Currently, automated commits are generated, tested, and pushed by the same script.
This has obvious separation of responsibility concerns. We have a land bot that can land
human-authored patches, but it makes heavy assumptions about the code review process.
This patch is the first step to giving automated patch support to the land bot. Since automated
patches do not go through the code review flow, we need some other way to keep track of these
patches. We can leverage Differential's existing "Diff" framework (not to be confused with the
colloqual term "diff" we often use when we are actually talking about Revisions).
A Differential Diff has a patch, a unique ID, and an author. Differential acts as the
permissioned central repository for these patches to ensure the author information can
be trusted. The unique ID can be used for automated patch generators to call the land bot.
The land bot fetches the Diff using the ID, validates the author information, and then lands
This script is the author validation piece. It will soon be used by other landbot machinary
to complete the process described above.