Printing the call site source on errors isn't really useful anyway - the
old backend points to a specific PsiElement, and JVM_IR wraps the
exception in FunctionCodegen (and printing the source never worked with
JVM_IR anyway).
rather than a fake inline function.
Also, generate more correct instructions for typeOf. Not sure how that
even worked before - `aconst(Boolean)` isn't even valid.
This avoids an extra call to 'analyze', which is rather costly.
Update debugger testData: Constant condition
elimination now performs DCE more consistently.
To avoid depending on a heavy module 'frontend' in parts of the compiler
where checking for optional annotations is needed, such as in
'ir.serialization.common'.
Anonymous types are not approximated by frontend for private declarations.
Class IDs for such types are replaced by StringTable before being written
to metadata. JVM string table mangles such types and keeps them generic
for reflection. For other purposes the types are replaced with `Any`.
Type arguments of the replaced type should be ignored in the latter case.
Otherwise decompiled text builder crashes on an attempt to restore `Any`
type with non-zero number of type arguments.
The code pretending to replace a type with its first supertype was dropped
from ApproximatingStringTable for two reasons:
- the first type from `getAllSuperClassifiers` is the original type itself
which doens't provide a ClassId for anonymous type, so it was a noop
- tracking potential type arguments of the first anonymous type's supertype
would be a complication with almost no practical value (types in decompiled
text would be slightly closer to the real type of a private declaration).
^KT-46393 Fixed
1. if an argument of a `pop` cannot be removed, then all other potential
arguments of that `pop` can't be removed either, and the same applies
to other `pop`s that touch them;
2. the same is true for primitive conversions, but this is even trickier
to implement correctly, so I simply did the same thing as with
boxing operators: replace the conversion itself with a `pop` and keep
the argument as-is.
Somehow this actually removes *more* redundant primitive type conversions
than the old code in a couple bytecode text tests, so I've patched them
to kind of use the value, forcing the instructions to stay.
#KT-46921 Fixed
CoroutineTransformermethodVisitor attempts to extend the ranges
of local variables in various situations. Probably in an attempt
to give a better debugging experience. However, all of these
range extensions lead to invalid local variable tables where
something is in the local variable table where nothing is in the
corresponding slot.
The code that extends variables to the next suspension point
instead of ending them when they are no longer live has issues
with loops. When resuming and reentering the loop, the locals
table will mention a local that we did not spill and which
is therefore not restored when resuming.
The code that extends local variable table entries if there
are no suspension points between two entries doesn't work
for code such as:
```
var s: String
if (suspendHere() == "OK") {
s = "OK"
} else {
s = "FAIL"
}
```
If the local variable ranges are collapsed into one, one of
the branches will have the local defined in the local variable
table before the slot is initialized.
Old compiler versions still won't be able to load default lambdas
generated by JVM_IR, but this way we avoid incorrect behavior of
function references taking inline class types that unbox to Any.
#KT-46601 Fixed
Also, produce more correct results in IrDefaultLambda's
`invokeMethodParameters` and `invokeMethodReturnType`. This affects
whether the inliner inserts inline class boxings/unboxings around lambda
calls; while this doesn't matter now due to KT-46601, it would if the
naming was fixed.
pyos