Hi all, [Greg, this is not directed only at you, but is a wider issue and you have given me the opportunity to dicsuss it]. On Thu, 30 Apr 2009 22:22:56 -0700 Greg KH wrote: > > On Fri, May 01, 2009 at 03:01:15PM +1000, Stephen Rothwell wrote: > > > > [Aside: the first I can find of this is the patch submission to LKML > > within the last day ... so "posted, reviewed, unit tested"? I assume > > this has had some discussion somewhere, but Google doesn't know where.] > > It was posted today for discussion, which didn't seem to happen. That was my point - waiting less than a day between the posting of a new feature publicly and its inclusion in linux-next seems a bit impatient. *And* you seem to have gotten some discussion now :-). > It has been unit tested a lot in SuSE's kernels with review from our > kernel developers (hence the 3 signed-off-bys). We use it to speed up > booting a lot because we have to use an initramfs (like all distros need > to for various reasons.) It aleviates the udev coldplug issues a lot, > and the embedded developers have been very happy to see this (for some > reason they only like writing private emails, not to the list, which is > unfortunate.) Having seen what is coming in this patch for linux-next on Monday it is clear that there is more work to be done on this before it is ready for Linus' tree. linux-next is for integration testing, it is not a development tree. Everything in it should, in the tree maintainers opinion, be ready for Linus' tree (if he happens to go insane and open the merge window unexpectedly). I am making no comment about this particular feature, just what should be in the linux-next tree. I would expect maintainers to have (at least) three (publicly available) trees (where "tree" can be a quilt (sub)series, or a git branch etc): development, ready and bug-fixes. Development is ongoing work and new features, not yet ready for Linus' tree. Ready is pretty much done - it may not be bug free, but it is (as far as is reasonably possible) tested and ready to go into Linus' tree. Bug-fixes is the stuff for Linus after the merge window closes. Maybe I have not made this clear enough in the past, but linux-next should really (except for a couple of clear exceptions like "staging") only contain the ready and bug-fixes trees. Discussion, anyone? I am open to changes if people think things should be run differently, but the above makes sense to me. -- Cheers, Stephen Rothwell sfr@canb.auug.org.au http://www.canb.auug.org.au/~sfr/