Codecov Report

❌ Patch coverage is 0% with 4 lines in your changes missing coverage. Please review.
✅ Project coverage is 82.05%. Comparing base (d2be704) to head (2535646).
⚠️ Report is 1 commits behind head on main.

@@            Coverage Diff             @@
##             main    #4039      +/-   ##
==========================================
- Coverage   82.07%   82.05%   -0.02%     
==========================================
  Files         385      385              
  Lines       15891    15895       +4     
==========================================
  Hits        13041    13041              
- Misses       2850     2854       +4     

Non-blocking: this improves the map-insertion failure path, but I think attributed observations can still terminate before this catch runs. The MetricAttributes construction copies attributes through noexcept callbacks/helpers that can allocate while storing keys/values. If that allocation throws, terminate happens before the outer catch in Observe() can handle it. We should either remove/catch those lower-level noexcept paths here or track it as a follow-up.

Wait - why are we doing POKEMON coding here? If there is an actual memory stomp, it'll obscure it even more. This is dangerous!

Wait - why are we doing POKEMON coding here? If there is an actual memory stomp, it'll obscure it even more. This is dangerous!

Fair point. This was approved and merged because it avoids std::terminate from the noexcept metrics path, but I agree the current catch (...) + silent return is too broad. I think we should improve this with a narrower follow-up fix.

Wait - why are we doing POKEMON coding here? If there is an actual memory stomp, it'll obscure it even more. This is dangerous!

Valid concern. As dbaker mentioned, I think the broader issue is how to make these failures observable without putting the application at risk.

If narrowing the catch clauses is needed here, I can send a follow-up fix.