On Wednesday, 16 January 2019 10:15:22 GMT Neil Bothwick wrote: > On Wed, 16 Jan 2019 09:50:58 +0000, Peter Humphrey wrote: > > This box acts as an emerge server for a 32-bit Atom. So I NFS-mount the > > Atom's portage tree in a 32-bit chroot, build the packages it needs and > > then emerge the packages on the Atom. > > > > This morning I found something odd: the Atom wanted to emerge perl > > 5.24, even though 5.26 was already present, so I got a long list of > > clashes. This is the emerge command on the Atom: > > > > emerge -auDvUK --jobs=2 --load-average=4 --changed-use --changed-deps \ > > > > --with-bdeps=y --nospinner --keep-going world > > > > The solution was to delete the 5.24 package left over from an earlier > > emerge. All was then well. > > > > It looks as though the -K switch caused portage to want to emerge the > > package even though it wasn't indicated by the update. Is it supposed > > to do that? > > Was there a suitable package for 5.26 in $PKGDIR? The -K switch forces > portage to use a package, unlike -k, so if the exact 5.26 version you had > installed had been removed from the tree in favour of an updated/fixed > version, portage would have to downgrade if you hadn't built the new > package.
No, the 5.26 package was there alongside the 5.24, and portage didn't want to downgrade. Once I'd removed the 5.24 package portage was no longer confused. I have checked that the host and client have identical world and package.* files. Also make.conf, except for things like --jobs and buildpkg. -- Regards, Peter.
