feat: add other-vcs-status-names-to-ignore

[bakayolo]

what

other-vcs-status-names-to-ignore is a comma-separated list of atlantis vcs status names that will be ignored when deciding if the PR is mergeable or not. This allows us to set mergeable with multiple atlantis running with multiple vcs status names for the same repository.

why

This PR is a proposal to close #2848

tests

I have only run make test but planning on testing in my own environment if the proposal is accepted.
Tested with make test and on my environment with bappr/atlantis:cud-ignore-vcs-status-names-ca1171e0dd174edbbd63f8f6ae560caa6af4c0c7-0.

references

closes #2848

[jamengual]

interesting proposition.
one quick thing I will say that instead of adding a flag for this it will be much better to add it to the server side repo config with something like :

vcs-status-to-ignore:
  - name1
  - name2
  - etc

[bakayolo]

@jamengual thanks for the feedback, why do you think it is much better? I would actually imagine that having other-vcs-status-names-to-ignore close to vcs-status-name would be less confusing no?

[jamengual]

Because we have a trillion flags, and we want to avoid having more that is the main reason, plus if is moved there then you can have different repos regexes with different statuses.

[bakayolo]

@jamengual I've been looking at the code and I believe it's gonna be pretty complex to do

plus if is moved there then you can have different repos regexes with different statuses.

because the FetchPullStatus is not called at the repo config level. It's basically a sort of boolean ("has been approved" and "is mergeable", no conditions)

Also, I am struggling to understand the benefit of having different repos regexes with different statuses. Simply put, #2848 is basically a bug where multiple atlantis instances with different vcs status name conflict with each other and forbid the apply when mergeable is set. Which means that if we do not make atlantis aware of ALL the others atlantis running for the repo, they will continue conflicting with each others. So I think that config should be global.

Wdyt?

[jamengual]

Sorry, this is a standard ask we ask everyone.
I did not look at the code to see if the context will be present at that point in the execution for you to do it.
As you mentioned, it could get pretty complex, so keeping the flag then is better.

[anryko]

Small nitpick, other in the name seems redundant. The vcs-status-names-to-ignore would be sufficient, but considering the pattern in other arguments naming (hide-, silence-, include-, allow-), the name should resemble ingnore-vsc-status-names.

[bakayolo]

FYI, I've tested this change on my environment (Dev, Staging, prod) with bappr/atlantis:cud-ignore-vcs-status-names-ca1171e0dd174edbbd63f8f6ae560caa6af4c0c7-0.