Raphael Hertzog wrote: > On Wed, 24 Oct 2007, Eddy Petrișor wrote: >>> I'm sorry, I don't see how triggers would help us here... care to explain? >> As I understand the purpose of acpi-support is to support and configure every >> acpi enabled system that exists in an attepmt to make sure the current HW is >> supported and acpi-support uses the appropriate scripts. >> >> Ok, suppose you leave the installation of the hw dependant packages to a >> third >> party (say a tool which would be smart about what it installs via discover >> information, or even d-i). So what is left for acpi-support to do is to be >> interested about a trigger enabled by a package such as nvclock. > > What for? We don't need to be informed that a package is installed. We
You would if acpi-suspend wouldn't blindly and needlessly depend on hw spcific packages on systems where is not necessary. > just need to do the right thing when the user ask the computer to suspend > itself. Yes, but this bug is about pulling unneeded packages on systems which don't need it. > People keep inventing scenarios of improper usage of triggers. :) I fail to see how this is an improper use of a trigger if you think that apci-support would leave dependency handling related to package specific to different pieces of hw to some tool that is able to do that discrimination instead of pulling those package on every system out there via a blind (wrt hw specific) tool like dpkg. Having nvclock installed as a consequence of a user choice of a system that is able to detect and install that package on the systems that need it seems (to me) like a nice and elegant way to enable different functionalities within acpi-support. >> I know the weak link of this approach is the magnificent tool that installs >> hw >> spcific packages based on discover findings, but discover already includes a >> script that can install packages based on those findings, I guess is >> necessary >> to check how much more it needs to be expanded[1] and integrate it somehow in >> the installation. > > Right, but all this is unrelated to triggers. If you think in terms of what's the current situation. To make it short, this is what it would look like if triggers would be used: 1. acpi-suport no longer pulls unconditionally nvclock and all other hw specific packages 2. acpi-support leaves the package pulling to a tool that is able to take those decisions based on the actual hw configuration 3. scripts are no longer ran in a bundle on all systems, but instead a system like it was proposed in #374734 is used; the symlinks are put in place via triggers This way you solve two+ issues at the same time, fixing properly this bug and similar bugs (#410918, #434566, #438665) and allowing flexibility in the system, as was requested in #374734. -- Regards, EddyP ============================================= "Imagination is more important than knowledge" A.Einstein
signature.asc
Description: OpenPGP digital signature
