Jasmine Tang

Now, I realize the title might be a bit heat-from-fire, fire-from-heat baiting :) I hope my mis-interested readers can still accept the bait and keep on reading :)

Bazel's transition, to me, is one of the most confusing docs in the world. This article is then an attepmt to try and demystify the operation in bazel.

Let's talk about configuration first. When a target's being built by an invocation on the command line, it's in the "target" configuration. In addition, the tools that's run on the platform that's building said target are in the "exec" configuration. Now, these are bazel-defined configurations, and to influence these configurations, you can use cfg = exec or cfg = target to transition from one configuration to another.

Hence, the act of transitioning, or, to transition, is to accept the input configuration and output some other build configuration.

Now, it is bar none helpful if all we did all day is talk about transitioning between "exec" and "target". We can actually rely on user-defined transitions.

User-defined transitions

Users can define their transition by defining a transition function

your_transition = transition(
    implementation = transition_impl,
 
 
 
)
 

Ingoing versus outgoing

Now, suppose your rule is rl(input, output), if you call your rule such as