azalea/kotlin-fork
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity
82,127 Commits 1 Branch 0 Tags
2c5a4dcb985ef2c91787d5bfd459fefbdbc47c95
Commit Graph

3844 Commits

Author SHA1 Message Date
Mads Ager 2c5a4dcb98 [JVM IR] Fix constant folding to use basic types always. 
Fixes KT-46540.
 2021-05-12 11:57:20 +02:00
pyos d5d6736e67 PSI2IR/FIR2IR: do not approximate T!! before translation 
This leads to weird effects when it's in a contravariant position,
because it's approximated by Nothing.
 2021-05-12 11:37:48 +03:00
pyos a37db99841 JVM_IR: remove a hack from InlineCallableReferenceToLambda 
It breaks other things, the same problem is hit by FunctionReference,
and it's the translation layer's fault anyway.

 #KT-46555 Fixed
 2021-05-12 11:37:47 +03:00
pyos cff74b31d4 JVM_IR: get unbound property reference receiver type from KProperty1 2021-05-12 11:37:45 +03:00
Mads Ager d397efb2bd [JVM IR] Copy attributes for JVM_STATIC_WRAPPERS. 2021-05-11 16:26:04 +02:00
Mads Ager 3db5ba98ad [JVM IR] Do not insert Nothing handling in JvmStatic wrapper. 
Fixes KT-46568.
 2021-05-11 16:26:04 +02:00
Ilmir Usmanov 640d263ae1 Do not unbox nullable Result, since before usage it is coerced 
#KT-46505
 2021-05-11 05:49:24 +02:00
pyos 656c2496a6 JVM_IR: omit bound receiver in IrExpressionLambdaImpl parameters 
and remove a hack intended to replace the missing type conversions this
caused.

 #KT-46505 Fixed
 2021-05-11 05:49:20 +02:00
Dmitry Petrov ac0aaff611 JVM_IR KT-46562 don't use LambdaMetafactory for Serializable SAMs 
TODO support serializable lambdas creation with LambdaMetafactory
 2021-05-10 14:54:14 +03:00
Jinseong Jeon bdfc879f00 FIR checker: report UNINITIALIZED_PARAMETER 2021-05-10 14:26:14 +03:00
Mikhael Bogdanov cd7841ceed Generate type annotations on class type parameters bounds 
#KT-13228
 #KT-46539 Fixed
 2021-05-10 12:12:21 +03:00
Mikhael Bogdanov cbe3c66156 Generate function type parameter annotations and type annotations on their bounds into bytecode 
Generate type parameter annotation by default for `-Xjvm-target 1.8` and above

 #KT-46539
 #KT-13228
 #KT-46545 Fixed
 2021-05-10 12:12:18 +03:00
Alexander Udalov d306d8a90e JVM IR: cast indy-SAM receiver type from KFunctionN to FunctionN 
#KT-46512 Fixed
 2021-05-07 00:53:31 +03:00
Mads Ager b6fa28ea81 [JVM_IR] Deal with inline-class arguments in large arity lambdas. 
FunctionNVarargBridgeLowering checked the name of the method instead
of whether the method overrides an invoke method. That doesn't work
when the name of the function gets mangled because of inline class
arguments.

Fixed KT-45084.
 2021-05-06 17:27:59 +02:00
pyos 1181854bd6 Add a couple more tests for array constructors and references 2021-05-06 18:25:56 +03:00
pyos 9f53d70109 JVM_IR: move ArrayConstructor below function reference phases 
This allows taking function references to inline array constructors.

Also, redundant classes are no longer generated when function references
are passed as arguments to the array constructors.

 #KT-46426 Fixed
 2021-05-06 18:25:53 +03:00
Mads Ager 6d9f02cfc6 [JVM_IR] Fix range of locals in connection with finally blocks. 
For code such as:

```
try {
  var y = "y"
  for (i in 0 until 1) {
    return y
  }
} finally {
  println("finally")
}
```

The local variables `y` and `i` ended up covering the finally block as
well in the debugger.

This change splits the range of the locals so that they do
not cover the finally block.

This change does not change the inliner to do similar transformations,
so the range of locals is still wrong win finally blocks when inlined
into an inline function. Added a failing test to that effect.
 2021-05-03 07:43:37 +02:00
Dmitry Petrov 83e3a702c5 JVM_IR KT-46408 properly map fake overrides in method handles 2021-04-30 22:24:02 +03:00
Mikhail Glukhikh 80a449862e FIR: implement diagnostics for qualifier as stand-alone expression case 2021-04-30 17:59:45 +03:00
Mikhail Glukhikh 84ccf7bbb1 FIR: use Java 8 rules in not implemented checker 2021-04-30 17:59:38 +03:00
Victor Petukhov 7c62e9aecd Introduce warnings reporting by missed constraints because of incorrect optimization in the constraints processor 2021-04-30 15:46:04 +03:00
Andrey Zinovyev 287ff3ed55 [FIR] Fix for AugmentedArraySetCall expression type 
Hacky fix so it's type is Unit and not error
 2021-04-30 15:25:44 +03:00
Dmitry Petrov 72804d2e8c JVM_IR KT-45779 don't generate intrinsic method refs via invokedynamic 2021-04-30 12:45:11 +03:00
Andi Wenger f9d2ca68ce KJS IR: Fix KT-45738 - Consider recursive checkForPrimitiveOrPattern 
Recursive results from checkForPrimitiveOrPattern were ignored. If it found a case/condition that could not be optimized the resulting "false" was not propagated. This would lead to a "optimized" when without all conditions.
- see KT-45738
- The return is now lifted out of the when to make it more obvious what is going on.
- New test for mixed multiple conditions in when
 2021-04-29 23:52:01 +03:00
Ilmir Usmanov cb89bd0e13 Generate SuspendFunction marker interface if supertype is suspend 
Otherwise, is/as checks will not work.
 #KT-18707
 2021-04-29 17:11:31 +03:00
Steven Schäfer 84d78035e4 JVM IR: Avoid double mangling of function reference invoke methods 2021-04-29 13:10:43 +02:00
Alexander Udalov 276ffd5a4b JVM IR: fix ::class reference and type mapping for scripts 
#KT-46284 Fixed
 2021-04-29 00:33:34 +03:00
Dmitriy Novozhilov 07b15f9de6 [FIR] Support effective visibility with @PublishedApi 
#KT-46270 Fixed
 2021-04-27 18:39:10 +03:00
Dmitriy Novozhilov 32c3f85679 [FIR] Add inline checker for bodies of inline functions 
This checker doesn't support `@PublishedAPI` yet, so some BB tests for it
 were muted. #KT-46270
 2021-04-27 18:39:09 +03:00
Ilmir Usmanov 0c0710bb79 Fix outer class accesses inside suspendImpl functions 
We need to generate this$0 fields to get to the outer class and generate
accesses to these fields inside suspendImpl function.
 #KT-46214 Fixed
 2021-04-27 10:44:18 +02:00
Ilmir Usmanov dc2485ae71 Support suspend functions as superinterfaces 
Forbid mixing suspend and non-suspend functional supertypes.
Since JVM BE generates suspend functional types as non-suspend ones
with SuspendFunction marker interface, there is not way to distinguish
non-suspend functional type from suspend one if they are mixed.
 #KT-18707 Fixed
 2021-04-26 22:14:32 +02:00
Steven Schäfer a1c1a32515 JVM: Fix unsigned literals in API version < 1.5 2021-04-26 15:11:45 +03:00
Mark Punzalan 9a4742c08d FIR: Properly build nullable suspend function types, and aggregate 
modifiers and annotations within KtTypeReference/REFERENCE_TYPE nodes.
 2021-04-26 15:11:38 +03:00
Mikhail Glukhikh f278de8768 FIR: don't call toString() inside string concatenations 2021-04-26 12:50:55 +03:00
Igor Chevdar 7dcdbf283b [box-tests] Added test 2021-04-26 11:53:27 +05:00
Alexander Udalov 0c3f2eefe0 Migrate bytecodeListing tests to new test infrastructure 2021-04-23 17:58:16 +02:00
Ilmir Usmanov f7a9bc3521 Minor. Make test actually suspend and add a test without suspension 2021-04-23 03:50:08 +02:00
pyos 23ffbe4d9e JVM_IR: do not box inline classes in suspend multifile bridges 2021-04-23 03:49:47 +02:00
pyos 7d95943b8b JVM_IR: do not box inline classes in suspend synthetic accessors 2021-04-23 03:49:43 +02:00
Alexander Udalov b2005302dc JVM, JVM IR: erase generic SAM supertypes 
Also, do not try to use invokedynamic on SAM calls with intersection
types, because intersection type is not allowed as an immediate type
projection of a supertype, and constructing a fake override in
LambdaMetafactoryArgumentsBuilder led to an exception. This fixes the
problem which was worked around earlier in e6c089ef, effectively
reverting that commit.

The main motivation for this change is that LambdaMetafactory also
doesn't generate generic signature for SAM wrapper classes at runtime.
Since these classes are synthetic, nobody should rely on the fact that
they have generic supertypes, which was observable only via Java
reflection.

 #KT-46149 Fixed
 #KT-46238 Fixed
 2021-04-22 10:53:15 +02:00
Dmitry Petrov 5324cbe729 Minor: update tests 2021-04-21 19:10:32 +03:00
Steven Schäfer d3d4e94cd6 JVM: Fix constant folding for unsigned values 2021-04-21 19:10:31 +03:00
Dmitry Petrov 732405895f Add tests for KT-45308 2021-04-21 17:36:12 +03:00
Dmitry Petrov f228c3bb04 Add tests for KT-45236 2021-04-21 17:24:41 +03:00
pyos e6d923f65c FIR: rename HIDDEN to INVISIBLE_REFERENCE 
Some of them should be INVISIBLE_MEMBER though
 2021-04-21 16:18:21 +03:00
Alexander Udalov 7fe10e27a1 Minor, fix failing codegen tests on different JDKs 
#KT-46238
 2021-04-21 14:43:10 +02:00
Dmitry Petrov 59de12c609 Minor: mute bytecode text check in some box tests failing in FIR 2021-04-21 11:27:12 +03:00
Dmitry Petrov f519150c08 JVM_IR KT-46189 lower tailrec functions after local declarations 2021-04-21 11:27:11 +03:00
Alexander Udalov 21e9bd7ea2 Add regression tests for obsolete issues 
#KT-9304
 #KT-14961
 #KT-16549
 #KT-21080
 #KT-28234
 #KT-30102
 #KT-31994
 #KT-34291
 #KT-38099
 #KT-41174
 #KT-44622
 #KT-44701
 #KT-44781
 #KT-44849
 #KT-44978
 #KT-45081
 #KT-45286
 #KT-45383
 #KT-45444
 #KT-45907
 2021-04-20 21:24:29 +02:00
Alexander Udalov e6c089ef40 IR: fix SAM conversion for types with contravariant intersection argument type 
In the added test, the problem was that the SAM type as computed by
`SamTypeFactory.createByValueParameter` was `Consumer<{BaseClass &
BaseInterface}>`, which was latter approximated in psi2ir during the
KotlinType->IrType conversion to `Consumer<out Any?>` (here:
https://github.com/JetBrains/kotlin/blob/3034d9d791cf1f9033104e12448e0d262d3bc3ce/compiler/ir/ir.psi2ir/src/org/jetbrains/kotlin/psi2ir/generators/ArgumentsGenerationUtils.kt#L606),
because intersection type argument is approximated to `out Any?`.

To avoid this, replace intersection type in immediate arguments of a SAM
type with the common supertype of its components at the same place where
we're getting rid of projections.

 #KT-45945 Fixed
 2021-04-20 17:31:14 +02:00