Andy Green wrote: > They really want a trickle of patches over time they can > digest without choking to death, not the "mochipatch"
*grin* True enough :-) > of way patchutils enables -- sort of "patch domain arithmetic": Hmm, "patch calculus" anyone ? ;-) > if we effectively marked up the tree like stuff in /arch/arm/kernel is > one kind of patch, stuff in /arch/arm/mach-s3c2440 is another and so on > maybe there is a way through it all that gets you the mokopatch effect > synthesized without the work. That assumes that overlaps aren't a big deal. But we still do have plenty of them :-( What would be really light on upstream would be a collection of patches that's basically 1/N fiq-generic (as a platform feature) 2/N neo1973-basis 3/N gta02-platform 4/N smedia-basis (device-independent) 5/M smedia-gta02-glue 6/N smedia-mmc 7/N smedia-mmc-gta02-glue etc. Then they could each be reviewed separately, then updated for final submission. If any fail review completely, nothing if the failed item would "leak" into the tree (such as some "here are all the resources on that platform" file), but the patches following it would have to be adjusted. - Werner
