Arch Linux

>makepkg - the complete file for anyone who wish to test this

oops... looks like attached the patch twice.
this is the complete makepkg with the patch applied

Hmmm, I like this... it's pretty close to what I was looking for, though the syntax itself seems a little odd. I just don't know how to refine it. Thanks a lot for the patch, i will look into it in the near future.

Your most welcome :)

I would really love to see makepkg support multiple pkgs whether it's in the form of my suggested patch or another.

Perhaps we can discuss more on the syntax? I think I can help.

Added Dan.

As for the syntax, I can't say I can think of anything better. However, let me throw this one out there for consideration: if all the (4 or 5) ndiswrapper modules could be built in one package, that'd be great.

See the attached file (how I'd redo the same PKGBUILD).

I was thinking that some sort of separate "build_foo" might be nice for given sub packages, but that would require either additional extraction of the source, or cleanup after each build

Hmmm... yes I agree. A separate "build_foo" coupled with source re-extraction would be the way to go.

I'm wondering though about patches that are in the form of tar.gz. How would makepkg selectively apply tar.gz patches only for pkg that require them?

Btw, in my sample PKGBUILD, I've ordered it such that ndiswrapper-utils precedes ndiswrapper. This is so that the "makepkg -i" command works. Otherwise pacman will complain about unresolved deps. I guess we could enhance makepkg to sort the pkgs in the order that pacman expects.

The bigger question though is that for the ndisPKGBUILD sample, what would be the correct behavior during "makepg -i"? Perhaps makepkg should not install any pkgs that have a corresponding "build_foo"?

Thoughts?

Ah, that one is more complicated - I didn't actually notice that. I was thinking in terms of "the package is ndiswrapper, regardless of how we split it" so Imoved that to the top.

As for -i behavior, I would assume it would check the subpackages for deps first, and use them if found.

> # default pkgname_foo to "${pkgname}-foo"

Oh btw, nice touch. I like it. +1 from me.

> The bigger question though is that for the ndisPKGBUILD sample, what would be the correct behavior during "makepg -i"?
> Perhaps makepkg should not install any pkgs that have a corresponding "build_foo"?

Ummm... I guess my question really is about what happens when the sub-packages provides the same module.
Silly me for incorrectly assuming that "ndiswrapper-ck" and "ndiswrapper-suspend2" provides "ndiswrapper".

Here's a proposal for the "build_foo" usecase.
See attached PKGBUILD that builds kernel26 and kernel26suspend2.

Is this useful for package maintainers?
Or is the syntax so convoluted that maintainers would prefer to maintain separate PKGBUILDs?

If this will be merged - I think Alessio's patch ( http://archlinux.org/pipermail/pacman-dev/2007-April/008171.html ) can be integrated into this one.

> Btw, in my sample PKGBUILD, I've ordered it such that ndiswrapper-utils precedes ndiswrapper.
> This is so that the "makepkg -i" command works. Otherwise pacman will complain about unresolved deps.

Hmmm... I can't seem to duplicate this problem anymore. Strange.
Pacman installs the packages correctly even if I place ndiswrapper before utils.

So I guess my original patch can be considered complete.

Please see http://bugs.archlinux.org/task/7982?project=3&pagenum=2 for another solution