On Thu, 2020-03-12 at 09:47 -0400, Theodore Y. Ts'o wrote: > On Thu, Mar 12, 2020 at 02:37:31AM -0700, Joe Perches wrote: > > As I have suggested a few times, better still > > would be to have a mechanism for scripted patches > > applied possibly as single treewide patch. > > > > Likely applied only at an -rc1. > > > > The stated negatives to a treewide mechanism > > have been difficulty to backport to -stable. > > Any time we do a massive, disruptive change to the code base, it's > going to cause problems to -stable. It means that bug fix patches > won't necessarily auto-apply, and some will require manual fixups > afterwards That's mostly a tools problem than a real problem. > Given that this change doesn't really fix any bugs, I'd have to ask > the question --- is it *worth* it? We really need to apply a certain > amount of cost/benefit analysis around this. > > If it were really important, the thing we could do is to apply a > single treewide patch at some point after the merge window. I'd > suggest after -rc2, myself, but reasonable people can differ. And > then, if it were *really* important we could run the same script on > the stable kernels. > > But for changing "/* fallthrough */" to "fallthrough;" > > Does this ***really*** matter? That depends a bit on whether clang is your compiler of choice. > Why are we tying ourselves up in knots > trying to do this all at once? Discretely or treewide, all at once or done over time, the impact problem to backports is the same.
