Update Go version to 1.20

[barrettj12]

Go 1.14 has been end-of-life for years, and now we're running into problems like not supporting io.ReadAll. Time to update it.

In our GitHub CI tests, I upgraded actions/setup-go to v3, and now we can get the Go version from go.mod.

[benhoyt]

Per discussion, I'm okay with this update (though I have a question below about testing on two versions). But it'd be good to run this past @niemeyer as there are other teams that use Pebble now.

[niemeyer]

Thanks or the PR, Jordan. A few questions below.

[jnsgruk]

@benhoyt @hpidcock do we need to revisit this, or should we just close the PR?

[barrettj12]

Considering we're now shipping Pebble binaries (which are built with Go 1.20 as well), I don't know if we still need to support old Go versions?

[jnsgruk]

I spoke with @niemeyer today about this PR. It looks like Juju is now building using 1.20 across all supported branches, so we shouldn't have a problem moving to 1.20 here - I've updated the branch to include the bump to 1.20 where it was previously 1.19 and rebased on master.

@barrettj12 Before we consider making a change like this, we need to make sure that we're not going to affect any of our downstream consumers, which I think is the piece that was missing from your original PR :)

So...

• @flotter - can you confirm that this isn't going to impact your projects?
• @cjdcordeiro or @tigarmo - can you confirm that we're okay from a Rockcraft perspective? My guess is we're fine since Pebble is sourced from the published snap
• @hpidcock can you confirm we're good from a Juju perspective here?

If we're all good on those fronts, then we can go ahead and merge this -- we're fine to live with a deprecation warning if we need to - the features won't go away until go2 in all likelihood - and we can reassess at that point.

@hpidcock / @benhoyt - feel free to merge once we get confirmation from the others :)

[flotter]

@jnsgruk @barrettj12 Our project selected Go 1.20 as the starting point, so I think this proposal aligns well with us. Thanks for checking.

Should we update the PR title to reflect Go 1.20 (I assume this is the latest proposal)?

[tigarmo]

@jnsgruk as far as Rockcraft goes, we should be fine - even before we moved to the snap we used go 1.19 to build locally.

[cjdcordeiro]

@jnsgruk as far as Rockcraft goes, we should be fine - even before we moved to the snap we used go 1.19 to build locally.

💯

https://github.com/canonical/rockcraft/blob/d81be44494fea5fd16f4d4baf2a6cafa66bfb083/rockcraft/pebble.py#L32-L36

---

The latest stable Go snap is also 1.20, which is what we use to build the Pebble snap:

pebble/snap/snapcraft.yaml

Lines 35 to 36 in 0906a62

	build-snaps:
	- go

Maybe it'll be good to pinpoint that Go snap version in Pebble's snapcraft.yaml -> #271

[sparkiegeek]

I spoke with @niemeyer today about this PR. It looks like Juju is now building using 1.20 across all supported branches, so we shouldn't have a problem moving to 1.20 here - I've updated the branch to include the bump to 1.20 where it was previously 1.19 and rebased on master.

@barrettj12 Before we consider making a change like this, we need to make sure that we're not going to affect any of our downstream consumers, which I think is the piece that was missing from your original PR :)

So...

* @flotter - can you confirm that this isn't going to impact your projects?

* @cjdcordeiro or @tigarmo - can you confirm that we're okay from a Rockcraft perspective? My guess is we're fine since Pebble is sourced from the published snap

* @hpidcock can you confirm we're good from a Juju perspective here?

If we're all good on those fronts, then we can go ahead and merge this -- we're fine to live with a deprecation warning if we need to - the features won't go away until go2 in all likelihood - and we can reassess at that point.

@hpidcock / @benhoyt - feel free to merge once we get confirmation from the others :)

You forgot about us!

We use golang-go package from Jammy to build Pebble, which is 1.18 AFAICS

[jnsgruk]

Sorry about that @sparkiegeek! You'll be on the list for next time, for sure.

Adam and I spoke briefly offline, and they can quite easily move to using the go snap for building, so this won't materialise as much of a problem.