On Thu, 29 Jun 2023, Mark Brown wrote: > On Thu, Jun 29, 2023 at 12:14:00PM -0600, Rob Herring wrote: > > On Thu, Jun 29, 2023 at 11:59 AM Mark Brown <broonie@xxxxxxxxxx> wrote: > > > > Right, I'm suggesting applying the core part without waiting for every > > > single leaf driver to be lined up rather than the other way around - > > > that way the core part is stable and the leaf drivers only have issues > > > with changes in their subsystems that they'll have anyway even with > > > waiting. Leaf drivers can be added on top as they're ready and if > > > something misses a release then it can go through the subsystem, and if > > > people do end up wandering off then you've still got whatever did get > > > merged in case someone else wants to pick things up. > > > I misunderstood. I thought you wanted to apply things to get them out > > of your queue. > > Well, I *do* but that's got issues especially when things get stuck so > I'm not going to. > > > That doesn't work when the leaf drivers depend on the > > core, so what do we do there? A branch or Lee takes everything? That's > > almost always the case with the bindings as the core binding > > references the child node bindings. My preference there would be that > > Lee picks up all the bindings with the core driver. > > My suggestion is that once the core is ready to apply that and also > start applying everything else to Lee's tree as it's ready. A branch > also works and might come in handy anyway in the case where there's some > subsystem wide updates in some other subsystem (since it avoids having > to pull the whole MFD tree in or anything like that) but it's not > essential to the idea. The issue we currently have is that the core usually comes with a header file which is included by some or all of the leaf drivers. If leaf drivers are pulled in without that header, the drivers will fail to build which will make people grumpy. The suggestion of a separate branch that's added to over time as leaf drivers become ready is even more work that a one-hit strategy. It will also mean littering the working branch which a bunch more merges and/or more frequent rebases than I'm happy with. -- Lee Jones [李琼斯]
