Am Sonntag, den 03.07.2005, 19:12 +0300 schrieb Ville Skyttä:
On Sun, 2005-07-03 at 09:21 -0500, Tom 'spot' Callaway wrote:
On Sun, 2005-07-03 at 16:11 +0200, Thorsten Leemhuis wrote:
I think you're mostly right. mainpkgversion and mainpkgrelease are redundant, however. Just use Version and Release for that.
Yep.
Removed already :))
A policy or at least guidelines exactly where to place the modules below /lib/modules/%{kver} would be nice.
Yes.
My proposal: Put it at the same place where the normal module would be places by a normal "make install" of that package. That has several benefits:
- If a external Documenation says "Look for the module in /lib/$(uname -r)/somewhere it is exactly found there where the Documentation says. No special fedora way
- People sometimes try to compile a module on their own and go for the rpm later (or the other way round). They might end with two different modules in the kernel tree -- witch one wins? This can't happen if the modules are installed at their usual place.
Personally I think /lib/modules/%{kver}/kernel/... is bad, these are not modules shipped with the kernel so intruding that space feels wrong.
I don't share that opinion. I think it's confusing for no real reason. If we have a gnome-app or enhancement it's installed in the gnome-space in the filesystem also.
"updates" is bad too, these aren't updates to in-kernel modules.
Yes.
The kernel-module itself is now placed in %package -n kernel-module-%{mainpkgname} So only one SRPM is created no matter how much different kernel-modules are build.
This makes sense, good thinking.
Except as described in my earlier mail to this thread, if the main package's NVR doesn't change between rebuilds, we have a problem with -debuginfo for these builds. https://www.redhat.com/archives/fedora-packaging/2005-June/msg00055.html
Uhh, sorry, I knew I was forgetting something. Yes, that should be fixed. So what are our options:
1) create the debug-pkg ourself and don't rely on the internal rpm solution. 2) use a spec similar two my first proposal where the actual source of the kernel-module is in an external rpm that is a BuildRequire. Resulting SRPMS are small then. 3) No sub-pkg -- gives a lot of SRPMS that might take a lot of space... 4) <your idea here>
If 1) is easy I'll vote for that. Otherwise I'd like to give 2) a try. 3) is a no go when I think of kernel-modules like ati-fglrx where each SRPM package would take around 4,5 MByte currently. Especially if we want to rebuild kernel-module for nearly all kernels that ever were published (someone suggested that somewhere in this thread iirc).