during tail-call optimization.
There can be code, where all next instructions are non-meaningful and
there is a back-edge, for example, while(true){}. Previously, analyzer
incorrectly assumed, that this cannot happen. Now, it keeps track of
visited instructions and says, that there is no meaningful instruction
in such case.
#KT-56815 Fixed
These type parameters where used in function parameters,
but since suspendImpl is static function, it has no access
to class type parameters. Solution is to copy them to
the function itself.
#KT-55125 Fixed
Being disabled by default
and not well-documented, these functions cause confusion among early
adopters as to why their code don't work properly.
Assert APIs need a proper design across Kotlin platforms.
Since APIs are not available in common code and K/JS, it is premature
to have such a general feature in a new experimental platform.
Compiler tests:
* Mute tests that rely on assert.
* Replace JVM-specific assert calls with require calls and unmute passed K/JS tests.
Merge-request: KT-MR-8636
Merged-by: Svyatoslav Kuzmich <svyatoslav.kuzmich@jetbrains.com>
in startCoroutineUninterceptedOrReturn. Otherwise, the coroutine will
not be interceptable later.
Add a test, which checks, that intercepted continuation is released.
#KT-55869
If the super class is in a file that has already been lowered, the base
method has an extra continuation parameter which breaks things.
Also, SAM wrappers around functional objects are tail-call and do not
need continuations ever, so don't even try.
^KT-50950 Fixed
We already check for {POP, Unit} sequence before ARETURN, but if the
there are multiple sequence before ARETURN, the compiler assumes, that
TCO misses.
The fix is to check, that the instruction after the sequence is either
ARETURN or another {POP, Unit} sequence.
#KT-50835
#KT-54152 Fixed
instead of bytecode text ones. Check the content of continuation
object instead of bytecode, since this is less prone to changes during
changes in coroutines codegen.
#KT-48678
1. merge(null of type A, null of type B) = null of unknown type;
2. merge(null of type A, something of type B) = merge(unknown null, B).
^KT-52311 Fixed
This is somewhat suboptimal since this results in `::suspendInline`
generating 2 classes while `{ suspendInline() }` only creates 1, but
it's the best allowed by the existing hierarchy of classes in stdlib. At
least it works?
^KT-50832 Fixed
Just see if every suspend call is followed only by safe instructions or
returns, then insert a suspension point check if there isn't a direct
return.
The first part of this is equivalent to the old implementation, just
refactored. The second part generates strictly more checks; see, for
example, the fixed test, in which the previous implementation failed to
insert a check before a CHECKCAST.
^KT-51818 Fixed
If we do not do this, the state-machine builder will not know the type
of the ACONST_NULL, defaulting to Object, leading to VerifyError.
Alternatively, we could use LVT to deduce the type, but getting types
from LVT is something I got rid of long time ago, and I have no desire
to return it back.
Generating CHECKCAST hints the state-machine builder the type of the
variable avoiding the issue of VerifyError. However, this CHECKCAST
replaces StrictBasicValue.NULL_VALUE with BasicValue in
OptimizationBasicInterpreter. To preserve optimization on not-spilling
known nulls, introduce BasicValues, which represent typed nulls and
create BasicInterpreter, which is aware of them. This way we have the
best of two worlds - we do not spill known nulls, and we know the type
of ACONST_NULL.
#KT-51718 Fixed
Ilmir Usmanov