cmd/go/internal/modfetch: TestCodeRepo/parallel/gopkg.in_natefinch_lumberjack.v2/v2.1 failing at head

[bcmills]

Likely another case of #28856, masked by #31263.

[…]
--- FAIL: TestCodeRepo (18.60s)
    --- FAIL: TestCodeRepo/parallel (0.00s)
        --- FAIL: TestCodeRepo/parallel/gopkg.in_natefinch_lumberjack.v2/v2.1 (3.35s)
            coderepo_test.go:386: info.Version = "v2.1.1-0.20170531160350-a96e63847dc3", want "v2.0.0-20170531160350-a96e63847dc3"
[…]
FAIL
FAIL    cmd/go/internal/modfetch        19.588s

CC @jayconrod

[bcmills]

I don't actually see any tagging changes that explain the failure. Will need to bisect.

[thepudds]

coderepo_test.go:386: info.Version = "v2.1.1-0.20170531160350-a96e63847dc3", want "v2.0.0-20170531160350-a96e63847dc3"

Drive by comment: https://golang.org/cl/174061 or https://golang.org/cl/174206 spring to mind.

But is this an internal test that was passing previously with those changes already in place? (If so, then perhaps not them ;-)

That looks like a "want" non-pseudoversion is now a pseudoversion?

[bcmills]

The base for the want pseudo-version went from v2.0.0 to v2.1.1.

I'm pretty sure I saw a couple pseudo-version changes go by while I was out, but I'm pretty sure git bisect will find it either way. 🙂

[thepudds]

That commit seems to have an "almost semver" tag of v2.1:

https://github.com/natefinch/lumberjack/tree/v2.1

[thepudds]

go/src/cmd/go/internal/modfetch/codehost/git.go

Line 681 in 3b8c804

if semver.IsValid(semtag) {

I think semver.IsValid might accept something like v2.1 as valid?

Maybe that needs to be semtag == semver.Canonical(semtag), or something like that?

[bcmills]

Ooh, that's a good find. Thank you!

[bcmills]

git bisect points to CL 174061, which (as it turns out) is where that call to semver.IsValid was added.

[gopherbot]

Change https://golang.org/cl/176539 mentions this issue: cmd/go/internal/modfetch/codehost: ignore incomplete semver tags in RecentTag