Conversation
brandur
force-pushed
the
brandur-go-1-25
branch
3 times, most recently
from
7bc728b to
2030523
Compare
| # Also run a couple previous Go versions against the latest Postgres version: | ||
| - go-version: "1.24" | ||
| postgres-version: 17 | ||
| - go-version: "1.23" | ||
| postgres-version: 17 |
There was a problem hiding this comment.
Thoughts on removing the v1.23 from here? I don't think it adds much tbh.
There was a problem hiding this comment.
I'm kindaa thinking that we should lave that in for the time being. It's still the version of Go in all the go.mods, and we should be testing against what our ostensible lowest version is.
We could bump all the go.mod's to 1.24, but I don't think we should — only the last two versions of Go are supported in practice, but a lot of people get stuck behind, at least for a while. If there's a feature we really want in the latest or something we really wanted to drop, I can see the logic for an aggressive push for new Go versions, but given there isn't this time around ... may as well leave that lower version constraint as low as is convenient. At least for now.
Thoughts?
Start testing against Go 1.25. I didn't change the `go.mod` version. I figure it's probably fine to leave that as low as is convenient for us to maintain as a courtesy to those at companies where it's harder to upgrade.
brandur
force-pushed
the
brandur-go-1-25
branch
from
2030523 to
3ca1ba1
Compare
| - go-version: "1.25" | ||
| postgres-version: 15 | ||
| - go-version: "1.25" | ||
| postgres-version: 14 |
There was a problem hiding this comment.
I ended up flattening this matrix out because I don't think it was working as intended before, or if it was, it was only doing so accidentally.
GitHub Actions matrixes operation quite unintuitively. I read those docs three times and still don't 100% understand it, but if I added only 1.24/17, it would buit it on all of Postgres 17/16/15/14, which is obviously not what was intended. Found it easier just to unroll the matrix like this which isn't too many more LOCs anwyay.
There was a problem hiding this comment.
I think the intention was to test the latest Go version against all recent supported Postgres versions, and then separately test each recent Go version against the latest Postgres. This was on the assumption that cross PG-Go version incompatibilities are super unlikely, whereas it's more likely that some dependency doesn't work as well in a particular Go version or against a certain PG version.
There was a problem hiding this comment.
Oh yeah, I totally understood that part of it. What I meant is just that the matrix functionality is kind of weirder than you'd expect. If you wanna play with it a bit, I'm sure there must be another way to get it working as intended, but my brain was too strained to figure it out.
| - go-version: "1.25" | ||
| postgres-version: 15 | ||
| - go-version: "1.25" | ||
| postgres-version: 14 |
There was a problem hiding this comment.
I think the intention was to test the latest Go version against all recent supported Postgres versions, and then separately test each recent Go version against the latest Postgres. This was on the assumption that cross PG-Go version incompatibilities are super unlikely, whereas it's more likely that some dependency doesn't work as well in a particular Go version or against a certain PG version.
|
Thanks! |
2 participants
Start testing against Go 1.25. I didn't change the
go.modversion. Ifigure it's probably fine to leave that as low as is convenient for us
to maintain as a courtesy to those at companies where it's harder to
upgrade.