set_status: use update_element since select_replace doesn't use index

[tsloughter]

No description provided.

[codecov]

Codecov Report

Merging #384 (6cf7b36) into main (f9996de) will increase coverage by 0.00%.
The diff coverage is 83.33%.

@@           Coverage Diff           @@
##             main     #384   +/-   ##
=======================================
  Coverage   73.29%   73.29%           
=======================================
  Files          53       53           
  Lines        1640     1644    +4     
=======================================
+ Hits         1202     1205    +3     
- Misses        438      439    +1     

Flag	Coverage Δ
api	68.70% <ø> (ø)
elixir	17.54% <ø> (ø)
erlang	74.73% <83.33%> (+<0.01%)	⬆️
exporter	74.41% <ø> (ø)
sdk	78.08% <83.33%> (-0.02%)	⬇️
zipkin	53.84% <ø> (ø)

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

Impacted Files	Coverage Δ
apps/opentelemetry/src/otel_span_ets.erl	72.09% <83.33%> (+0.29%)	⬆️

---

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 f9996de...6cf7b36. Read the comment docs.

[ferd]

I figure there's a risk where you can run into races with this, but given the current format of the writes and the possibility of getting a false result if a delete has occured, this appears to be safe in context given the current operations.

[ferd]

There could be a small race where the call to make a span status as ok after it was error interleaves oddly with scheduling and fails (by setting ok before error). I'm not sure how concerned you are with that.

[tsloughter]

Yea, technically the possibility of a race is part of every operation in this module. But a span should only be updated by a single process. This is not currently enforced since it would require somehow tracking the "owner" (can't just be the process that started it since it could be created to be passed to a newly spawned process), but since the majority of usage will be as the current span stored in the process dictionary it isn't something too be concerned about I think.