@SteveL-MSFT Do you agree to merge the PR before GA?

@adityapatwardhan Please review the PR again.

After looking in depth how the method is used I very wonder that we implement the conversion of these two enums if we can request current value from an assembly using Reflection.
On my system typeof(object).Assembly.GetName().ProcessorArchitecture returns Amd64.

@iSazonov , can you please double check that currently GetProcessorArchitecture() returns system-wide info and comment on this function is wrong?
If that's the case, then probably to avoid future confusion:

1. current GetProcessorArchitecture() should be renamed across the code to something like GetSystemProcessorArchitecture(), and comment updated;
2. since the functionality of getting current process arch is also needed, code of this PR should go into PsUtils as a separate method like GetCurrentProcessProcessorArchitecture().

@anmenaga If the name confuses you we could remove the method at all because it is used in one place/scenario. Th scenario is a module loading. We read ProcessorArchitecture parameter from the module manifest and check that we can load the module. The check is to compare the ProcessorArchitecture parameter and ProcessorArchitecture property from current assembly. And we can do this directly. I pulled the commit.

@anmenaga Please update your review.

@anmenaga Could you please continue?

@SteveL-MSFT @anmenaga I think the PR could be in 7.0.

@iSazonov can you add a test to cover #6533?

@SteveL-MSFT Sorry, I did not saw the issue. New tests are added.

@SteveL-MSFT It seems I need help to fix tests on Unix-s. :-( I guess SMA compiled to MSIL on Unix-s. In the case I don't know how make functional tests on Unix-s.

@iSazonov You can probably not even worry about Add-Type, the module manifest issue was at validation time when it is imported so you can just generate the module manifest with ProcessorArchitecture and it should succeed on compatible arch and fail on not-compatible arch

@SteveL-MSFT I try to create functional tests to ensure that scenario like #6533 really works. But test failures show that perhaps the PR doesn't fix #6533 in general (the PR only removes broken p/invoke but do not ensure that import-module works) because Roslyn works in another way on Unix-s and probably loader too. Now I set to skip new tests on Unix-s.

Notice. The code was not covered by tests. Now only Windows on Intel is covered. Unix-s and Arm are not covered.

🎉v7.0.0-rc.2 has been released which incorporates this pull request.:tada:

Handy links:

• Release Notes