Conversation

Replying to
Once toolchain hardening is added back for the OS, it will need to be ported over the Chromium again. Separate toolchain versions for AOSP and Chromium is an annoyance but projects of this scale tend to hit lots of compiler bugs and it's best to stick with what's tested upstream.
1
I've noticed that traditional Linux distributions (Debian, Fedora, Arch Linux) build Chromium with security features like CFI crippled or even completely disabled due to using GCC and/or using component builds which are a development feature, especially with system libraries.
1
2
Nearly every non-Google Chromium build is also being deployed with messed up field trial configuration. I think it's best to disable field trials, but they all seem to ship with either seed-based field trials or the testing configuration aimed at maximizing testing code coverage.
1
1
I don't think it makes sense to ship intentionally maximized experimental attack surface / bugs and many annoyances never shipped for most Chrome users. Having a random set of experimental features isn't much better. For Chrome, Google controls it from a service for A/B testing.
1
I can also tell Chromium forks don't understand what they're doing from the fact that they leave the field trials enabled, notice that changing the default values of features doesn't work (since they are overriding it with field trials) and then hard-wire overrides in the code.
1
4