On Nov 16, 2007 3:41 PM, Jason Chu <[EMAIL PROTECTED]> wrote:
> On Fri, Nov 16, 2007 at 11:09:13AM -0600, Aaron Griffin wrote:
> > On Nov 16, 2007 6:04 AM, Thomas Bächler <[EMAIL PROTECTED]> wrote:
> > > Roman Kyrylych schrieb:
> > > >> bash 3.2.025-2 -> depends on readline>=5.2
> > > >> readline 5.2-5 -> depends on ncurses
> > > >> ncurses 5.6-6 -> depends on glibc
> > > >> glibc 2.7-7 -> depends on bash -> oops!
> > > >>
> > > >> But it looks like there are other ones as well.
> > > >>
> > > >
> > > > I'd say we remove glibc dependency from ncurses or bash dependency
> > > > from glibc. :-/
> > > > /me scratches his head
> > >
> > > This cyclic dependency cannot be avoided: bash needs glibc to be
> > > executed, but glibc needs bash in its post_install.
> >
> > I do know one way to resolve this.... we can ship something like dash
> > for /bin/sh _only_ and depend on that... sounds like a headache, but
> > it'd work
>
> Wouldn't that mean that anything that used /bin/sh as its interpreter would
> have to be valid dash script and not the bash script that we expected?

If you expected bash, that is an invalid assumption.

<http://projects.archlinux.org/git/?p=pacman.git;a=blob;f=lib/libalpm/trans.c;h=bcd0707d567befa000b161dc08b2051c2dd82111;hb=2f55733be35e7ec4188ddab17a92ed67faa7210d#l577>

We should make that more clear in the PKGBUILD.5 manpage.

-Dan
_______________________________________________
arch mailing list
[email protected]
http://archlinux.org/mailman/listinfo/arch

Reply via email to