[aur-general] xorg packages

[Ronald van Haren]

now for real, cc-ing

On 10/2/09, Ronald van Haren <pressh at gmail.com> wrote:
> cc-ing to speedvin so he has a change to reply and take action.
>
> (speedvin if you reply please do so to the aur-general mailing list)
> (sorry for top posting)
>
> Ronald
>
>
> On 10/2/09, Xavier <shiningxc at gmail.com> wrote:
>> On Fri, Oct 2, 2009 at 1:33 PM, Ronald van Haren <pressh at gmail.com>
>> wrote:
>>>>
>>>>
>>>> Actually, I would not recommend using any of these packages because of
>>>> their poor quality :
>>>> http://aur.archlinux.org/packages.php?K=SpeedVin&SeB=m
>>>>
>>>
>>> I haven't really time to look at them. Are they just poor quality or do
>>> they
>>> possible cause system breakage ?
>>>
>>> btw, isn't that the guy who recently applied to become a TU?
>>>
>>
>> Oh right, I didn't realize that. Actually I already complained :
>> http://mailman.archlinux.org/pipermail/aur-general/2009-September/006559.html
>> I didn't realize how correct my comment was !
>>
>> About the other packages, I was not sure whether to mention them or
>> not. The easiest way would have been to stay silent. If I start giving
>> details, I should address them to the maintainer directly, in the AUR
>> comments. But explaining everything that is wrong (there are all
>> levels of problems : big ones, average, and minor details) will take
>> much more time.
>>
>> My biggest complaint was on the xorg-server-git one :
>> http://aur.archlinux.org/packages/xorg-server-git/xorg-server-git/PKGBUILD
>> extra package :
>> http://repos.archlinux.org/viewvc.cgi/xorg-server/repos/extra-i686/PKGBUILD?revision=51034&view=markup
>>
>> - it disables a lot of important features making it unusable (why
>> using xserver-git if you disable ipv6 dri dri2 xvfb composite)
>>
>> - it replaces one change of xorg-server pkgbuild for integration with
>> libgl
>> :
>>   # Needed for non-mesa drivers, libgl will restore it
>>   mv "${pkgdir}/usr/lib/xorg/modules/extensions/libglx.so" \
>>      "${pkgdir}/usr/lib/xorg/modules/extensions/libglx.xorg" || return 1
>> by a completely broken symlink :
>>   msg "Fixes for non-Mesa drivers..."
>>   ln -s "${pkgdir}/usr/lib/xorg/modules/extensions/libglx.so" \
>>      "${pkgdir}/usr/lib/xorg/modules/extensions/libglx.xorg" || return 1
>>
>> Not only doing this makes xorg-server-git fileconflicts with libgl. It
>> also creates a symlink which points to the $pkgdir where the package
>> was built...
>>
>> - it disables libwfb.so without any explanation (the old extra
>> pkgbuild did that too, saying it libwfb was no longer needed. but last
>> pkgbuild doesnt do it anymore. ok that one is a bit confusing)
>>
>> - it does not provide/conflict xorg-server
>>
>> - it cleans the build dir at the end of build() so you cannot check /
>> re-use anything
>>
>> - it defines unused variables (_gitver)
>>
>> - instead of depending on -git packages, it would be better to depend
>> on a specific version, which a real git package could provide. then it
>> would also be possible to use a normal (non git) package if it is
>> recent enough
>>
>> List goes on and on.
>>
>> Comments about other packages :
>> - glib2-newest also seems to be a duplicate of extra
>> - grub2-git has the date reversed ?
>> - kernel26-bfs really had 300 releases ?
>> - dwm-git ? why ? the official dwm repo is just hg, and actually even
>> the use of dwm-hg is debatable (see archives of last month)
>> - several PKGBUILDs have more than one empty md5sums array, for
>> example this one has 3 :
>> http://aur.archlinux.org/packages/xf86-input-mouse-git/xf86-input-mouse-git/PKGBUILD
>> - actually variables which are just empty array should not be declared at
>> all
>> - all git packages do something funny :
>>   if [ -d $startdir/src/$_gitname ] ; then
>>   cd $_gitname
>> why not if [ -d $_gitname ] , like the prototype does ? much less
>> confusing.
>> http://projects.archlinux.org/?p=abs.git;a=blob;f=prototypes/PKGBUILD-git.proto;hb=HEAD
>>
>> I am sure I would find much more to say by looking closer and going
>> more into smaller details.
>>
>