> > > I don't like dependencies, and kernel-patch is one extra to build the > > > images. The kernel-image packages is from the method the same as for > > > i386 and alpha. Without the dependencies we must not play with debhelper, > > > kernel-package and other possible buggy packages. > > > > I have a working powerpc kernel-image solution. I see this thread never > > came to a resolution so we better do it now. > > > > kernel-patch approach allows one to add other generic kernel patch packages > > into the build easily. On the other hand, it requires a bunch of > > dependencies. > > Well, one thing here is a style/attitude issue - adding additional > dependencies for the kernel build is not such a bad thing, and it means > that there is a central place to fix such problems instead of a host of > kernel-image packages. Kernel-package is a complex piece of software > with a lot of potential for problems, but I still think it is a better > idea to use that.
The kernel-package package should be for end-users. If we think about this way we don't need the kernel-patch package, because it makes no sense if we put the patch(es) in another place then the kernel-image package, where we build the image(s). If we want a new image (with the old way) and we need a little patch for the kernel-package and then also for kernel-patch, we must wait for manoj to accept this patch and upload the newer kernel-package. After that we must play with the kernel-patch _and_ the kernel-image package. In my eyes this is a horror cenario. This is a very bad style for programming and projecting. Thats the reason i don't like such dependencies. One package (the kernel-image-powerpc) is enough. You must not wait and the turnarounds are much faster. > that the Universal IDE patch would apply cleanly. We should definitely > sit down and figure out what patches we need. Yes. > > Comments? What can the kernel-patch approach really do that kernel-image > > can't? > > Well, one of the motivations for the kernel-patch packages, if I > understand correctly, was the x86 IDE kernel flavor (although I'm > greatly confused, as I can not find the IDE patch package any more, so > I may be misremembering). This patch now applies to powerpc as well; No, the idea for the kernel-patch approche (inventet by me) was to split the dawm i386 centric off. Before the kernel-patch and the _new_ kernel-image method, you couldn't build an image for powerpc. I'll now (as the maintainer for the powerpc kernel-patch package) remove this package. The kernel-patch package has nothing to do with x86 stuff. > in fact, I need it to boot my desktop. And, being an occasional kernel > hacker, I've also got a few local patches I regularly apply. Thus I > would prefer to do kernel packages in a way that easily supports > building with alternate sets of installed patches. Doit in the kernel-image package, much easier in this package as to play with serveral other packages. > I think it's more useful, on the whole. No! For an example: How will you remove older kernel-patches from /usr/src/kernel-patches/* ? Note: we have no kernel-package-base, that will generate (and also delete) the arch directory. Regards, Hartmut