Yep... the thing to remember is that the goal of radeon-rewrite is to be able to use a single code base for both TTM and non-TTM systems, since both of them will need to be fully supported over the next year or two. On a non-TTM system, radeon-rewrite is supposed to be "no worse, and easier to maintain in the future", while on a TTM system it should provide a good base for adding new features (eg the VBO and OQ work already in progress).
Between non-TTM and TTM, testing on non-TTM systems is definitely the most time-critical since all future Mesa releases will be using this code base and any problems translate directly into regressions for current users.
Michael's testing confirmed that there is more work to do before the next Mesa release. This was understood by the devs before the merge, but the consensus was that it made more sense to merge now and get more people testing the code than to try and "make the code perfect in a branch" and drop it into master at the last minute.
Last edited by bridgman; 06-15-2009 at 02:05 PM.