On Sun, Jan 12, 2014 at 3:11 PM, Lukas Fleischer <archlinux@cryptocrack.de> wrote:
On Sun, 12 Jan 2014 at 15:51:48, Anton Larionov wrote:
Hello,
I was under the impression that .AURINFO was introduced to override some fields in PKGBUILD when they are written in format which can't be properly displayed by AUR (or maybe I've missed something). But why do you want to force it's usage for all packages? In most cases AURINFO will just duplicate same fields from PKGBUILD.
The long-term plan is to use it for all AUR packages, improve the format and finally make it an official feature of makepkg(8) (the file will probably be called .SRCINFO then but we're far from there). See my other reply to Sebastien for some reasons on why it should be used.
So the official goal is to have it generated as part of makepkg -S? Because I see that as the only way the format will get popular: if it's nobody's problem. J. Leclanche
Also I have some questions about it's format:
1) If package has different dependencies for 86_64 and 686, what should I put in depend array?
Good question. This cannot be answered properly, though, since dependencies actually are a property of the binary package and not a property of the source package. Maybe we should loosen the format for dependencies of source packages and allow optdep-like comments? Something like:
depends = foo depends = bar depends = foobar: x86_64 only
Just an idea. Comments welcome.
2) Which 'pkgname' will be unique - from PKGBUILD or AURINFO? E.g if I upload package with name 'foo' and overriden name 'bar' will someone be able to upload new package with name 'foo'? Or 'bar'?
Only the information from .AURINFO will be used. You can already trick the AUR into reading a completely different name from the PKGBUILD than it actually produces (and that problem is unavoidable), so that isn't a (new) issue.
--
Regards, Anton Larionov