Codecov Report

Attention: 1 lines in your changes are missing coverage. Please review.

Comparison is base (5140a36) 64.90% compared to head (1ed3ab1) 65.31%.

@@              Coverage Diff              @@
##             develop     #666      +/-   ##
=============================================
+ Coverage      64.90%   65.31%   +0.40%     
- Complexity      3368     3406      +38     
=============================================
  Files            313      313              
  Lines          14960    14974      +14     
  Branches        2525     2525              
=============================================
+ Hits            9710     9780      +70     
+ Misses          4351     4304      -47     
+ Partials         899      890       -9     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

Hi folks, I've been running some tests on my end with this patch and I wanted to note I still see some issues with indys.
For example

public Stream<LocalDate> test(LocalDate endExclusive, Period step) {
        long months = step.toTotalMonths();
        long days = step.getDays();
        int sign = months > 0 || days > 0 ? 1 : -1;
        long steps = 200000 / (months + days);
        return LongStream.rangeClosed(0, steps).mapToObj(
                n -> endExclusive.plusDays(days * n));
    }

issues the following error:

Error: java.lang.RuntimeException: Failed to convert <jlink.Q: java.util.stream.Stream test(java.time.LocalDate,java.time.Period)>

Printing the stack trace

java.lang.RuntimeException: Illegal Assignment statement. Make sure that right hand side has a valid operand.
	at [email protected]/sootup.core.jimple.common.stmt.JAssignStmt.<init>(JAssignStmt.java:55)
	at [email protected]/sootup.core.jimple.Jimple.newAssignStmt(Jimple.java:530)
	at [email protected]/sootup.java.bytecode.frontend.StackFrame.mergeIn(StackFrame.java:160)
	at [email protected]/sootup.java.bytecode.frontend.AsmMethodSource.convertInvokeDynamicInsn(AsmMethodSource.java:1402)
	at [email protected]/sootup.java.bytecode.frontend.AsmMethodSource.convert(AsmMethodSource.java:1774)
	at [email protected]/sootup.java.bytecode.frontend.AsmMethodSource.resolveBody(AsmMethodSource.java:215)
	at [email protected]/sootup.core.model.SootMethod.lazyBodyInitializer(SootMethod.java:98)
	at [email protected]/com.google.common.base.Suppliers$NonSerializableMemoizingSupplier.get(Suppliers.java:181)
	at [email protected]/sootup.core.model.SootMethod.getBody(SootMethod.java:177)

I haven't had a chance to do much digging to see if this is still somehow related to the operand stack but I thought I would bring it up. Thanks!

With Sonias example input there was null on the rhs of the assigment.
Then I merged the other porting regression bug fix where we missed handling one Operand.
And now we have:
a) "Invalid in operands length! " in StackFrame with the example

seems the static invoke itself is missing from before..?

b) SwitchExprWithYieldTest fails due to a stack underrun

so it seems there are missing operands that are not added.. or we are taking too much.

I have investigated Test: SwitchExprWithYieldTest of java.bytecode.
The stack underrun is probaly caused by handling a virtual invoke by the dynamic invoke, but the invoke should be static. That consumes one element from the stack which is missing for the parameter of the dynamic invoke call.
This seems to only happen if a dynamic invoke is converted for a second time, which is caused by the switch since there are two branches over the same dynamic invoke instruction.

this line is probably the issue:
final boolean isStaticInvokeExpr = expr instanceof JStaticInvokeExpr;
since the expr is always a dynamic invoke. the coverage report also shows that the if branch (boolean true) is never entered

But couldnt find the time to further investigate this, today.

@SoniaZaldana does the current state of this PR fix the indy problems in your project?