The einstall() function in ebuild.sh prepends the ABI target libdir to the global variable EXTRA_EINSTALL.
Since the variable is global, the modifications accumulate across calls, which messes multilib loops that call einstall().
I think the EXTRA_* variables should not be modified by portage, if they're supposed to be user settable.
Patch to follow.
Created attachment 119249 [details, diff]
preserve EXTRA_EINSTALL in einstall() function
This patch preserves the contents of EXTRA_EINSTALL, by using the EI_DESTLIBDIR variable directly instead.
`einstall` shouldnt be used in the first place, but that's a different topic
the usage should be changed to match EXTRA_ECONF usage in econf() ... in other words, introduce a local LOCAL_EXTRA_EINSTALL and utilize that
Created attachment 119319 [details, diff]
updated patch to einstall() in ebuild.sh
This patch incorporates spanKY's suggestions to solve the problem.
thanks, ive merged that one
I have another bug, similar to this one, reported as #178378, which I think got lost between the cracks last week. It started out as an eclass bug, but the problem also exists in ebuild.sh, so it might be relevant to core also.
(In reply to comment #5)
> I have another bug, similar to this one, reported as #178378, which I think got
> lost between the cracks last week. It started out as an eclass bug, but the
> problem also exists in ebuild.sh, so it might be relevant to core also.
In svn r6558.
This has been released in 220.127.116.11.