[mono] Fix StackTrace from a dim and Vtable offsets for static interface method

[thaystg]

• Fix StackTrace when called from a DIM.
• Fix the other test case that was added for @bholmes, and this case when the method TestMethod5 was being called it was executing TestMethod10, and this was fixed skipping static interface methods when was calculating vtable offsets.

The fix was completely done by @vargaz, I just opened the PR.
Thanks @vargaz .

Fix#60486

[ghost]

Tagging subscribers to this area:
See info in area-owners.md if you want to be subscribed.

Issue Details

---

I'm not sure if this is the right fix, but with it I can see the callstack correctly as I can see if I run using CoreCLR.

Output running on mono:

at System.Environment.get_StackTrace()
at DefItf.ISubscriber`1.InternalSubscribe(ISubscriber`1 subscriber, IPublisher`1 publisher) in /Users/thaysgrazia/runtime/src/mono/sample/HelloWorld/Program.cs:line 27
at DefItf.Program.Subscribe(IPublisher`1 publisher) in /Users/thaysgrazia/runtime/src/mono/sample/HelloWorld/Program.cs:line 68
at DefItf.Program.Start() in /Users/thaysgrazia/runtime/src/mono/sample/HelloWorld/Program.cs:line 61
at DefItf.Program.Main(String[] args) in /Users/thaysgrazia/runtime/src/mono/sample/HelloWorld/Program.cs:line 53
do something
at System.Environment.get_StackTrace()
at DefItf.ISubscriber`1.InternalUnsubscribe(ISubscriber`1 subscriber, IPublisher`1 publisher) in /Users/thaysgrazia/runtime/src/mono/sample/HelloWorld/Program.cs:line 33
at DefItf.ISubscriber`1.Unsubscribe(IPublisher`1 publisher) in /Users/thaysgrazia/runtime/src/mono/sample/HelloWorld/Program.cs:line 21
at DefItf.Program.Start() in /Users/thaysgrazia/runtime/src/mono/sample/HelloWorld/Program.cs:line 63
at DefItf.Program.Main(String[] args) in /Users/thaysgrazia/runtime/src/mono/sample/HelloWorld/Program.cs:line 53

Author:	thaystg
Assignees:	-
Labels:	area-VM-meta-mono
Milestone:	-

[lambdageek]

Related to #60486. @thaystg can you add a test to https://github.com/dotnet/runtime/tree/main/src/tests/Loader/classloader/DefaultInterfaceMethods/regressions

/cc @bholmes

[vargaz]

The stacktrace is supposed to contain the exact generic instance not DefItf.ISubscriber`1.

[vargaz]

This seems to fix it for me:
https://gist.github.com/vargaz/2ded43b6598b9e2e117325eeed4cf6f0

[thaystg]

@vargaz , your fix looks like it works the result running on mono with your fix would be:

 at System.Environment.get_StackTrace()
at DefItf.ISubscriber`1[[DefItf.InputData, HelloWorld, Version=7.0.0.0, Culture=neutral, PublicKeyToken=31bf3856ad364e35]].InternalSubscribe(ISubscriber`1 subscriber, IPublisher`1 publisher) in /Users/thaysgrazia/runtime/src/mono/sample/HelloWorld/Program.cs:line 27
at DefItf.Program.Subscribe(IPublisher`1 publisher) in /Users/thaysgrazia/runtime/src/mono/sample/HelloWorld/Program.cs:line 68
at DefItf.Program.Start() in /Users/thaysgrazia/runtime/src/mono/sample/HelloWorld/Program.cs:line 61
at DefItf.Program.Main(String[] args) in /Users/thaysgrazia/runtime/src/mono/sample/HelloWorld/Program.cs:line 53
do something
at System.Environment.get_StackTrace()
at DefItf.ISubscriber`1[[DefItf.InputData, HelloWorld, Version=7.0.0.0, Culture=neutral, PublicKeyToken=31bf3856ad364e35]].InternalUnsubscribe(ISubscriber`1 subscriber, IPublisher`1 publisher) in /Users/thaysgrazia/runtime/src/mono/sample/HelloWorld/Program.cs:line 33
at DefItf.ISubscriber`1.Unsubscribe(IPublisher`1 publisher) in /Users/thaysgrazia/runtime/src/mono/sample/HelloWorld/Program.cs:line 21
at DefItf.Program.Start() in /Users/thaysgrazia/runtime/src/mono/sample/HelloWorld/Program.cs:line 63
at DefItf.Program.Main(String[] args) in /Users/thaysgrazia/runtime/src/mono/sample/HelloWorld/Program.cs:line 53

But it's different of what I get when I run on CoreCLR where I have:

[vargaz]

Its not defined what the result should be, we try to obtain the exact generic instance if possible.

[bholmes]

I created a new program with more cases.

https://gist.github.com/bholmes/e3ccaf9d044b7261d6865c8f0173dc64

Method5 appears to be broken but I think this patch is good.

[thaystg]

/azp run runtime-manual

[azure-pipelines]

Azure Pipelines successfully started running 1 pipeline(s).

[lambdageek]

We're going to need to revisit this (and line 1535, below) when non-abstract static virtual methods are going to be supported

[bholmes]

If this is approved, I want this to go back to mono/mono. Let me know if and when you want me to back port these changes to that repo. (assuming that automation will fail.)

[lambdageek]

/azp run sync-runtime-to-mono

[azure-pipelines]

Azure Pipelines successfully started running 1 pipeline(s).

[lambdageek]

There are CoreCLR and Mono test failures from the new test.

1. CoreCLR failures are because the new test methods should all be marked NoInlining
2. Mono failure is because the validation for

    t1.TestMethod5(null, new [] {new StackFrame ("TestItf1`1", "TestMethod3"), new StackFrame ("TestItf2`1.TestItf1", "TestMethod5")})
   

   is too strict - the actual string in Mono is TestItf2`1[[InputData, github60486, Version=0.0.0.0, Culture=neutral, PublicKeyToken=null]].TestItf1<TT>. So we should make something like new StackFrame (new [] {"TestItf2`1", "TestItf1"}, "TestMethod5") so that the class name can be in multiple parts.

In both cases, it's an issue with the testcase, not with the implementation. (So the mono backport is still ok)

[thaystg]

/backport to release/6.0