vmm_tests: memory validation with release baseline

[dpaliulis-msft]

Description:
This PR adjusts the memory validation tests to compare test results on the release build against their own set of baseline memory usage values.

Changes:

• Added small and large GP memory validation test for release only builds
• Added new specifier for "build_version" in memstat baseline comparison method
• Added "release" memory usage baseline values to memstat_baseline.json

[Copilot]

Pull Request Overview

This PR separates memory validation tests by build type (debug vs release) to enable more accurate baseline comparisons. The changes allow memory usage validation tests to compare against build-specific baseline values rather than using a single set of baseline values for both debug and release builds.

• Added build version parameter to memory validation baseline comparison logic
• Split existing memory validation tests into separate debug and release variants using conditional compilation
• Added release-specific baseline memory usage values to the test data configuration

Reviewed Changes

Copilot reviewed 3 out of 3 changed files in this pull request and generated no comments.

File	Description
vmm_tests/vmm_tests/tests/tests/multiarch/memstat.rs	Added build_version parameter to compare_to_baseline method and updated all baseline JSON lookups
vmm_tests/vmm_tests/tests/tests/multiarch.rs	Split memory validation tests into debug/release variants with conditional compilation and renamed existing functions
vmm_tests/vmm_tests/test_data/memstat_baseline.json	Added release baseline values alongside existing debug values in nested JSON structure

[github-actions]

At least one Petri test failed.

[mattkur]

Thanks for this, and I see where you're going here. Adding memory usage tests with release config is important and valuable!

I think it would make more sense to have your test explicitly depend upon the release-built variant of openhcl. I suspect there are a few things we'd need to think through to make that work, and maybe this is why you chose to implement the tests this way :)

1. Our default CI does not build a release variant of the openhcl binary (I think that's what happens when someone tags a PR as release-ci-required, which I just did here). Maybe that's okay: this PR adds tests that should only run in release builds (which means the CI pipe, not per PR).
2. We would need to explicitly declare an artifact for the release-built IGVM, and resolve it only on release builds. I see you chose the most appropriate built-in cfg() flag, but I suspect we will want to be more explicit.

Interested in what you, and also what other reviewers, think.

[github-actions]

At least one Petri test failed.

[smalis-msft]

I initially suggested he do things this way for exactly those reasons Matt haha. Since we only produce a release build in the CI and optional release gates, this test can only really run there. This is somewhat of an ongoing experiment to gather data and refine what we're looking for here. If we see it failing a lot in CI we'll adjust.

[mattkur]

I initially suggested he do things this way for exactly those reasons Matt haha. Since we only produce a release build in the CI and optional release gates, this test can only really run there. This is somewhat of an ongoing experiment to gather data and refine what we're looking for here. If we see it failing a lot in CI we'll adjust.

I must be on the right track if you and I are sharing thoughts :)

I only have a few nits then, @dpaliulis-msft :

1. This matrix is getting large, so suggest that you define a macro to get rid of much of the boilerplate.
2. "version" is typically used to mean something akin to the contents of the code that was compiled, suggest you use something more explicit (like optimizations, config, etc.)

[smalis-msft]

For your point 2, i commonly see "flavor" as well.

[smalis-msft]

For 1 I'm not sure a macro is really that needed yet, but maybe it's worth moving the tests themselves into memstat.rs just to keep things contained.

[github-actions]

At least one Petri test failed.