Messages in this thread | | | Date | Fri, 13 Jul 2012 16:09:11 -0700 | Subject | Re: [PATCH v2 05/12] rbtree: performance and correctness test | From | Michel Lespinasse <> |
| |
On Fri, Jul 13, 2012 at 3:45 PM, Andrew Morton <akpm@linux-foundation.org> wrote: > On Fri, 13 Jul 2012 15:33:35 -0700 Michel Lespinasse <walken@google.com> wrote: >> Ah, I did not realize we had a precedent for in-tree kernel test modules. > > hm, well, just because that's what we do now doesn't mean that it was a > good idea ;) These things arrive as a result of individual developers > doing stuff in their little directories and no particular thought was > put into overall structure. > > It could be that it would be better to put all these tests into a > central place, rather than sprinkling them around the tree. If so, > then your patch can lead the way, and we (ie: I) prod past and future > developers into getting with the program. > > otoh, perhaps in-kernel test modules will rely on headers and constants > which are private to the implementation directory. So perhaps > sprinkled-everywhere is the best approach.
I think it is at least reasonable. Where we could improve, however, would be on the Kconfig side of things.
>> I don't think my proposal was significantly better than this >> precedent, so I'll just adjust my patch to conform to it: >> - move rbtree_test.c to lib/ >> - modify just lib/Makefile and lib/Kconfig.debug to get the module built. >> >> Will send a replacement patch for this (so you can drop that one patch >> from the stack and replace it with) > > OK, you could do that too. That way you avoid the problem and we can > worry about it later (if ever), as a separate activity.
Going to attach as a reply to this email.
-- Michel "Walken" Lespinasse A program is never fully debugged until the last user dies.
| |