chore: Migrate from dep to go modules

[lcostea]

Migrate from dep to go modules:

• Removed the files used by dep to install dependencies
• Moved to using go.mod
• Updated documentation
• Updated ci jobs to stop using dep

[dthomson25]

This PR is great, but the code-gen script we use depended on the vendor folder no longer works with this change. When you run make codegen you get the following error:

mockery -dir ./metricproviders -name Provider -output ./metricproviders/mocks
Generating mock for: Provider in file: metricproviders/mocks/Provider.go
./hack/update-codegen.sh
./hack/update-codegen.sh: line 7: ../code-generator/generate-groups.sh: No such file or directory
make: *** [codegen] Error 1

To get around this, we could follow Linkerd's approach: https://github.com/linkerd/linkerd2/blob/master/bin/update-codegen.sh but I don't like how that still requires vendoring

[lcostea]

@dthomson25 Thanks, I started looking at other projects on how they handle the code-gen and mod vendor is used by all, I couldn't see an alternative. So I will give that a try.

[dthomson25]

Hey @lcostea, I wanted to check in to see if you had a chance to try to make those changes. If not worries! I know the end of the year can get busy at times!

[lcostea]

@dthomson25 Just got back today from the New Year's vacation, I will get back on track with everything in the following days.

[lcostea]

@dthomson25 Re the codegen I found a solution that doesn't need vendoring, inspired from https://github.com/weaveworks/flagger/blob/master/hack/update-codegen.sh

[dthomson25]

Awesome! That's great that we don't need to vendor the folder!

[lcostea]

@dthomson25 only now I saw that all my CircleCI pipelines have been failing.
Let me work on fixing those first, no need for any review yet

[dthomson25]

I took a look at your failed pipeline and I think the Prometheus library is getting upgraded and that version introduces a new method to the API interface which the mock doesn't implement.

[lcostea]

Thanks, passed that step by installing golangci-lint binary directly and not as a dependency. That go get was modifying the go.mod, while before with dep it didn't have any effect.
It looks ok now, I still have failures on modules installs and builds, but it looks like an OOM issue: I get a lot of /usr/local/go/pkg/tool/linux_amd64/link: signal: killed and locally it passes when I run the same commands in the circleci container.
So it looks ready to be reviewed

[codecov]

Codecov Report

Merging #331 into master will decrease coverage by 0.34%.
The diff coverage is n/a.

@@            Coverage Diff            @@
##           master    #331      +/-   ##
=========================================
- Coverage   84.04%   83.7%   -0.35%     
=========================================
  Files          70      70              
  Lines        6589    6555      -34     
=========================================
- Hits         5538    5487      -51     
- Misses        762     778      +16     
- Partials      289     290       +1

Impacted Files	Coverage Δ
utils/template/template.go	93.1% <0%> (-6.9%)	⬇️
rollout/bluegreen.go	76.95% <0%> (-5.33%)	⬇️
pkg/kubectl-argo-rollouts/info/replicaset_info.go	79.41% <0%> (-1.42%)	⬇️
experiments/experiment.go	93.3% <0%> (-0.34%)	⬇️
utils/experiment/experiment.go	93.04% <0%> (-0.29%)	⬇️
pkg/kubectl-argo-rollouts/cmd/get/get_rollout.go	83.76% <0%> (-0.21%)	⬇️
rollout/experiment.go	86.82% <0%> (-0.08%)	⬇️
rollout/pause.go	100% <0%> (ø)	⬆️
pkg/kubectl-argo-rollouts/info/info.go	100% <0%> (ø)	⬆️
experiments/replicaset.go	83.87% <0%> (+0.28%)	⬆️
... and 1 more

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 4bf6fa3...9c00e31. Read the comment docs.

[dthomson25]

Hey @lcostea, I pushed your branch to the main repo so the code could take advantage of the increased resources. Looks like it passed!

I noticed two issues when I ran through the Makefile:

1. I got line 10: realpath: command not found when I ran ./hack/update-codegen.sh. When I added the first answer from this stackoverflow, I was able to get it to work.

2. I think the version of the open-api gen is different as I see

modified:   pkg/apis/api-rules/violation_exceptions.list
modified:   pkg/apis/rollouts/v1alpha1/openapi_generated.go

Do you have this behavior too?

[dthomson25]

Hey @lcostea, I talked with @jessesuen, and we decided that we will move forward to the new openapi gen version and address the changes in another PR. So, we can merge this once we get the realpath fix in and fix this last merge conflict.

[lcostea]

@dthomson25 fixed the conflict
regarding realpath I see the command is available on the ci and I also checked alpine, debian, ubuntu, centos images and they all have it
On mac I see it is available on coreutils formula, so what about if I add this to the docs: you need to install coreutils with brew? I wouldn't want to use a dirty replacement, when the original one works on all major distros. Otherwise I will look for a standard replacement.

[jessesuen]

I wouldn't want to use a dirty replacement, when the original one works on all major distros. Otherwise I will look for a standard replacement.

DIR="$( cd "$( dirname "${BASH_SOURCE[0]}" )" >/dev/null 2>&1 && pwd )"

Is a pretty standard bash trick to get the real directory of a script -- insofar as I wouldn't even call it that dirty. I prefer using something like this which is platform agnostic, versus increasing the barrier to contribution.

[lcostea]

@jessesuen @dthomson25 using now what you suggested for getting the script directory

[dthomson25]

LGTM!

Unless you want to make any other changes, I'll merge it in!

[lcostea]

no other changes, please go ahead