chore: create GitHub action to update the canary branch

[targos]

No description provided.

[targos]

I created an empty main branch.
Here's a run on my fork: https://github.com/targos/node-v8/runs/1107906165?check_suite_focus=true

Things to note:

• We need to find a way to notify interested individuals when there's a failure, because I think errors in scheduled actions do not generate GitHub notifications
• Fetching V8 in node-core-utils takes a lot of time. Since I already clone the lkgr branch in the action, that step shouldn't be necessary and we can probably enhance ncu for that

[targos]

/cc @mmarchini

[mmarchini]

We need to find a way to notify interested individuals when there's a failure, because I think errors in scheduled actions do not generate GitHub notifications

It generates for the committer and for the author (maybe also for the pusher). Are you on the OpenJS Foundation slack? We could have a nodejs-v8-canary channel and then send the notifications there. Otherwise there's probably a SMTP email notifier Action we could use (edit: https://github.com/marketplace/actions/send-email).

[mmarchini]

Fetching V8 in node-core-utils takes a lot of time. Since I already clone the lkgr branch in the action, that step shouldn't be necessary and we can probably enhance ncu for that

It doesn't respect the --base-dir option? Another option is to use a cache instead of checking out the repository (and let ncu check out like it does today, only the first run will take time to fetch V8).

[targos]

It doesn't respect the --base-dir option?

It does respect that option. The issue is that it fetches all branches with git fetch origin. That's not a problem with Jenkins because the workspaces are kept between runs so only the first run on a new machine would be slow.

[targos]

Using a Slack channel SGTM.

[mmarchini]

It does respect that option. The issue is that it fetches all branches with git fetch origin. That's not a problem with Jenkins because the workspaces are kept between runs so only the first run on a new machine would be slow.

Ah, that sounds like a good use for cache then

[mmarchini]

For Slack there are many actions out there: https://github.com/marketplace?type=actions&query=slack. I have tested https://github.com/marketplace/actions/slack-notify but haven't tested it to "notify on failure". Other Actions look easier to use to notify on failure (or after every run, if that's what we want).

[targos]

I created the #nodejs-v8-canary channel on Slack

[cclauss]

Sorry @targos I did not mean to undraft this!

[mmarchini]

@targos fyi I added SLACK_WEBHOOK Secret to the repo, so we can use it to notify the channel when sync fails

[targos]

Result:

[targos]

Last try: https://github.com/targos/node-v8/runs/1800903471?check_suite_focus=true

Note that we apparently need a custom GitHub token with "workflows" permission, otherwise the push to canary fails if a workflow was updated in nodejs/node.
Apart from that, I think it's ready. The build took around 1hour. I don't know if we can optimize that somehow but the good news is that it's much faster than the Jenkins job!

[targos]

@mmarchini @gengjiawen would you like to have a new look?

[gengjiawen]

LGTM

[targos]

Starting with a manual run: https://github.com/nodejs/node-v8/actions/runs/737865140