Following the discussion on the ml, I'm opening a bug to keep track of it. I think we haven't reached a definite solution yet. As for multilib, it would be really helpful if different CHOSTs were used indeed. Since toolchain uses only the default CHOST for the prefix and I can't think of anything that would actually rely on the alternate ABI CHOSTs, I think it should be reasonably safe to change the two other CHOSTs. And since multilib has limited scope of application and is very experimental ATM, I think we can handle the fallout of changing CHOST to the Debian one.
I still think this need discussion in the mailing list otherwise this bug will end up with too many comments.
(In reply to Markos Chandras from comment #1) > I still think this need discussion in the mailing list otherwise this bug > will end up with too many comments. Ditto. Can we please close this bug for now and continue discussion on the list.
Any way that works for you. Though the list didn't see any new reply for a few days...
(In reply to Michał Górny from comment #3) > Any way that works for you. Though the list didn't see any new reply for a > few days... Please restart the discussion. Mike suggested that you should not assume the uniqueness of the CHOST to do your multilib stuff. He suggested CHOST-ABI or something. If that does not work for you lets discuss on the mailing list why. If you really want us to chance CHOST, then we can also discuss that. But not on this bug ;)
