jlord / reporobot Goto Github PK

Dear me, update server with the latest v of maxogden/taco. x, me

Hi,

I wasn't sure whether this issue should go on Reporobot, or on the Patchwork repo - it doesn't look like Reporobot is merging all PRs that go through to Patchwork at the moment. Looking through the list of PRs on that repo, there are a number that have been open for some time, but that have not been commented on by Reporobot (suggesting that there are no issues with the PR.) It does, however, look like merges are still taking place some of the time (one within the last hour in fact), so it doesn't appear to be a blanket stoppage.

Here are a couple of sample PRs that have been opened within the last 24 hours, but not commented on or merged:
jlord/patchwork#18888
jlord/patchwork#18878
jlord/patchwork#18877

It looks as though the oldest un-merged, un-commented PR dates back to 25th Jan 2017.

It's good to do.

Currently @reporobot server assumes the PR will be mergeable in the near future so it tries every 3 seconds. But if for some reason it never becomes mergeable @reporobot will be stuck in a loop and API limits will be reached and it's not good!

Am adding a limit to how many times @reporobot will try before he stops 👍

The PR that put the server in a loop: jlord/patchwork#797
Noticed via email from user in: jlord/patchwork#808

Hi,

I would like to know what is/are the license/s applied to the repository?
Please can you specify it/them? it would be very useful if somebody decide to use it (e.g. gitevents/core#20).

Many thanks.

I noticed that @reporobot was no longer writing to contributors.json. Without checking, I decided to check a hunch that the GitHub API wasn't letting it write because the file was now too large (over 1MB). I created an archive and perhaps someone will try and complete it soon and we'll see what happens...otherwise 🔜 I'll go digging around for what's up.

and make them a json file you require so all the text doesn't clutter up the js file ✨

It's not working!

This directory is getting too big and emptying it will make it easier to navigate, smaller and possibly help out with that random capitalization bug.

The new Repository Invites change to GitHub.com breaks @reporobot where it gets added as a collaborate then draws art to the users repo.

This potentially could be fixed by using the current system of parsing the username from the email subject and then going to the predicable invite site, parsing that body for the ID and the accept button. Make a request using All the Right Things.

Also if there was a setting on GitHub to auto accept invites that would be great 👍

🔗 jlord/git-it-electron#137 + jlord/patchwork#11760

See current open pull requests lists on @jlord/patchwork.

• Forks are getting thousands of commits along with them..
• @reporobot isn't responding the way he should to some.
• Check that he is handling deleted files correctly (not counting them in total)

So if I change the template of the page I don't have to wait for someone to complete Git-it to see it.

A tracking issue for building a suite of tests for @reporobot 🐱 🎇

Currently only I can run these tests as I have the tokens for @reporobot's account 😁

Test Scenarios

• Correct Pull Request
• Incorrect PR: Wrong branch name #22
• Incorrect PR: File is empty
• Incorrect PR: Wrong Filename
• Incorrect PR: Too many files #23

Other bits

• scripts
  • run suite
  • run each test

Just in case people have to reopen PRs, make sure everything still works like it should and the JSON you get and the variable sand what not are the same...

• Unmerged PR
• Unmerge PR type
• If unmerged PR type is too many files, how many files (a range)

• don't commit twice

Because sometimes these are people doing Git-it, but they're not realizing (I guess) that their branch name is wrong so @reporobot isn't looking at their stuff at all.

He should leave a comment saying something like "Hey! Thanks for the PR, the branch name isn't ... so I'm not checking it for Git-it. If you meant it to be...."

That's not good, but I don't like any others as much as I like isometric2...

Add the total number of users that have completed it in the template, cause that would be fun!

It actually kills the server. I need to go in and add a check for that scenario and have @reporobot leave a comment on that PR.

PRs are already closed by RR by the time git-it goes to check that the user submitted one, so check the closed PRs to verify.

I noticed that ~2 months ago @reporobot stopped drawing 🎨 and I have no idea why 🎉

Cross ref: 🔗 jlord/patchwork#16453

Hi jlord.

I'm a big fan of git-it/nodeschool. I recently passed its_a_small_world, but I managed to pass it with the wrong collaborator (I had typoed and put Reprobot as a collaborator instead - https://github.com/ReproBot). I removed Reprobot and ran the verifier again by mistake (I'm not the slickest in terminal), this time without any collaborator on my patchwork repo, yet still passed.

I noticed that in reprobot/collaborate.js, we are checking permissions.push against the username supplied by its_a_small_world/verify.js:

 var repo = github.getRepo(username, 'patchwork')
 var collab = false

 repo.show(function(err, repo) {
 if (err) return callback(err)

var permissions = repo.permissions
if (permissions.push) {
  collab = true
  callback(null, collab)
} else {
  collab = false
  callback(null, collab)
}
})

Is it possible that the username from its_a_small_world/verify.js:

exec('git config user.username', function(err, stdout, stdrr) {
var username = stdout.trim()
collaborating(username)
})

will always result in collab = true, as the repo we are checking is a fork (and therefore I assume I have full permissions)?

If this is the case, then it might be hard to catch if you have made a mistake when completing the exercise. If you are a bit of a crappy typist (like me), you will not trigger reporobot's confirmation email, but still pass and possibly give somebody else push permission to your repo (and maybe end up a git pull away from being totally pwned 😱).

Is there a way we can check for repo.collaborators instead? I do not know enough about the output of git.show to tell.

Anyway, sorry about the lengthy message - let me know if I'm barking up the wrong tree.

All the best,

Mike :)

I keep forgetting to go in and shorten it (right now the server log spits out the whole thing). Hopefully this reminds me.

Newest on top.