On Mon, Mar 11, 2024 at 7:40 PM Paolo Bonzini <pbonzini@xxxxxxxxxx> wrote: > > On 3/8/24 16:40, Sean Christopherson wrote: > > You're missing the point. I don't care when patches land in the RISC-V tree, nor > > do I care that you made a last minute tweak to fix a bug. I care when commits > > show up in linux-next, and*none* of these commits were in linux-next until > > yesterday. > > > > $ git tag -l --contains 2c5af1c8460376751d57c50af88a053a3b869926 > > next-20240307 > > next-20240308 > > > > The*entire* purpose of linux-next is to integrate*all* work destined for the > > next kernel into a single tree, so that conflicts, bugs, etc. can be found and > > fixed*before* the next merge window. > > Indeed, and this is more important as more work is routed towards > different trees. At this point we have 5 more or less active > architectures, and especially in selftests land it's important to > coordinate with each other. > > Anup, ideally, when you say that a patch is "queued" it should only be a > short time before you're ready to send it to me - and that means putting > it in a place where linux-next picks it up. For x86 I generally compile > test and run kvm-unit-tests on one of Intel or AMD, and leave the > remaining tests for later (because they take a day or two), but in > general it's a matter of days before linux-next get the patches. > Currently, I was collecting patches in the queue and allowing people (including myself) to test the queue for a longer time and updating "next" branch only a couple of weeks before I send PR. Going forward, I will follow your suggestion. This means once a series is tested on queue, I will move it to the "next" branch sooner so that "linux-next" picks it sooner. Regards, Anup