[BC ABI] gcj-abi-2-dev-branch created

[Tom Tromey]

>>>>> "Per" == Per Bothner <per@bothner.com> writes:

Per> If the plan is to for the foreseeable future to support two ABI's
Per> in the same source tree, do we really need a separate branch?

I think this code is pretty intrusive, so a branch makes sense while
we iron out the bugs.

Per> What time horizon are you thinking about in terms of when to
Per> merge the branch back?

I don't know.  Personally I'm hoping this could go in "3.5" as
supported, with "real ABI stability" guaranteed starting in "3.6".

Per> In any case, is there a design document?  If not, we need one now.

There was an IBM paper about an approach to this that serves as a
useful starting point.  It is easily googled.  Also there were some
long posts from Bryce describing the approach on this list in the
past; I don't have a URL handy though.  They might turn up under
"-findirect-dispatch".  These sources cover everything except the
verification stuff, which probably won't affect your goal anyway.

Also Andrew and I are working on a paper for the GCC Summit, I wrote
the outline last night.

Tom