Swap to go-junit-report for parsing go test output #982
Conversation
| @@ -169,11 +169,6 @@ func TestGoIgnoreUnknownOutput(t *testing.T) { | |||
| assert.Equal(t, 0, results.Skips()) | |||
| } | |||
|
|
|||
| func TestGoFailIfUnknownTestPasses(t *testing.T) { | |||
There was a problem hiding this comment.
I'm not sure how often we expect this particular case to come up. Go Junit report doesn't error out in this scenario - but it does 'misattribute' a failure to a Test case, so I'm not sure it's 100% correct in that regard.
There was a problem hiding this comment.
That dates back to literally before the dawn of time (initial commit). I don't think I've ever seen it actually happen though; not terribly bothered about losing this one case if everything else is still passing.
src/test/go_results.go
Outdated
| testPackage, err := parser.Parse(bytes.NewReader(data), "") | ||
|
|
||
| if err != nil { | ||
| return core.TestSuite{}, fmt.Errorf("Failed to parser go test output: %w", err) |
There was a problem hiding this comment.
nit: Failed to parser -> Failed to parse
I doubt anyone deeply cares about the older versions; we support I think 1.8+ (https://golang.org/pkg/testing/#MainStart is probably the most critical thing there) but I don't imagine lots of people deeply care about building 1.8 or 1.9 with the latest version of plz. |
…achine#982)" This reverts commit 13adc19.
Swap to using a 3rd party library for parsing Go test output.
This should cover more edge cases than our previous simple parser.
An additional note - by reading various PRs in
go-junit-reportaround the different edge cases in Go test output. I've found that go test has a json output which should be used for parsing tasks like we're doing. However there's discussions on thego-junit-reportrepo which suggest that there aren't much gains to be had just yet by using the go test json output.Also the json option only exists for Go 1.10+. I think the best thing to do going forwards is to revisit this if we officially stop support Go versions < 1.10, to see if we hit any more edge cases that would be solved by using a json format parser.