Actually I see we ran vet only with -vettool which replaces the default checks:

The -vettool=prog flag selects a different analysis tool with alternative
or additional checks.

So I've added a command that now runs plain vet as part of lint-go-vet.

Need to check whether gitea-vit replaces or extends the builtin checks. If the latter, we only need to run go vet once.

On a lower powered machine this definitely helps.

Yeah I agree, go test in general is a massively slow test runner compared to others like vitest, I guess it's partly because it runs vet. But let's wait on the investigation above until I'm certain whether one or two vet calls are needed.

Before this can merge, I need to know whether gitea-vet extends or replaces the default vet checks, e.g. whether it's sufficient to only run vet once with the -vettool option.

@techknowlogick or @yp05327 do you perhaps know?

Actually I was able to confirm that -vettool=gitea-vet replaces the default vet checks, by adding version = version to main.go:

$ go vet ./...
# code.gitea.io/gitea
# [code.gitea.io/gitea]
./main.go:52:2: self-assignment of version to version
$ go vet -vettool=gitea-vet ./...
$

So it needs to run it twice and the current approach is fine.

Actually I found out we were still running it twice because regular go vet also runs as part of golangci-lint:

So I removed the explicit call to go vet again and renamed the make target to lint-go-gitea-vet to accurately reflect what it does (it has nothing to do with default go vet).

Also I've moved the flag to GOTESTFLAGS, making it being used in all go test commands.