Introduce seven stages:
1) Analyze postponed arguments with fixed parameter types
2) Collect parameter types from constraints and lambda parameters' declaration
3) Fix not postponed variables for parameter types of all postponed arguments
4) Create atoms with revised expected types if needed
5) Analyze the first ready postponed argument and rerun stages if it has been analyzed
6) Force fixation remaining type variables: fix if possible or report not enough information
7) Force analysis remaining not analyzed postponed arguments and rerun stages if there are
^KT-37952 Fixed
^KT-32156 Fixed
^KT-37249 Fixed
^KT-37341 Fixed
#KT-36247 fixed
A lot of testdata changed because significanly less (error) descriptors
are created for unresolved types, so diagnostics became different.
Consider lower `Nothing` constraint non-proper only if there is a dependant not analyzed postponed atom.
Early completion to `Nothing` provides data flow info for smart casts.
KT-35668 Fixed
Treating special functions for `if`, `when`, `try`, `?:` as not accepting `Nothing` result type is incorrect.
Making so leads to cases with uninferred `Nothing` result type for inner calls and lost data flow info.
There is added a new service named `SubstitutingScopeProvider`, that
provides factory that creates captured types and approximator for them.
In OI they are the same as before commit, for NI they are empty, because
that approximation interferes with NI algorithm
That service is injected into function descriptors and property descriptors
and used for creating `SubstitutingScope` with correct services
Also there is changed time when we approximate captured types in NI
(after all call checkers)
#KT-25290 Fixed
In 1.3, due to changes in language, testdata for some tests can be
different from 1.2
We want to simlultaneously test both versions, so instead of fixing
language version in such tests, we split them into two: one with fixed
1.2, another with fixed 1.3
During subtyping/incorporation we transform types (e.g. changing nullability,
form of the type) and, basically, we're doing this to some FIXPOINT.
It's important that we use `KotlinType.hashCode()` to compare types, but
for error types hashCode is a hashCode of its supertype and, for example,
`makeNullableAsSpecified` method recreate type every time. So, we continue
to generate new constraints and we'll never stop incorporation algorithm
The problem is that when performing full analysis we do it in
a backward order while result for trivial vals is filled
in a forward one.
It turns out that reversedInstuctions might return a superset of
forward traversed instructions, e.g. in case of dead code in lambda.
At the same time result for trivial vals is constant
for any instruction, thus we can just return its constant value
and use it in the full analysis
#KT-20895 Fixed
It's anyway is obvious for write instruction, while effectively
after optimizations we treat val with syntactic initializer
uninitialized until their immediate initiazer
Parameters/vals with an immediate initializer (which we assume is a
rather common situation) do not require any kind of complicated CFA
- Unused vals can be simply determined by linear traversal of
the pseudocode
- Definite assignment is a bit more complicated: a read-instruction of val
can be considered as a safe if it's located *after* the first write in
the pseudocode. It works almost always beside the case with do/while
(see the test changed). This case will be fixed in the further commits
The test for kt897.kt will also be fixed further, all other changes
might be considered as minor as they mostly change diagnostics for
already red code
- Tell user what exactly is not supported (e.g., local inline function)
- Reduce diagnostics range to a keyword or an identifier
where appropriate
#KT-16223 Fixed Target versions 1.1.50
Ilmir Usmanov