Permissive handling for runtimeconfig.json

[szilvaa-adsk]

This is a follow up to #91238. With these changes users can omit .runtimeconfig.json for C++/CLI dlls (i.e. dlls on Windows that link to ijwhost.dll) that are loaded after the runtime is already initialized. Note that missing .runtimeconfig.json will continue to cause a failure in all other situations.

[ghost]

Tagging subscribers to this area: @vitek-karas, @agocke, @VSadov
See info in area-owners.md if you want to be subscribed.

Issue Details

---

This is a follow up to #91238. With these changes users can omit .runtimeconfig.json for dlls that are loaded after the runtime is already initialized. Note that missing .runtimeconfig.json will continue to cause a failure in all other situations.

Author:	szilvaa-adsk
Assignees:	-
Labels:	area-Host, community-contribution
Milestone:	-

[vitek-karas]

Sorry for the possible misunderstanding. I perceived the discussion in #91238 as concluding that we would only change ijwhost. The fxr_resolver.cpp is shared among all the hosts, so your change modifies the behavior for all of them. So I think the change is good but I think it should add a new parameter to load_fxr_and_get_delegate whether the "non-existent runtime config" is allowed or not. And ijwhost would pass in true, everybody else would pass in false.

[szilvaa-adsk]

Ah! I will update the PR accordingly.

[szilvaa-adsk]

@dotnet-policy-service agree company="Autodesk"

[vitek-karas]

The product changes look good.

We need to add a test for the new behavior as well - and ideally also test that comhost correctly fails without a runtime config (not sure if we have such a test).

The ijwhost tests are here: https://github.com/dotnet/runtime/blob/main/src/installer/tests/HostActivation.Tests/NativeHosting/Ijwhost.cs

What I would do is modify this

runtime/src/native/corehost/test/nativehost/nativehost.cpp

Line 519 in 84a8f67

else if (pal::strcmp(command, _X("ijwhost")) == 0)

to accept additional parameter, which tells it if it should load a runtime - something similar to what other tests do (load from runtimeconfig and get some runtime delegate), and THEN trigger the ijwhost part of the test.

And then just add a new test into the C# part which moves/renames the runtimeconfig (make sure to make a copy of the test asset so that it doesn't corrupt test state for the other tests).

[elinor-fung]

comhost correctly fails without a runtime config (not sure if we have such a test)

Comhost.ActivateClass_ValidateIErrorInfoResult indirectly does this. It uses the missing runtime config scenario as the failure that should be set in error info.

[szilvaa-adsk]

@vitek-karas PR has been updated as you requested.

[szilvaa-adsk]

@vitek-karas @elinor-fung Is there anything else you need to approve this?

[elinor-fung]

One last suggestion for tests. Everything else looks good to me.

[elinor-fung]

Thank you!

[szilvaa-adsk]

@elinor-fung I can't merge this. Can you please merge?

[vitek-karas]

@szilvaa-adsk : Thank you very much for the change!