build: bump oapi-codegen to 2.4.1

[lzap]

Since the project relaxed the hardcoded Go requirement, we can now update. I propose to keep the pin and thus to keep updating this generator manually. Or alternatively, we can just remove the hack completely.

Quite honestly, I am not huge fan of go generate and I would suggest to get rid of it completely calling oapi-codegen directly. This is what works for us quite nicely in provisioning app (this is from makefile but you get the idea):

• installation target
• generation target

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Files with missing lines	Coverage Δ
internal/clients/composer/openapi.v2.gen.go	21.45% <ø> (ø)
internal/v1/api.go	49.02% <ø> (ø)
internal/v1/server.go	73.52% <ø> (ø)

[mgold1234]

/lgtm

[ezr-ondrej]

I'd love to get rid of the hack.
To minimize the dependencies, my vote goes to not directly depend on it. If it's not used in production code, it should not be packaged IMO :)

[lzap]

I'd love to get rid of the hack.

Before we do that, I will say that the hack makes generated content more deterministic. Instead of people regenerating with effectively "random" version, we can upgrade this and create commits which will only show upgrade changes. These generator upgrades did introduce some bugs in provisioning in the past.

[lzap]

How about this?

[croissanne]

Overall I'm not sure what the real gain is here in changing up the tooling, I don't feel strongly either way however. But each change we'll have will need to update the individual go generate directives.

Anyway LGTM, but one comment on the use of which.

[croissanne]

Will which still work here? $TOOLS_PATH wouldn't be part of the user's path.

[lzap]

Oh good catch! Rebasing.

[lzap]

But each change we'll have will need to update the individual go generate directives.

I was testing this and as far as I can tell, the tool picks the tool which is installed. I hope this will work, but it is hard to test now that there is only one release that we can use due to Go stack incompatiblity. The next update will tell. But I manually removed the version from all comment headers and the shell script does not add it back now! Looks good.

I need to figure out how to silence shellcheck on this line tho, it wants me to enquote it:

TOOLS_PATH="$(realpath $(dirname $0)/../bin)

[ezr-ondrej]

/retest

[github-actions]

This PR is stale because it has been open 30 days with no activity. Remove "Stale" label or comment or this will be closed in 7 days.

[lzap]

Requesting re-review @mgold1234 and @croissanne thank you.