Stop using packman for Ruby gRPC client generation

[googleapis]

Stop using packman for Ruby gRPC client generation

[jmuk]

@landrito, this is the one we chatted.

[landrito]

Gotcha, I'll look into it!

[garrettjonesgoogle]

I did want to note that the "similar to Java" comment is not correct - Java was converted to use packman (opposite conversion as proposed here for Ruby).

At a higher level, I'm not a huge fan of this mix of doing some generation in packman and some in artman. It seems preferable to make only one component responsible.

[jmuk]

Hmmm, I didn't know that, thank you for the info.

Then, I am quite unsure what needs to be done here... possibly:

• fix packman templates -- remove README or gemspec files. Then purely run protoc.
• modify artman task to copy the packman results to the final_repo_dir
  ?

[garrettjonesgoogle]

I think it could make sense for artman to copy the packman results to final_repo_dir.

[landrito]

For clarifications sake, we are copying rather than generating directly to the final_repo_dir to avoid the case where generation fails partway through only generating some of the files?

[garrettjonesgoogle]

Yes, we always generate directly into a temp dir under output/.

[jmuk]

Let me reopen this, I think the original description still makes sense.

The justification is that we actually don't need package definition files for gRPC clients, because we don't create an independent package for gRPC client. Instead Gapic client package should contain .rb files for both Gapic and gRPC clients. Now it's better for artman to invoke proto-compiler rather.