that could bring the questioning if is a good idea use md5

Doesn't sound like a constructive way to approach this in my opinion. Florian -- http://www.the-compiler.org | me@the-compiler.org (Mail/XMPP) GPG: 916E B0C8 FD55 A072 | http://the-compiler.org/pubkey.asc I love long mails! | http://email.is-not-s.ms/

On Tuesday, December 16, 2014 at 06:54:34 PM, Doug McMillan wrote:

...

Date: Mon, 15 Dec 2014 16:54:04 +0100 From: marcel.korpel@gmail.com To: aur-general@archlinux.org Subject: Re: [aur-general] requesting a critique for my build of xen-git-4.5.0-rc3 Please rework the PKGBUILD and post it at Github gist or Pastebin, so it's easy for all of us to have a look at it.

Regards, Marcel

[1] https://wiki.archlinux.org/index.php/VCS_PKGBUILD_Guidelines [2] https://aur.archlinux.org/packages/gt/gtk-engine-murrine-git/PKGBUILD [3] this will be changed in the (near?) future, when pacman supports the generation of .SRCINFO files

Thank you for the critique. I have fixed the name, changed from manual to automated git, switched frommd5 to sha256, enclosed all variable containing directories to be in " ", cleaned up the 2 unnecessary directory changes. I have posted the updated PKGBUILD at http://pastebin.com/j0MTRciV I have updated the sourceforge files also. There I have created the source package with mkaurball.

Hi Doug, a few other nitpicky comments: - doing `unset CFLAGS` *after* make looks pointless, especially given that you also have !buildflags in your options=() - "xen::git+git://" looks redundant, just "git://" will suffice (because the name of the repository is already "xen") - messages with lots of "====" seem awkward -- makepkg already highlights messages; you may try to use "msg2" instead of "msg" to have them in different color - are you sure that exports from package() have effect on build()? I'd rather create a separate function and call it from build() and maybe package() Otherwise, the PKGBUILD is pretty good AFAICS. Cheers, -- Ivan Shapovalov / intelfx /

Subject: Re: [aur-general] requesting a critique for my build of xen-git-4.5.0-rc3> - doing `unset CFLAGS` *after* make looks pointless> - "xen::git+git://" looks redundant> - messages with lots of "====" seem awkward > - are you sure that exports from package() have effect on build()?> Ivan Shapovalov / intelfx / From: marcel.korpel@gmail.com> One more thing: you are dlm1065@etcetera, but the maintainer is listed> as Triton (jacek@etcetera). You should me him/her a contributor and> list yourself as maintainer.> > And about what Ivan said: he probably meant exports in prepare(), that> might not be available during build(); even more, I can use makepkg> --nobuild to only get the sources and run prepare() and at a later time> run makepkg --noextract to run build() etc.> Regards,> Marcel Ivan Removed the unset as suggested, removed the git+, altered msg to msg2 (<sigh> had that in there for debugging first few runs had so many errors needed an easy way to break were what error went at a glance). Marcel Changed my name to maintainer. Ivan, Marcel thank you thank you.. for having me look at the exports I goofed here big time I originally had set these all up in my .profile and never removed that(done now). I am not sure if prepare would of passed to the other sections or not but if done with the --noextract it would of definitely broke the build. Custom function in and working