[krajzeg]

Their fix included a similarly large refactor, they just used the "security screen type" as a newly introduced parameter instead of something unique to the screen instance.

I do agree that in the real world, sometimes you have to settle for a less-than-ideal solution. I hope my post reads less like "those people are idiots", which was not my intent, but more like: this specific fix isn't ideal, and knowing this type of code is live in a device doesn't fill me with confidence, even if I can understand reasons for why it was done that way.

[btown]

Right? This was absolutely the "right" level of refactor for a hotfix, as the full refactor would introduce much more state management that could itself introduce bugs. And especially if behind the scenes there was a detailed audit of what things can currently access the current security screen, it would be fine for now.

But I sincerely hope that in the postmortem, there would be a larger meta-discussion around code review practices and how something like this "global dismiss" became part of the API surface to begin with, and a sincere prioritization of a larger review within the backlog. Though with everyone on edge at this time in big tech, I doubt that ends up happening :(

[bragr]

>Their fix included a similarly large refactor

Their change is hardly a big refactor. This includes all the new code, all the parameter changes everywhere the function is used, and two additional test cases. This is a tiny change.

>12 changed files with 102 additions and 26 deletions. [1]

https://github.com/aosp-mirror/platform_frameworks_base/comm...