Ensure commit sha is always passed to Message

[tobischo]

In #18 the issue is raised the gitlab API cannot be correctly addressed with the gem in the state it is right now.

This seems to be happening because the commit sha is set to nil in any case

pronto-eslint/lib/pronto/eslint.rb

Line 41 in 46848c9

Message.new(path, line, level, offence['message'], nil, self.class)

Lines would provide the commit sha correctly for the change and the error caused by it, but those are only accessed for non fatal errors.

The commit sha is also stored at the top level @patches variable which can provide it for the commit that was last pushed/on which the check is running. In for fatal errors it is possible then to use this commit message to attach the messages

[ashkulz]

I'm not sure about this change, as I looked at pronto-rubocop and pronto-haml and it has the same logic -- so if it's a bug, it's more wide-spread than this linter. Can you reproduce it with other linters?

[tobischo]

You are right, for pronto-rubocop it is not there and it currently is also not causing issues for us.

I know that I added it for pronto-golang

But let's go through the code:

The runners are executed here:
https://github.com/prontolabs/pronto/blob/999d0b92fc2f916cb1c77fa4f792c31428a0b560/lib/pronto.rb#L66
https://github.com/prontolabs/pronto/blob/999d0b92fc2f916cb1c77fa4f792c31428a0b560/lib/pronto/runners.rb#L20

We are using the -f gitlab option, so
https://github.com/prontolabs/pronto/blob/999d0b92fc2f916cb1c77fa4f792c31428a0b560/lib/pronto/formatter/formatter.rb#L19

And the gitlab formatter
https://github.com/prontolabs/pronto/blob/999d0b92fc2f916cb1c77fa4f792c31428a0b560/lib/pronto/formatter/gitlab_formatter.rb#L3
is coming from the commit formatter
https://github.com/prontolabs/pronto/blob/999d0b92fc2f916cb1c77fa4f792c31428a0b560/lib/pronto/formatter/commit_formatter.rb#L10
which is coming from
https://github.com/prontolabs/pronto/blob/999d0b92fc2f916cb1c77fa4f792c31428a0b560/lib/pronto/formatter/git_formatter.rb#L3

And the messages are mapped to comments here:
https://github.com/prontolabs/pronto/blob/999d0b92fc2f916cb1c77fa4f792c31428a0b560/lib/pronto/formatter/git_formatter.rb#L74

and the commit_sha is coming from the message when mapping it:
https://github.com/prontolabs/pronto/blob/999d0b92fc2f916cb1c77fa4f792c31428a0b560/lib/pronto/formatter/git_formatter.rb#L65

And the commit_sha in the message, if left to nil, needs to come from the line:
https://github.com/prontolabs/pronto/blob/999d0b92fc2f916cb1c77fa4f792c31428a0b560/lib/pronto/message.rb#L18

For the fatal errors it therefore needs to get the commit_sha but it does not need it for other types of errors, so we could change it to not pass in the commit_sha for the places where we have a line object.
That's probably also why we do not need it for pronto-rubocop and pronto-haml.

[ashkulz]

Thanks for the detailed analysis, @tobischo! I'll try to take a look at it over the weekend 👍

[tobischo]

Hey @ashkulz
did you maybe have some time in between to take a look at this change?

[istisiki]

Can confirm this is still a problem.
I've added pronto-eslint to our Github Actions recently and it's gives the same error.

/opt/hostedtoolcache/Ruby/2.7.3/x64/lib/ruby/gems/2.7.0/gems/octokit-4.21.0/lib/octokit/response/raise_error.rb:14:in `on_complete': POST https://api.github.com/repos/ORG_NAME/REPO_NAME/commits//comments: 404 - Not Found // See: https://docs.github.com/rest (Octokit::NotFound)
	from /opt/hostedtoolcache/Ruby/2.7.3/x64/lib/ruby/gems/2.7.0/gems/faraday-1.8.0/lib/faraday/middleware.rb:19:in `block in call'
	from /opt/hostedtoolcache/Ruby/2.7.3/x64/lib/ruby/gems/2.7.0/gems/faraday-1.8.0/lib/faraday/response.rb:59:in `on_complete'
	from /opt/hostedtoolcache/Ruby/2.7.3/x64/lib/ruby/gems/2.7.0/gems/faraday-1.8.0/lib/faraday/middleware.rb:18:in `call'
	from /opt/hostedtoolcache/Ruby/2.7.3/x64/lib/ruby/gems/2.7.0/gems/octokit-4.21.0/lib/octokit/middleware/follow_redirects.rb:73:in 
	...

The runners we are using are:

• pronto-rubocop
• pronto-eslint

Removing pronto-eslint resolves the issue for me but I would really want to have that in our CI.

[leonardoalemax]

this will solve my problem.

[ashkulz]

Sorry, this got lost in my to do list. I'll take a look at it in the evening.

[tobischo]

@ashkulz Gentle reminder for this 🙂

[ashkulz]

@tobischo thanks for the reminder, I'll review this on the weekend. Sorry for the delay 🙈

[tobischo]

Maybe this weekend? 😄

[tobischo]

Or this one? 😉

[ashkulz]

Sorry, there's been a lot of sickness in sequence in the family. I'll do it this weekend, I promise 🙈

[tobischo]

No need to explain or apologize.
As far as I am concerned I am basically asking you to spend your spare time for free on something for me.

Therefore, no rush, just gentle reminders as I also know how easily something like this can be forgotten about

[tobischo]

Alright, I resolved the conflict that got in from another change in the meantime.

Gentle reminder for a review as well