I've been using bazel+REAPI for remote execution of many different kinds of actions and running into some situations that require lots of hacks. Basically, I've run into config issue where I need to add some execution properties but they don't seem to quite fit platform idioms or work with how some features work in Bazel. However, the properties I'd like to add already exist in, execution_requirements.The existing requires-* block-* attributes already do what I'd like them to do locally on the actions but aren't available in the REAPI "platform" fields on action/command.
On GitHub it seems like there are 2 examples of execution_requirments that I'd like to have working on exec_properties.
This would also allow people to feed properties from the analysis phase into the REAPI execute calls. Is there a good reason we shouldn't have a feature that opts some execution_requirments into being part of the REAPI to generally solve this? I'm happy to write up a proposal to at least handle the use cases listed in this document, or a more general way to handle execution_requirments