Some functions have -1 as their end token positions

[chloestefantsova]

It is visible in the VM when the coverage report data is printed for those functions. One of such functions is replaceRange for some classes.

CL 43720 is written to mitigate the impact for the VM.

[peter-ahe-google]

Do you have an example of such a function?

[chloestefantsova]

Sorry I didn't provide the examples and the repro steps. I needed the issue number to refer to from the code (I'm currently fixing #32282). Currently I know the name of one of such functions. I've got it from a debug session when I printed out the names of the functions. I'm not sure what class it is from, though. I'll come back soon with concrete examples and repro steps.

[jensjoha]

At least when compiling the sdk, they all come from forwarding stubs, and the following fixes it:

diff --git a/pkg/front_end/lib/src/fasta/type_inference/interface_resolver.dart b/pkg/front_end/lib/src/fasta/type_inference/interface_resolver.dart
index be842cf29d..4b91c19f30 100644
--- a/pkg/front_end/lib/src/fasta/type_inference/interface_resolver.dart
+++ b/pkg/front_end/lib/src/fasta/type_inference/interface_resolver.dart
@@ -480,6 +480,7 @@ class ForwardingNode extends Procedure {
         fileUri: enclosingClass.fileUri,
         forwardingStubInterfaceTarget: finalTarget)
       ..fileOffset = enclosingClass.fileOffset
+      ..fileEndOffset = enclosingClass.fileOffset
       ..parent = enclosingClass
       ..isGenericContravariant = target.isGenericContravariant;
   }

I'm, not quite sure the offsets make sense though...

[jensjoha]

Fields also contain endOffsets, and doesn't (at least always) have them. In the VM they are used to set the end offsets for setters and getters and to setup scopes (when the field has an initializer).

Also, not all constructors have an endOffset.