go_vet reports unknown identifier in example code

[xthezealot]

For example, the github.com/gowww/compress package.

The package name compress is suffixed with _test inside the example_test.go file.
That's maybe why this go vet doesn't recognize the identifier reference.

But it should.
And locally, my go vet github.com/gowww/compress returns no error.

[shawnps]

@ArthurWhite that is strange behavior indeed.

i can't reproduce it on my local, but i see it failing on the go report card server. i opened a ticket with gometalinter, will see what the author says:

alecthomas/gometalinter#108

[shawnps]

Looks like it's not gometalinter but rather something strange going on with vet on the server...I'll try to investigate further soon

[shawnps]

Weird, go vet passes on command line on my local, but fails with the same errors when running Go Report Card locally.

[hectorj]

What go versions are you guys using?

I have a similar problem in one of my projects: it works with Go 1.5.X, but fails with Go 1.6.X

Edit:

Also see openshift/origin#4682 and golang/go#12636

[xthezealot]

Personally, I have the 1.6.2 darwin/amd64 installed with Homebrew.

What is even weirder is that Go Report Card returns a 100% success for another package (github.com/gowww/log) which follows exactly the same "template".

[xthezealot]

I just pushed a new commit and Go Report Card fails for this package too. 💣

	example_test.go
Previously	e31412b
Now	a2e97ea

[db47h]

Same for this package: https://goreportcard.com/report/github.com/db47h/cache

I use go 1.7 and go-tip as test code is not compatible with go <1.7,

[korthaj]

Maybe this is a caching issue? In my case, I got the false positive

radix/example_test.go
Line 8: error: ExampleSortSlice refers to unknown identifier: SortSlice (vet)

Perhaps go vet is looking at the new version of the file example_test.go and an older cached version of the file that actually contains SortSlice. I did, in fact, use goreportcard earlier, when the new function and the new test didn't yet exist.

[shawnps]

@korthaj unfortunately I don't believe it's a caching issue but rather an issue with go tool vet itself. I just ran it manually against your repo and I get the error:

$ go tool vet *.go
example_test.go:8: ExampleSortSlice refers to unknown identifier: SortSlice

gometalinter uses go tool vet rather than go vet, I believe for historical reasons. Someone brought up a similar issue here:

alecthomas/gometalinter#218

If I can find a way to have gometalinter use go vet instead of go tool vet, it might be fixable.

[xthezealot]

If I can find a way to have gometalinter use go vet instead of go tool vet, it might be fixable.

At first glance, just calling go vet instead of gometalinter at check/go_vet.go:21 should do the fix.
'Doesn't work like this?

[arsham]

I'm experiencing the same problem here
from this report. The identifier is defined and on local version it reports no issues.

[GoLangsam]

Same experience with my new repo github.com/GoLangsam/template using unnamed import aka import . "html/template" (as does standard html/template/template_test.go).

Local use of go vet and go tool vet remains silent.

Local use of gometalinter gives same "noise"!
Local use of goreportcard (under Windows - mind You :-) with latest gometalinter gives same "noise"!

After patching check/go_vet.go to:

return GoTool(g.Dir, g.Filenames, []string{"go", "tool", "vet"})

things are fine :-)

(except for small and unrelated Windows-issues - and that's a separate story!)

[shawnps]

I will take a look at this soon.

[GoLangsam]

Looking @ alecthomas/gometalinter#218 I feel that they'r pushing this off to the go tool chain. Thus, there is no hope for any quick solution soon.

Having browsed cmd/go/main.go and friends such as pkg.go I understand the trouble of the go team in supporting the variety of usages of their tools.

I also understand and experience: the best (if I may say so) way to point to packages is inside a workspace ($GOPATH set) and via import-path-of-package\....

Looking @ Your implementation, I see:

• You supply individual filenames ... (to be discussed separtely)
• nowhere do not (attempt to) touch $GOPATH

Thus: I just set my $GOPATH the _repo - and: All is fine and copathetic :-) :-) :-)

So, I'd humbly like to recommend: Stay far away from non-standard usage of package pointers when using go tools!

(Note: There's even a myriad of issues, blogs etc. around explaining very good reasone as of why some usage styles have to behave in ways unexpected to the naive and innocent user.)

[GoLangsam]

@shawnps Having found a.m. evidence, I'll have to read and think a little more in order to come up with a feasible and robust solution which suits Your fine product.

[GoLangsam]

@shawnps @hermanschaaf

here we go - Sorry for keeping You waiting :-)

[stevenh]

This is still an issue, I've raised it against go vet:
golang/go#22530

[stevenh]

As suggested by Russ Cox if the calls to go tool vet are replaced with calls to go vet examples pass.

[breml]

I have a similar problem with my package logstash-config.

Locally go vet ., go tool vet . as well as gometalinter . do not return an error. I use go version go1.10.1 linux/amd64. On goreportcard.com the following error is shown (https://goreportcard.com/report/github.com/breml/logstash-config):

logstash-config/logstash_config_test.go
Line 12: error: ExampleParseReader refers to unknown identifier: ParseReader (vet)

The ParseReader function is defined in logstash_config.go, which is a generated file and therefore has the comment // Code generated by pigeon; DO NOT EDIT., which instructs the go tools to not perform linting and vetting on this file (see golang/go#13560 (comment) for details).

[rameshvk]

I hit this regularly when working with examples (with a _test package name). Gometalinter will pass locally but goreportcard reports these false warnings. In most cases, I am able to change the example to use the same package name but sometimes that causes import cycles....

[mrinalwadhwa]

Have the same problem in this report https://goreportcard.com/report/github.com/ockam-network/did

Locally go vet github.com/ockam-network/did returns no errors.

[blockloop]

I'm having the same issue on my repo.

I was able to reproduce by running go vet <file> against only the example_test file rather than against the whole repo.

λ go vet example_scanner_test.go
# command-line-arguments_test
./example_scanner_test.go:33:1: ExampleRow refers to unknown identifier: Row
./example_scanner_test.go:55:1: ExampleRow_scalar refers to unknown identifier: Row
./example_scanner_test.go:74:1: ExampleRows refers to unknown identifier: Rows
./example_scanner_test.go:96:1: ExampleRows_primitive refers to unknown identifier: Rows
- master
λ go vet github.com/blockloop/scan
- master
λ

[ljmsc]

hi,
i had the same issue. Please note the naming convention described here:
https://golang.org/pkg/testing/#hdr-Examples

[github-actions]

This issue is stale because it has been open 60 days with no activity.