On Sun, Jan 12, 2025 at 08:48:44PM -0800, Christoph Hellwig wrote: > On Fri, Jan 10, 2025 at 12:50:26PM -0500, Brian Foster wrote: > > Yeah, I agree this is a wart. Another option I thought about was > > creating a new flag to declare which iteration mode a particular > > operation uses, if for nothing else but to improve clarity. > > I actually really like the model where the processing loop always > advances. It'll make a few things I have on my mind much easier. > > That doesn't mean I want to force you to go all the way for the initial > patch series, but I'd love to see a full switchover, and preferably > without a too long window of having both. > Ok, thanks. I'm on board with that, just need to dig back into it to be certain of details or roadblocks.. Another thing that crossed my mind is that it might be preferable to convert across at least one release cycle regardless, just from a risk management standpoint. I.e., introduce for zero range in one release, let the test robots and whatnot come at me with whatever issues that might exist, and then follow up with broader changes from there. But anyways, one thing at a time.. Brian > > reason.. would we think this is a worthwhile iteration cleanup on its > > own? > > Yes. >
