mirror of https://github.com/pulumi/pulumi.git
7b563a49dc
During the execution of a normal Pulumi program, the various components (engine, language hosts, and providers) communicate over gRPC. When running conformance tests, we want to simulate this environment as closely as possible. To this end, even though we have mocked in-memory providers for the tests, we wrap them in gRPC servers and the test host communicates with them over gRPC, just as the engine would. Moreover, the language host under test communicates with the test host over gRPC, just as it would with a real engine. Presently, however, there is one missing piece -- when a language host asks the engine to load a schema, this call is typically answered by a delegated gRPC to the provider offering that schema. Presently, the conformance test does not do this, instead loading the schema using the in-memory provider objects it already possesses. This commit fixes that to match the "real-world" behaviour, having the test host delegate to its already-gRPC-wrapped providers instead. To support this new behaviour the test host needs to learn how to load unversioned providers as well, but this is relatively simple and shouldn't impact any of the tests we run presently. |
||
---|---|---|
.. | ||
pulumi-test-language |