Huawei's phones in China are still affected Google's revocation of the Android license. Think security patch updates and platform betas - those still matter to Chinese Android phones.
Conversation
Replying to
From Google, they mostly only lose the 30 day early access to AOSP security updates and it really doesn't take 30 days to make a build with those and thoroughly test it. Being cut off from driver and firmware updates from companies adhering to this policy is a bigger problem.
1
2
I don't know how many phones they have with a Snapdragon SoC or a Qualcomm baseband, but I don't know how they would get updates for those. The same thing applies to other components like Qualcomm Atheros or Broadcom Wi-Fi. There are lots of components with firmware updates.
2
2
The worst case scenario for security updates in AOSP code is being delayed by 30 days, but it would really be on them if it actually takes them 30 days to do what should take 3 days at most. Google has already implemented / tested the patches when they give them to vendors early.
1
So, they need to build, and run through the VTS / CTS along with a bunch of other testing. For unlocked devices without carrier nonsense, a couple days or at most a week is really more than a enough to deal with it, especially with a proper team dedicated to dealing with this.
1
Replying to
I mean, you still need to do that to make sure it works properly. I don't care about conforming to the compatibility rules for GrapheneOS but I still run through the whole CTS regularly. I have a file where I record all of the failures and mark the ones that are intended to fail.
1
1
An example, from making INTERNET into an enabled by default runtime permission:
* CtsPermission2TestCases - pass
- intentional failure: android.permission2.cts.PermissionPolicyTest#testPlatformPermissionPolicyUnaltered
Not going to include the traceback since it's too long.
1
1
Another example:
* CtsAppSecurityHostTestCases - pass
- intentional failure (serial not visible to legacy apps): android.appsecurity.cts.PermissionsHostTest#testSerialAccessPolicy
The tests are still very important, and if I really wanted I could build a modified CTS.
1
2
I have access to the bulletin previews, so I know what is and isn't in those. They might not be as helpful as you'd think. They only fully cover the YYYY-MM-01 patch level and only have a tiny bit of what's needed for the YYYY-MM-05 patch level. OEMs need vendor support for that.
The zip file with the fixes is divided into a patches directory with the YYYY-MM-01 fixes and a code_snippets directory with a tiny subset of YYYY-MM-05 fixes. It usually only has the generic kernel patches not tied to an SoC for supported kernel versions (3.18, 4.4, 4.9 ATM).
1
1
I could post one of the zip files for an earlier patch level since it's not under embargo once the bulletin goes out, if you're curious about the format vendors receive these fixes. There's some other supplementary info other than the zips that I might not be allowed to publish.
1
Show replies