#4990 made the definition of RawModuleDefV10 invalid, so that generating it to TypeScript or C++ would result in code that didn't compile. We merged that PR without knowing, as we didn't actually re-run codegen during the PR; we planned to do so in a follow-up. This became a problem in #4636 and friends, when we attempted to make a separate change to RawModuleDefV10 in all of our supported languages. We reverted (well, as of writing, will revert) the broken change in #5098.
We should have a test which runs in CI which checks that the RawModuleDefV10 changes are up to date in all languages. Because we already check that the module libraries compile in all languages, this will mean that PRs like #4990 cannot merge. IIRC, we already have such a test for the WebSocket schema.
#4990 made the definition of
RawModuleDefV10invalid, so that generating it to TypeScript or C++ would result in code that didn't compile. We merged that PR without knowing, as we didn't actually re-run codegen during the PR; we planned to do so in a follow-up. This became a problem in #4636 and friends, when we attempted to make a separate change toRawModuleDefV10in all of our supported languages. We reverted (well, as of writing, will revert) the broken change in #5098.We should have a test which runs in CI which checks that the
RawModuleDefV10changes are up to date in all languages. Because we already check that the module libraries compile in all languages, this will mean that PRs like #4990 cannot merge. IIRC, we already have such a test for the WebSocket schema.