Better compliance with status spec

[bryannaegele]

Statuses are not allowed to be freeform and descriptions are only acceptable for errors. For the other statuses, descriptions are to be ignored.

https://github.com/open-telemetry/opentelemetry-specification/blob/main/specification/trace/api.md#set-status

[codecov]

Codecov Report

Merging #246 (f4b7356) into main (9110035) will increase coverage by 0.04%.
The diff coverage is 100.00%.

@@            Coverage Diff             @@
##             main     #246      +/-   ##
==========================================
+ Coverage   36.33%   36.37%   +0.04%     
==========================================
  Files          37       37              
  Lines        3162     3164       +2     
==========================================
+ Hits         1149     1151       +2     
  Misses       2013     2013              

Flag	Coverage Δ
api	63.04% <100.00%> (+0.16%)	⬆️
elixir	16.08% <0.00%> (-0.08%)	⬇️
erlang	36.37% <100.00%> (+0.04%)	⬆️
exporter	19.58% <ø> (ø)
sdk	78.89% <ø> (ø)

Flags with carried forward coverage won't be shown. Click here to find out more.

Impacted Files	Coverage Δ
apps/opentelemetry_api/lib/open_telemetry.ex	20.00% <ø> (ø)
apps/opentelemetry_api/lib/open_telemetry/span.ex	18.75% <ø> (ø)
...pps/opentelemetry_api/lib/open_telemetry/tracer.ex	46.15% <ø> (ø)
apps/opentelemetry_api/src/opentelemetry.erl	74.19% <100.00%> (+0.86%)	⬆️

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 9110035...f4b7356. Read the comment docs.

[bryannaegele]

One callout is the spec calls for the API to have the function be called set_status taking the arguments we're currently using to build the status record and then passing to this function. I'm reluctant to make the change at this point but wanted to point it out if anyone cares.

[hauleth]

Why not unicode:chardata/0? It would make the API a little bit simpler for the Erlang users.

[bryannaegele]

If I recall, we ran into some interop issues in some places around unicode charlists versus binary, so it settled out to using the binary everywhere as the default. I don't think it was this specific place (I believe it was around headers) but I think that was the origin of just doing the unicode binary.

[tsloughter]

My only other question before approving is related, why not allow undefined?

[bryannaegele]

I think we'd need to check the exporter would work ok with that since it has to export a binary for that either way, but it wasn't allowed to begin with. There was always a binary value guard on it. I was just aiming to tighten down the allowed values for the status and the description with this.