Inconsistent setField method in generated examples about ts-protoc-gen HOT 4 CLOSED

2 seems to be a problem as protoc is usually installed globally. A server developed on the same machine may require a newer version.

1 is nice, as it serves as a bit of a unit test against unwanted regressions, but I assume most changes to the code would cause output to differ, and CI tests that fail always are a bit annoying.

You could also add the version to Contributing.md and manually note unwarranted changes to the generated output when reviewing PRs

from ts-protoc-gen.

Oh yes, both of these are flawed, probably not a great idea to respond to GitHub issues at 7am 😂

You're right, updating the contribution guide is an obvious first step, however another spin on option #2 may be to download protoc as part of generate.sh as it is a standalone binary

from ts-protoc-gen.

I'm leaning towards a pinned version of protoc for the repository. This makes the experience consistent for all contributors. Furthermore, an upgrade is only a PR away which helps us track the protoc target we're building against.

I can see three ways to get this done:

1. Docker container with protoc
2. Use grpc-tools which bundles protoc but hasn't been updated in a while
3. Use Bazel to generate the examples. This would require contributors to install bazel.

I like 1 and 3 or going with downloading protoc

from ts-protoc-gen.

Ah yes, sorry about that. I think we need to enforce a given version of protoc for consistency. Two options I see are:

1. Add a CI build step which regenerates the example protos and then checks for changes, if there are any it will fail the CI build with a descriptive error message

2. Add a local build step which checks the users locally installed version of or protoc.

However I'm welcome to other suggestions 😀

from ts-protoc-gen.