On 18 March 2016 at 08:47, Markus Armbruster <arm...@redhat.com> wrote: > Paolo Bonzini <pbonz...@redhat.com> writes: > >> On 15/03/2016 16:03, Markus Armbruster wrote: >>> This series tackles two issues: >>> >>> * qapi/error.h and therefore qapi-types.h get included into every file >>> via qemu/osdep.h. Touch the QAPI schema, recompile the world. As >>> the QAPI maintainer, I get to touch it regularly, so this is an itch >>> I need to scratch. PATCH 01-04 drop qapi/error.h from qemu/osdep.h. >>> Unfortunately, the number of .o depending on qapi-types.h barely >>> changes, because it gets included by so many other headers. More >>> work is needed here. >>> >>> * qemu-common.h explicitly asks not to be included in headers, but we >>> do it anyway. PATCH 05-15 clean that up. Cuts the number of .o >>> depending on qemu-common.h from ~95% to ~35%. >>> >>> For additional motivation, see the "Our use of #include is >>> undisciplined, and what to do about it" thread. >> >> Patches 1-13 and possibly 15 are great. > > Is this an R-by or Acked-by for 01-13? > > Could they still be considered for 2.6? Rebasing them eats > compile-time like no tomorrow.
If Paolo wants to provide an R-by I'm happy to take them for 2.6 -- the possible failure modes are basically just "something stops compiling" which should be an easy fix if it happens. You could make my life easier by sticking them in a pullreq. PS: master already has a patch to remove config.h. thanks -- PMM
