Mark instance method bodies lazily

[mrvoorhe]

Instance method bodies do not need to be marked until an instance of the type could exist. Any instance methods that were marked on types that are never instantiated will be converted to a throw.

• A number of existing tests needed to be updated to avoid lazy body marking

• Add a new CodeOptimization to allow disabling this mechanism. Reasoning was, (1) it's simple to support an option. (2) It may be helpful if it causes problems. (3) I could see using this option to make writing certain tests easier.

• Marking of some small bodies will not be deferred. The idea is that the size cost of some methods is less than converting them to a throw. So we might as well leave them alone. See IsWorthConvertingToThrow

• Factored out MarkAndCacheNotSupportedCtorString since I need to call it from multiple places now.

• Expose the RewriteBody* methods in CodeRewriterStep for overriding. A few motivations for this. (1) We have a runtime that doesn't support exceptions. When we target that runtime we will need to override ConvertToThrow with something else. (2) I will likely override all of the RewriteBody* methods so that I can record which bodies are changed and hook it up to logging mechanisms that we have.

Note. This PR has a test that depends on #501 . That PR will need to land first and I will need to rebase after that lands. There will be 1 failing test until that happens.

[mrvoorhe]

Rebased and should be ready to go pending CI results 🤞

[marek-safar]

Did you forget to push the changes ;-)

[mrvoorhe]

Did you forget to push the changes ;-)

I didn't finish making all of the changes before I had to go.

All changes are done and pushed now.

[mrvoorhe]

I'm looking into the test failures with illink

[mrvoorhe]

I refactored IsWorthConvertingToThrow a bit to handle noop. A bunch of unrelated tests were failing during the illink build because roslyn likes to write empty methods as

noop
ret

I can't get the .net core tests running locally. Hopefully everything is fixed now, but I'll have to wait for CI to finish to know for certain.

[mrvoorhe]

fyi, I had to add this so that System.NotSupportedException could be found. This could become a tedious pattern to maintain

[mrvoorhe]

@sbomer BCL.FindPredefinedType can't find System.NotSupportedException on .netcore. Any idea how I should locate that type?

I tried adding

// System.NotSupportedException lives here on .netcore
			if (cache.TryGetValue ("System.Runtime", out corlib))
				return corlib.MainModule.GetType (ns, name);

To FindPredefinedType but that didn't work.

I can't get the .netcore tests working locally so that's making iterating on this challenging.

[sbomer]

@mrvoorhe I can take a look. Could you try running the tests locally on core using ./eng/dotnet.sh test illink.sln and let me know if you have any issues?

[sbomer]

Also, it looks like System.NotSupportedException in System.Runtime is just a type forwarder to System.Private.CoreLib. I'm not sure why it wouldn't have been found in SPC in the first place.

[sbomer]

Switching System.Runtime -> System.Private.CoreLib in FindPredefinedType fixes at least some of the test failures when I run locally (I haven't run them all yet). I'm not sure though exactly what the intention of FindPredefinedType is or why it wouldn't work with type forwarders. If it's just to resolve a set of types that we assume are present in some particular assembly for the linker to work, then it seems like hard-coding System.PrivateCoreLib might be ok. Maybe we could use a define for the target's core library similar to https://github.com/mono/linker/blob/31555535b160c1de7ba72fd7c51ec584ef395fdb/test/Mono.Linker.Tests.Cases.Expectations/Support/PlatformAssemblies.cs?

[mrvoorhe]

@sbomer How can I debug this? I tried what you said and I can reproduce the failure locally. However, Rider EAP and VS 2019 are not happy with illink.sln and I tried to do some Console.WriteLine debugging but it looks like dotnet.sh test illink.sln is doing something other than let stdout go to the console

[mrvoorhe]

I think I know what the issue is now. The test that is failing is an IL test.

That said, I would still like some pro tips on how to debug individual tests on .netcore for future reference

[sbomer]

#471 may help with IDE support for the .sln, but I'm not sure. I've had lots of issues with tooling support for projects that mix configurations/targetframeworks and projectreferences. Sorry it's causing you trouble.

For debugging, I've been using VS code. Here's the best way I've found to do it so far:

• Add this to the linker somewhere:

  while (!System.Diagnostics.Debugger.IsAttached)
      System.Threading.Thread.Sleep(500)
  

• Set breakpoints
• Run an individual test:

  ./eng/dotnet.sh test illink.sln --filter TestCaseName
  

  (See https://github.com/Microsoft/vstest-docs/blob/master/docs/filter.md for some docs on the filter expression)
• Wait for it to hit the while loop above, and attach to testhost.dll.

[mrvoorhe]

@marek-safar Good to go!

[alexanderkyte]

This has been shown to regress the mono harness when mono is bumped to this commit.

[mrvoorhe]

@alexanderkyte Can you investigate a bit? If any types are created from native and do not get an instance .ctor marked it will cause problems. If this is the case it is not necessarily a problem with these changes. Instead the fix is to properly annotate w/e instance .ctor is needed by the type.

It's certainly possible there is a bug with these changes, but I'd prefer to rule out the above before making any changes to disable this.

[alexanderkyte]

@mrvoorhe I have a PR testing whether disabling the unreachablebodies optimization will fix the build. If so, I will triage what breaks when we enable the optimization. That way we can make use of other changes to the linker while we get unreachablebodies working separately.