CLI wrapper to rebase, pre-fill and run CI on PRs

[silverwind]

Spawned off #324 (comment)

The basic idea would be a CLI script that does:

1. Take a PR number argument
2. Pull patches from Github
3. Aks y/N if commits shall be squashed
4. Rebase commits and apply metadata (use GitHub API to parse for LGTMs)
5. Preview commit messages and ask y/N if they shall be further edited in $EDITOR
6. Start a Jenkins job, possibly authorized through a user's SSH key or other token
7. Sleep the process until Jenkins is complete
8. If CI is green, ask y/N whether the commits shall be landed

The only step on which I'm unsure on how to do would be step 5. Can anyone think of a git workflow to add metadata to existing commits without rebase --interactive?

cc: @jbergstroem

[Trott]

Maybe add a commit message linting/validation step that makes sure the final message:

• has a first line that is not empty and is not longer than 50 chars
• has a first line that matches a regular expression along the lines of this (untested): /^[a-z,]+: \w.*/
• has a blank second line
• has a non-blank third line
• has a blank line before the metadata
• has metadata consistent with what it knows about the PR (e.g., PR-URL exists and is correct)
• has at least one Reviewed-by metadata line
• individuals in Reviewed-by metadata are in fact collaborators (basically compare against a whitelist of people who can give a LGTM that counts)
• all non-metadata lines are shorter than 72 chars

[jbergstroem]

Excellent list by @Trott. I'd like to add:

• metadata ordering (suggesting: PR-URL -> Fixes -> Refs -> Reviewed-By)

It'd be really nice to have a LGTM validation/"check" (ok/not ok to land); but I think it's too hard to make realistic LGTM counts seeing how new commits could introduce new code.

[refack]

Cross ref: #705

There's an issue of racing with humans/concurrent jobs. That's where the concept of a queue comes into play. (#324 (comment))

[Trott]

Closing due to long period of inactivity. Feel free to re-open if this is a thing. I'm just trying to close stuff that has been ignored for sufficiently long that it seems likely it's not something we're going to get to.