FS#8840 - Request loosening of dependency requirements for module packages

Attached to Project: Arch Linux
Opened by Michael Towers (gradgrind) - Wednesday, 05 December 2007, 08:10 GMT
Last edited by Allan McRae (Allan) - Sunday, 15 March 2009, 02:45 GMT
Task Type Feature Request
Category Packages: Core
Status Closed
Assigned To Paul Mattal (paul)
Aaron Griffin (phrakture)
Architecture All
Severity Low
Priority Normal
Reported Version 2007.08-2
Due in Version Undecided
Due Date Undecided
Percent Complete 100%
Votes 1
Private No

Details

In the module packaging guidelines (http://wiki.archlinux.org/index.php/Kernel_Module_Package_Guidelines) one is required to make a module package depend on its associated utilities.

In most cases this will be eminently reasonable. However there are cases (for example aufs) where the utilities are optional extras and these utilities themselves have further dependencies which might be quite extensive and in certain circumstances undesirable (for example because of size).

I admit this case might well be rare. Nevertheless this same requirement doesn't exist for modules included in the kernel (it would be a pain if it did!), thus the kernel doesn't depend on, for example, unionfs-utils or squashfs-tools.

My suggestion, which I believe would be painless to implement (no existing packages would _need_ to be changed) and in line with general Arch policy (not to have unnecessary dependencies), would be to change the line:

"... make sure nvidia depends on nvidia-utils ..."

to:

"... if nvidia-utils is required for the use of nvidia, make sure nvidia depends on nvidia-utils ..."

Thus cases where the utilities are optional would be left to the discretion of the package maintainer.
This task depends upon

Closed by  Allan McRae (Allan)
Sunday, 15 March 2009, 02:45 GMT
Reason for closing:  Fixed
Comment by Aaron Griffin (phrakture) - Wednesday, 05 December 2007, 17:05 GMT
How about a compromise: we change the wording to "foo SHOULD depend on foo-utils, but it is not necessary. This decision is up to the maintainer of the package"
Comment by Roman Kyrylych (Romashka) - Sunday, 06 April 2008, 09:42 GMT
status?
Comment by Gavin Bisesi (Daenyth) - Tuesday, 07 October 2008, 15:56 GMT
I like Aaron's suggestion.
Comment by Glenn Matthys (RedShift) - Thursday, 11 December 2008, 07:15 GMT
We have optdepends now, so this can be implemented. Michael, are you referring to any specific packages at this point?
Comment by Michael Towers (gradgrind) - Thursday, 11 December 2008, 08:40 GMT
Originally I had aufs in mind, as mentioned in the original post, but that may be no longer relevant. I bet it's not a very significant point, but a reformulation along the lines discussed would seem logical to me.
Comment by Paul Mattal (paul) - Thursday, 11 December 2008, 11:10 GMT
I also like where Aaron was going, and have tried to incorporate that sentiment minimally into the existing language as follows:

http://wiki.archlinux.org/index.php?title=Kernel_Module_Package_Guidelines&diff=55245&oldid=53935

Feel free to hack away, Aaron, if you prefer some other wording.
Comment by Aaron Griffin (phrakture) - Thursday, 11 December 2008, 17:37 GMT
That seems fine to me, though perhaps we should add an example of an acceptable reason?
Comment by Allan McRae (Allan) - Sunday, 15 March 2009, 02:45 GMT
Seems fixed now and this is a wiki page so anybody can make further adjustments. Closing.

Loading...