RFR: Section on PRs

[Pavel Rappo]

On Wed, 6 Apr 2022 17:15:21 GMT, Pavel Rappo <prappo at openjdk.org> wrote:

>> Some tips for creating pull requests.
>
> 1. Do we need to say anything about testing? For example, we might want to say that each commit into a PR branch should compile and pass tests.
> 2. Do we need to add more rationale why PR should be generally kept up to date with respect to mainline, and especially just before they are integrated?

> @pavelrappo
> 
> 1. I think the section on testing should cover these parts. Me may want to say something about the PR specific testing though, like GHA. Could you please provide an example of what you want to add?

I don't have any examples at the moment.

> 2. Maybe :-)  Any specific phrasing you would like to see?

I cannot think of any specific phrasing; I only have an idea of what might need to be conveyed.

Before JEP 369 a committer was guaranteed a particular state of the mainline. A commit was atomic. A commit would either fail or succeed; there were no other options. If a commit failed, the committer would pull, merge and re-attempt to commit. Rinse, repeat.

After JEP 369 a committer generally operates in what one might call the "I'm Feeling Lucky" mode. Unless there are conflicts, a committer is guaranteed a mainline commit derived from their PR; a commit, but not the state: the commit's parent is unknown until after integration.

Along with the above information, that guide section could also provide:

  * a link to atomic integration, i.e. `/integrate <hash>`
  * a command similar to that used by Skara, to "dry run" the integration locally

-------------

PR: https://git.openjdk.java.net/guide/pull/79