fix: Fix missing metric response attributes in HTTPX instrumentation

[diogosilva30]

Description

This PR fixes an issue where HTTP response attributes (status code, http version) are only recorded in metrics if span are recording. I believe this should not be the case because metrics should not be coupled with tracing. Based myself of requests instrumentation here, where it even states:

Also the decision to include status code on a histogram should
not be dependent on tracing decisions.

Fixes #3614

Type of change

Please delete options that are not relevant.

• Bug fix (non-breaking change which fixes an issue)

How Has This Been Tested?

Please describe the tests that you ran to verify your changes. Provide instructions so we can reproduce. Please also list any relevant details for your test configuration

• Added unit tests for both old and new sem conv
• Ran script provided in minimum reproducible code and saw after this change behaviour is as expected

Does This PR Require a Core Repo Change?

• No.

Checklist:

See contributing.md for styleguide, changelog guidelines, and more.

• Followed the style guidelines of this project
• Changelogs have been updated
• Unit tests have been added
• Documentation has been updated

[linux-foundation-easycla]

The committers listed above are authorized under a signed CLA.

• ✅ login: diogosilva30 / name: Diogo Silva (658a1ce, 0682ab0, dd4f3de, 8df6a6d, 2067c9e, 4dc0c8a, 96814d7, 27f63e2)
• ✅ login: emdneto / name: Emídio Neto (1079fed, 91b3e74)
• ✅ login: xrmx / name: Riccardo Magliocchetti (c025178)

[diogosilva30]

From manual testing this looks good and it's a simple change. Still need to figure out unit test updates tho

[diogosilva30]

Added unit tests for both old and new sem conv ✅
Should be ready for review!

[diogosilva30]

@xrmx let me know if further changes are required

[xrmx]

@xrmx let me know if further changes are required

I think you missed this comment #3615 (comment)

[diogosilva30]

@xrmx Indeed I've missed, apologize.

If you check previously existing tests those attributes are only asserted when test ends in both_semconv which means its running with opt-in as http/dup. For "default" behaviour (old semantic) only HTTP_STATUS_CODE, HTTP_METHOD and HTTP_SCHEME are added. To have HTTP_FLAVOR, HTTP_HOST , NET_PEER_NAME , NET_PEER_PORT it needs to run in http/dup (both sem conv).

We can see this explanation in original implementation by @emdneto : #2631

Don't know if it makes sense, but it's the current behaviour. I've added a new test case so it covers all 3 scenarios according to logic previously implemented by @emdneto :

• old sem conv: test_metrics_have_response_attributes_with_disabled_tracing
• both sem conv: test_metrics_have_response_attributes_with_disabled_tracing_both_semconv
• new sem conv: test_metrics_have_response_attributes_with_disabled_tracing_new_semconv

[diogosilva30]

@xrmx any chance to review this again?

[diogosilva30]

@xrmx @emdneto any update on this? Happy to make further changes if needed

[diogosilva30]

Thanks for review @emdneto @xrmx ! Everything good for merge?