provider/azure: strip off UTF-8 BOM from responses

[axw]

When handling responses from the AAD REST endpoint,
strip off the UTF-8 BOM, if any, from JSON responses.
These are only sometimes returned.

Fixes https://bugs.launchpad.net/juju/+bug/1657448

[axw]

QA

add-credential azure, interactive, 10 times

[rogpeppe]

Given that we're now reading the whole body, why not:

 var utf8BOM = []byte("\ufeff")    // global
 data, err := ioutil.ReadAll(resp.Body)
 if err != nil ...
 if err := json.Unmarshal(bytes.TrimPrefix(data, utf8BOM), &oe); err != nil ... {
 }
 resp.Body = io.NopCloser(bytes.NewReader(data))

?

BTW I think it might be better to strip the BOM in all JSON decoding code - surely
this isn't the only place that can include it?

[axw]

BTW I think it might be better to strip the BOM in all JSON decoding code - surely
this isn't the only place that can include it?

Maybe, but we don't control the other code. That will need to go in go-autorest. The reason this code exists at all is because the AAD (Azure Active Directory) API endpoint is a little different to the other Azure API endpoints. Different error structures. So I wouldn't be surprised if this is the only place that the BOM shows up, because it's probably getting routed to a different backend.

[axw]

As for the simplification: I'm intentionally preserving the behaviour of leaving the body as-is for the caller. But I think you're right, we should just chuck out the BOM and pretend it was never there. It's no use to anyone.

EDIT: and now I see yours preserves it too, I'm just a bit thick. I'll go with that.

[rogpeppe]

LGTM with one thought, thanks.

[axw]

$$bill$$

[jujubot]

Status: merge request accepted. Url: http://juju-ci.vapour.ws:8080/job/github-merge-juju