On Mon, Jun 22, 2020 at 10:44:02AM -0400, Sasha Levin wrote:
On Mon, Jun 22, 2020 at 02:27:57PM +0100, Mark Brown wrote:
On Mon, Jun 22, 2020 at 08:31:18AM -0400, Sasha Levin wrote:
How come? This is one of the things stable rules explicitly call for: "New device IDs and quirks are also accepted".
I would expect that to be data only additions, I would not expect that to be adding new code.
These come hand in hand. Take a look at the more complex cases such as sound/pci/hda/patch_*
There are more complex cases, I'm just not sure how good an idea backporting them.
It would be much better to not have to watch stable constantly like we currently do - we're seeing people report breakage often enough to be a concern as things are, we don't need to be trying to pile extra stuff in there because there's some keywords in a changelog or whatever. The testing coverage for drivers is weak, increasing the change rate puts more stress on that.
Shouldn't we instead improve testing here? nvidia for example already provides Tegra testing for stable releases, if the coverage isn't sufficient then let's work on making it better.
Obviously it'd be good to improve the test coverage, but I think that's something that needs doing before backporting loads of stuff rather than after. For this automated stuff I'd much rather see positive confirmation that the change had been tested on relevant systems (not just something with a similar SoC), especially on the edges where we're getting to board specific things.