Hi all, As you may know, the NVIDIA drivers are split across a multitude of packages. While it is of no consequence when using the ones in the repo, it becomes a bit more of a problem when building packages from the AUR. I for example have to install nvidia-utils-beta, lib32-nvidia-utils-beta, nvidia-source-beta and nvidia-hook (mkinitcpio hook using dkms) to get all the required files. That means downloading the x86_64 blob 2 times and the i686 one 1 time (I'm using a dkms version to not download them one more time for each of my kernels). Moreover, all those packages have hard dependencies on version numbers, making AUR helpers quite helpless when there is an update (sure you could download all tarballs and have a centralized sources folder, build all packages, and then installing all at the same time, but keeping up to date would become very time consuming). I heard a nvidia-total-beta package was uploaded not long ago but swiftly deleted by gtmanfred who thought it was not needed. I don't mean to criticize, but I do think it would be of use to quite a lot of people. I'm using the catalyst-total on my laptop and I love it, having all in one place is good for the bandwidth and every update is just a breeze. Therefore I have made two new PKGBUILDs, nvidia-total [1] and nvidia-total-beta [2] which group the packages mentioned above for the regular and beta drivers, and are the NVIDIA counterparts to the catalyst-total package. I'm asking the TUs to reconsider their usefulness before I upload the to the AUR, only to see them deleted. Thx in advance for your answers. Regards, [1] http://pastebin.com/sKa4pDfp [2] http://pastebin.com/V5FSLmUi -- Alucryd "In the name of God, impure souls of the living dead shall be banished into eternal damnation. Amen."
Hi, According to the AUR guidelines, it should be okay with several packages for the same application, as long as they all provide something different, like applying different patches etc. "Exception to this strict rule may only be packages having extra features enabled and/or patches in compare to the official ones." - https://wiki.archlinux.org/index.php/Arch_Packaging_Standards#Submitting_pac... If the *-total packages provide the same software, without any extra features enabled and/or patches, perhaps they could be kept instead of the old packages? Are there reasons to keep the old nvidia packages, if they provide the same content, instead of only having the *-total packages? -- Sincerely, Alexander Rødseth xyproto / TU
hi before two days i ve upload to AUR this package: http://paste.ubuntu.com/1567724 (not the same, the name is nvidia-total-beta). and ipsofact deleted by @gtmanfred with this message: "this is completely un necessary, please don't litter the aur with more trash" well, i ve upload this package for the reason explain by @alucryd, need download up to 4 times the nvidia blob drivers for install dkms-nvidia-beta (was mine and left to use this package), nvidia-utils-beta, lib32-nvidia-utils-beta and lib32-libcl. everyone can use the package for your needs, that's what AUR, I think. the *-total* packages not add nothig in special like individual packages, only group in one some diferent packages for hel save several bandwidth/space and time i think can live with individual packages greetings pd: sorry my english 2013/1/25 Alexander Rødseth <rodseth@gmail.com>:
Hi,
According to the AUR guidelines, it should be okay with several packages for the same application, as long as they all provide something different, like applying different patches etc.
"Exception to this strict rule may only be packages having extra features enabled and/or patches in compare to the official ones." - https://wiki.archlinux.org/index.php/Arch_Packaging_Standards#Submitting_pac...
If the *-total packages provide the same software, without any extra features enabled and/or patches, perhaps they could be kept instead of the old packages? Are there reasons to keep the old nvidia packages, if they provide the same content, instead of only having the *-total packages?
-- Sincerely, Alexander Rødseth xyproto / TU
On Fri, Jan 25, 2013 at 9:19 AM, SpinFlo <sl1pkn07@gmail.com> wrote:
well, i ve upload this package for the reason explain by @alucryd, need download up to 4 times the nvidia blob drivers for install dkms-nvidia-beta (was mine and left to use this package), nvidia-utils-beta, lib32-nvidia-utils-beta and lib32-libcl.
Bottom-post please, like this. And why will symlinks not work for your purpose? Or is the purpose to be able to use yaourt or another AUR helper?
i ve use yaourt, yes, yaourt when make and install sucessful package delete all sources., then the symlink don't work greetings 2013/1/25 Oon-Ee Ng <ngoonee.talk@gmail.com>:
On Fri, Jan 25, 2013 at 9:19 AM, SpinFlo <sl1pkn07@gmail.com> wrote:
well, i ve upload this package for the reason explain by @alucryd, need download up to 4 times the nvidia blob drivers for install dkms-nvidia-beta (was mine and left to use this package), nvidia-utils-beta, lib32-nvidia-utils-beta and lib32-libcl.
Bottom-post please, like this.
And why will symlinks not work for your purpose? Or is the purpose to be able to use yaourt or another AUR helper?
sorry for the quotes after message, this is automatic by gmail :S greetings
On Fri, Jan 25, 2013 at 9:32 AM, SpinFlo <sl1pkn07@gmail.com> wrote:
i ve use yaourt, yes,
yaourt when make and install sucessful package delete all sources., then the symlink don't work
Don't use yaourt then. On Fri, Jan 25, 2013 at 9:37 AM, SpinFlo <sl1pkn07@gmail.com> wrote:
sorry for the quotes after message, this is automatic by gmail :S
And it's not 'automatic' by gmail, I use gmail, just click the 'in-line' button (3 dots). On Fri, Jan 25, 2013 at 11:07 AM, Xyne <xyne@archlinux.ca> wrote:
/etc/makepkg.conf -> SRCDEST
Set that to the directory of your choice and all source files should be saved there. Common sources will be automatically detected.
(yeah, via symlinks... but if this is the approach that you meant, it wasn't clear)
Yes SRCDEST is better but won't help for yaourt users. I can't remember when/why I started making my own symlinks, perhaps to prevent one directory being way too full (harder to spring clean). In any case, my apologies, SRCDEST is obviously the better solution here.
On Fri, Jan 25, 2013 at 5:10 AM, Oon-Ee Ng <ngoonee.talk@gmail.com> wrote:
On Fri, Jan 25, 2013 at 9:32 AM, SpinFlo <sl1pkn07@gmail.com> wrote:
i ve use yaourt, yes,
yaourt when make and install sucessful package delete all sources., then the symlink don't work
Don't use yaourt then.
On Fri, Jan 25, 2013 at 9:37 AM, SpinFlo <sl1pkn07@gmail.com> wrote:
sorry for the quotes after message, this is automatic by gmail :S
And it's not 'automatic' by gmail, I use gmail, just click the 'in-line' button (3 dots).
On Fri, Jan 25, 2013 at 11:07 AM, Xyne <xyne@archlinux.ca> wrote:
/etc/makepkg.conf -> SRCDEST
Set that to the directory of your choice and all source files should be saved there. Common sources will be automatically detected.
(yeah, via symlinks... but if this is the approach that you meant, it wasn't clear)
Yes SRCDEST is better but won't help for yaourt users. I can't remember when/why I started making my own symlinks, perhaps to prevent one directory being way too full (harder to spring clean). In any case, my apologies, SRCDEST is obviously the better solution here.
Just tried setting an SRCDEST, and yaourt uses it. The redundant downloads issue is then solved. Remains the dependency problem, I thought I would build all packages separately and install all of them: building the utils packages works, they don't rely on anything but building the other which depends on the utils package requires it to be installed, except the old modules package need the old utils package and won't complete unless you force yaourt to ignore dependencies, or purge your system from all said packages before installing the new ones. I believe this is very sub-optimal and this is making NVIDIA users' lives harder, but you guys are boss. I will use the packages I made in my own little corner and won't upload them to the AUR then (I already had at least one user use them happily now, maybe I'll share them elsewhere). Thank you all for your answers. Regards, -- Alucryd "In the name of God, impure souls of the living dead shall be banished into eternal damnation. Amen."
2013/1/25 Maxime Gauduin <alucryd@gmail.com>:
I will use the packages I made in my own little corner and won't upload them to the AUR then (I already had at least one user use them happily now, maybe I'll share them elsewhere).
Thank you all for your answers.
same here greetings
On Fri, Jan 25, 2013 at 10:25 AM, Maxime Gauduin <alucryd@gmail.com> wrote:
On Fri, Jan 25, 2013 at 5:10 AM, Oon-Ee Ng <ngoonee.talk@gmail.com> wrote:
On Fri, Jan 25, 2013 at 9:32 AM, SpinFlo <sl1pkn07@gmail.com> wrote:
i ve use yaourt, yes,
yaourt when make and install sucessful package delete all sources., then the symlink don't work
Don't use yaourt then.
On Fri, Jan 25, 2013 at 9:37 AM, SpinFlo <sl1pkn07@gmail.com> wrote:
sorry for the quotes after message, this is automatic by gmail :S
And it's not 'automatic' by gmail, I use gmail, just click the 'in-line' button (3 dots).
On Fri, Jan 25, 2013 at 11:07 AM, Xyne <xyne@archlinux.ca> wrote:
/etc/makepkg.conf -> SRCDEST
Set that to the directory of your choice and all source files should be saved there. Common sources will be automatically detected.
(yeah, via symlinks... but if this is the approach that you meant, it wasn't clear)
Yes SRCDEST is better but won't help for yaourt users. I can't remember when/why I started making my own symlinks, perhaps to prevent one directory being way too full (harder to spring clean). In any case, my apologies, SRCDEST is obviously the better solution here.
Just tried setting an SRCDEST, and yaourt uses it. The redundant downloads issue is then solved.
Remains the dependency problem, I thought I would build all packages separately and install all of them: building the utils packages works, they don't rely on anything but building the other which depends on the utils package requires it to be installed, except the old modules package need the old utils package and won't complete unless you force yaourt to ignore dependencies, or purge your system from all said packages before installing the new ones.
I believe this is very sub-optimal and this is making NVIDIA users' lives harder, but you guys are boss. I will use the packages I made in my own little corner and won't upload them to the AUR then (I already had at least one user use them happily now, maybe I'll share them elsewhere).
There is a dependency in nvidia/PKGBUILD which is totally wrong depends=( [...] "nvidia-utils=${pkgver}") It leads to a circular dependency at ***build*** time and isn't really needed at run time for nvidia to work. depends=( [...] "nvidia-utils") should be enough to pull it in when installing/updating I've reported it last year but didn't got any response from the maintainer.
Thank you all for your answers.
Regards,
-- Alucryd
"In the name of God, impure souls of the living dead shall be banished into eternal damnation. Amen."
On Jan 26, 2013 6:10 AM, "Rob Til Freedmen" <rob.til.freedman@gmail.com> wrote:
There is a dependency in nvidia/PKGBUILD which is totally wrong depends=( [...] "nvidia-utils=${pkgver}")
It leads to a circular dependency at ***build*** time and isn't really
at run time for nvidia to work.
depends=( [...] "nvidia-utils") should be enough to pull it in when installing/updating
I've reported it last year but didn't got any response from the
needed maintainer.
Have you tried installing different versions of nvidia-utils and nvidia? There's a reason for that versions dependency.
On Fri, Jan 25, 2013 at 11:43 PM, Oon-Ee Ng <ngoonee.talk@gmail.com> wrote:
On Jan 26, 2013 6:10 AM, "Rob Til Freedmen" <rob.til.freedman@gmail.com> wrote:
There is a dependency in nvidia/PKGBUILD which is totally wrong depends=( [...] "nvidia-utils=${pkgver}")
It leads to a circular dependency at ***build*** time and isn't really
at run time for nvidia to work.
depends=( [...] "nvidia-utils") should be enough to pull it in when installing/updating
I've reported it last year but didn't got any response from the
needed maintainer.
Have you tried installing different versions of nvidia-utils and nvidia? There's a reason for that versions dependency.
I did once by mistake, that wasn't pretty. No the hard dependencies are required, the problem is the depends array implies makedepends, but sometimes it is wrong like here. Making depends and makedepends independent, then have makepkg check for depends only at install time would make more sense. However that would mean a rewrite of these 2 arrays in all the PKGBUILDs in the AUR, I don't see that happening. -- Alucryd "In the name of God, impure souls of the living dead shall be banished into eternal damnation. Amen."
On Jan 26, 2013 7:42 AM, "Maxime Gauduin" <alucryd@gmail.com> wrote:
On Fri, Jan 25, 2013 at 11:43 PM, Oon-Ee Ng <ngoonee.talk@gmail.com>
wrote:
On Jan 26, 2013 6:10 AM, "Rob Til Freedmen" <rob.til.freedman@gmail.com> wrote:
There is a dependency in nvidia/PKGBUILD which is totally wrong depends=( [...] "nvidia-utils=${pkgver}")
It leads to a circular dependency at ***build*** time and isn't really
at run time for nvidia to work.
depends=( [...] "nvidia-utils") should be enough to pull it in when installing/updating
I've reported it last year but didn't got any response from the
needed maintainer.
Have you tried installing different versions of nvidia-utils and nvidia? There's a reason for that versions dependency.
I did once by mistake, that wasn't pretty. No the hard dependencies are required, the problem is the depends array implies makedepends, but sometimes it is wrong like here. Making depends and makedepends independent, then have makepkg check for depends only at install time
would
make more sense. However that would mean a rewrite of these 2 arrays in all the PKGBUILDs in the AUR, I don't see that happening.
If that makes sense I'm sure a feature request on pacman wouldn't be out of place. Minimal benefit though, as this really only affects a handful of packages. Doesn't makepkg have a flag to ignore dependencies? Arch laptop died so I can't easily check, that would be easier on everyone involved (except yaourt and similar users)
On Sat, Jan 26, 2013 at 3:25 AM, Oon-Ee Ng <ngoonee.talk@gmail.com> wrote:
On Jan 26, 2013 7:42 AM, "Maxime Gauduin" <alucryd@gmail.com> wrote:
On Fri, Jan 25, 2013 at 11:43 PM, Oon-Ee Ng <ngoonee.talk@gmail.com>
wrote:
On Jan 26, 2013 6:10 AM, "Rob Til Freedmen" <
rob.til.freedman@gmail.com>
wrote:
There is a dependency in nvidia/PKGBUILD which is totally wrong depends=( [...] "nvidia-utils=${pkgver}")
It leads to a circular dependency at ***build*** time and isn't
at run time for nvidia to work.
depends=( [...] "nvidia-utils") should be enough to pull it in when installing/updating
I've reported it last year but didn't got any response from the
really needed maintainer.
Have you tried installing different versions of nvidia-utils and nvidia? There's a reason for that versions dependency.
I did once by mistake, that wasn't pretty. No the hard dependencies are required, the problem is the depends array implies makedepends, but sometimes it is wrong like here. Making depends and makedepends independent, then have makepkg check for depends only at install time would make more sense. However that would mean a rewrite of these 2 arrays in all the PKGBUILDs in the AUR, I don't see that happening.
If that makes sense I'm sure a feature request on pacman wouldn't be out of place. Minimal benefit though, as this really only affects a handful of packages. Doesn't makepkg have a flag to ignore dependencies? Arch laptop died so I can't easily check, that would be easier on everyone involved (except yaourt and similar users)
makepkg does have such an option (-d) and yaourt should too since it is a wrapper. I have nevertheless posted a feature request on makepkg here: https://bugs.archlinux.org/task/33574 While I too believe it's not worth the trouble seeing it concerns only a few packages, I'll see what feedback it gets. -- Alucryd "In the name of God, impure souls of the living dead shall be banished into eternal damnation. Amen."
On Sat, Jan 26, 2013 at 3:30 PM, Maxime Gauduin <alucryd@gmail.com> wrote:
On Sat, Jan 26, 2013 at 3:25 AM, Oon-Ee Ng <ngoonee.talk@gmail.com> wrote:
On Jan 26, 2013 7:42 AM, "Maxime Gauduin" <alucryd@gmail.com> wrote:
On Fri, Jan 25, 2013 at 11:43 PM, Oon-Ee Ng <ngoonee.talk@gmail.com>
wrote:
On Jan 26, 2013 6:10 AM, "Rob Til Freedmen" <
rob.til.freedman@gmail.com>
wrote:
There is a dependency in nvidia/PKGBUILD which is totally wrong depends=( [...] "nvidia-utils=${pkgver}")
It leads to a circular dependency at ***build*** time and isn't
at run time for nvidia to work.
depends=( [...] "nvidia-utils") should be enough to pull it in when installing/updating
I've reported it last year but didn't got any response from the
really needed maintainer.
Have you tried installing different versions of nvidia-utils and nvidia? There's a reason for that versions dependency.
I did once by mistake, that wasn't pretty. No the hard dependencies are required, the problem is the depends array implies makedepends, but sometimes it is wrong like here. Making depends and makedepends independent, then have makepkg check for depends only at install time would make more sense. However that would mean a rewrite of these 2 arrays in all the PKGBUILDs in the AUR, I don't see that happening.
If that makes sense I'm sure a feature request on pacman wouldn't be out of place. Minimal benefit though, as this really only affects a handful of packages. Doesn't makepkg have a flag to ignore dependencies? Arch laptop died so I can't easily check, that would be easier on everyone involved (except yaourt and similar users)
makepkg does have such an option (-d) and yaourt should too since it is a wrapper. I have nevertheless posted a feature request on makepkg here: https://bugs.archlinux.org/task/33574 While I too believe it's not worth the trouble seeing it concerns only a few packages, I'll see what feedback it gets. -- Alucryd
"In the name of God, impure souls of the living dead shall be banished into eternal damnation. Amen."
Maybe I'm missing something ... but what could possible go wrong if we go with "nvidia-utils" instead of "nvidia-utils=${pkgver}" ? There's only one package named nvidia-utils build from the same source and always updated in tandem in the repo. nvidia really only depends on the kernel header and nvidia-utils is build from the same source - depends on nvidia at build time. Please be patient with me, I don't get it yet.
On Jan 27, 2013 4:59 AM, "Rob Til Freedmen" <rob.til.freedman@gmail.com> wrote:
Maybe I'm missing something ... but what could possible go wrong if we go with "nvidia-utils" instead of "nvidia-utils=${pkgver}" ?
There's only one package named nvidia-utils build from the same source and always updated in tandem in the repo.
nvidia really only depends on the kernel header and nvidia-utils is build from the same source - depends on nvidia at build time.
Please be patient with me, I don't get it yet.
AUR is not always in sync with the specific mirror you're using. Mirrors sync at different times. Perhaps nvidia-utils could have a different version in [testing].
On Fri, Jan 25, 2013 at 11:43 PM, Oon-Ee Ng <ngoonee.talk@gmail.com> wrote:
On Jan 26, 2013 6:10 AM, "Rob Til Freedmen" <rob.til.freedman@gmail.com> wrote:
There is a dependency in nvidia/PKGBUILD which is totally wrong depends=( [...] "nvidia-utils=${pkgver}")
It leads to a circular dependency at ***build*** time and isn't really
at run time for nvidia to work.
depends=( [...] "nvidia-utils") should be enough to pull it in when installing/updating
I've reported it last year but didn't got any response from the
needed maintainer.
Have you tried installing different versions of nvidia-utils and nvidia? There's a reason for that versions dependency.
Do you mean something like nvidia-3xx and nvidia-utils-295? For what reason would you want it? Or do you mean nvidia-3xx/nvidia-utils-3xx and also nvidia-295/nvidia-utils-295? How would it possible works? Or do you think there are two different versions with the same name but different versions?
On Sat, Jan 26, 2013 at 1:07 AM, Rob Til Freedmen < rob.til.freedman@gmail.com> wrote:
On Fri, Jan 25, 2013 at 11:43 PM, Oon-Ee Ng <ngoonee.talk@gmail.com> wrote:
On Jan 26, 2013 6:10 AM, "Rob Til Freedmen" <rob.til.freedman@gmail.com> wrote:
There is a dependency in nvidia/PKGBUILD which is totally wrong depends=( [...] "nvidia-utils=${pkgver}")
It leads to a circular dependency at ***build*** time and isn't really
at run time for nvidia to work.
depends=( [...] "nvidia-utils") should be enough to pull it in when installing/updating
I've reported it last year but didn't got any response from the
needed maintainer.
Have you tried installing different versions of nvidia-utils and nvidia? There's a reason for that versions dependency.
Do you mean something like nvidia-3xx and nvidia-utils-295? For what reason would you want it?
Or do you mean nvidia-3xx/nvidia-utils-3xx and also nvidia-295/nvidia-utils-295? How would it possible works?
Or do you think there are two different versions with the same name but different versions?
He meant having, for example, nvidia 313.18 and nvidia-utils 310.32 installed at the same time. -- Alucryd "In the name of God, impure souls of the living dead shall be banished into eternal damnation. Amen."
On Sat, Jan 26, 2013 at 1:20 AM, Maxime Gauduin <alucryd@gmail.com> wrote:
On Sat, Jan 26, 2013 at 1:07 AM, Rob Til Freedmen < rob.til.freedman@gmail.com> wrote:
On Fri, Jan 25, 2013 at 11:43 PM, Oon-Ee Ng <ngoonee.talk@gmail.com> wrote:
On Jan 26, 2013 6:10 AM, "Rob Til Freedmen" <rob.til.freedman@gmail.com> wrote:
There is a dependency in nvidia/PKGBUILD which is totally wrong depends=( [...] "nvidia-utils=${pkgver}")
It leads to a circular dependency at ***build*** time and isn't really
at run time for nvidia to work.
depends=( [...] "nvidia-utils") should be enough to pull it in when installing/updating
I've reported it last year but didn't got any response from the
needed maintainer.
Have you tried installing different versions of nvidia-utils and nvidia? There's a reason for that versions dependency.
Do you mean something like nvidia-3xx and nvidia-utils-295? For what reason would you want it?
Or do you mean nvidia-3xx/nvidia-utils-3xx and also nvidia-295/nvidia-utils-295? How would it possible works?
Or do you think there are two different versions with the same name but different versions?
He meant having, for example, nvidia 313.18 and nvidia-utils 310.32 installed at the same time.
Would that make any sense? Why would I want it? Really?
-- Alucryd
"In the name of God, impure souls of the living dead shall be banished into eternal damnation. Amen."
On Fri, Jan 25, 2013 at 6:34 AM, Maxime Gauduin <alucryd@gmail.com> wrote:
That means downloading the x86_64 blob 2 times and the i686 one 1 time (I'm using a dkms version to not download them one more time for each of my kernels).
Symlinks, man.... I fail to see the benefit beyond download size (which symlinks solve). AUR helpers do a fundamentally tough job and 'because its easier for them' isn't really an argument I think, especially with the bash hackery some PKGBUILDs use.
Oon-Ee Ng wrote:
On Fri, Jan 25, 2013 at 6:34 AM, Maxime Gauduin <alucryd@gmail.com> wrote:
That means downloading the x86_64 blob 2 times and the i686 one 1 time (I'm using a dkms version to not download them one more time for each of my kernels).
Symlinks, man....
I fail to see the benefit beyond download size (which symlinks solve). AUR helpers do a fundamentally tough job and 'because its easier for them' isn't really an argument I think, especially with the bash hackery some PKGBUILDs use.
/etc/makepkg.conf -> SRCDEST Set that to the directory of your choice and all source files should be saved there. Common sources will be automatically detected. (yeah, via symlinks... but if this is the approach that you meant, it wasn't clear)
participants (6)
-
Alexander Rødseth
-
Maxime Gauduin
-
Oon-Ee Ng
-
Rob Til Freedmen
-
SpinFlo
-
Xyne