FqName/FqNameUnsafe tend to eagerly compute things like
safe/parent/shortName which take a lot of space.
This change helps to lower strongly reachable memory just before codegen
in JVM IR by about 5% on a relatively big project.
this change fix issue with inlining lambda in inline function which(function) inlined from other library.
E.g.
> cat i-lib.kt
class _special_class(val v:Int)
class _special_class1(val v:Int)
class __helper(val v:Int)
inline fun foo(h: __helper): Int {
val sum = h.op {
_special_class(it.v)
}.v
return sum
}
inline fun __helper.op(block:(_special_class1) -> _special_class) = block(_special_class1(v))
> cat i-main.kt
fun main() {
val h = __helper(42)
println(foo(h))
}
Here how the incorrect parent affects debug information:
(lldb) target create "program.kexe"
Current executable set to '/Users/minamoto/ws/kotlin-native/program.kexe' (x86_64).
(lldb) command source -s 0 'i-test.lldb'
Executing commands in '/Users/minamoto/ws/kotlin-native/i-test.lldb'.
(lldb) b i-lib.kt:8
Breakpoint 1: where = program.kexe`kfun:#main(){} + 435 [inlined] foo + 98 at i-main.kt:3, address = 0x00000001000540e3
(lldb) r
Process 70550 stopped
* thread #1, queue = 'com.apple.main-thread', stop reason = breakpoint 1.1
frame #0: 0x00000001000540e3 program.kexe`kfun:#main(){} [inlined] foo at i-lib.kt:9:7
6 inline fun foo(h: __helper): Int {
7 val sum = h.op {
8 _special_class(it.v)
-> 9 }.v
^
10 return sum
11 }
12
Target 0: (program.kexe) stopped.
the parent of lambda is i-main.kt instead of i-lib.kt, and offsets calculated against wrong file.
Here is behaviour with fix:
(lldb) target create "program.kexe"
Current executable set to '/Users/minamoto/ws/.git-trees/minamoto/debug-info/subprograms-with-missed-scopes/program.kexe' (x86_64).
(lldb) command source -s 0 '/Users/minamoto/ws/kotlin-native/i-test.lldb'
Executing commands in '/Users/minamoto/ws/kotlin-native/i-test.lldb'.
(lldb) b i-lib.kt:8
Breakpoint 1: where = program.kexe`kfun:#main(){} + 337 [inlined] <anonymous>_2 at i-lib.kt:14, address = 0x0000000100054bb1
(lldb) r
Process 70560 stopped
* thread #1, queue = 'com.apple.main-thread', stop reason = breakpoint 1.1
frame #0: 0x0000000100054bb1 program.kexe`kfun:#main(){} [inlined] <anonymous>_2 at i-lib.kt:8:24
5
6 inline fun foo(h: __helper): Int {
7 val sum = h.op {
-> 8 _special_class(it.v)
^
9 }.v
10 return sum
11 }
Target 0: (program.kexe) stopped.
In the newly added test, prior to this change, JVM IR was generating
DefaultImpls classes with calls to things like
`kotlin/collections/MutableList$DefaultImpls.spliterator` and other
default methods present in JDK 8+. This obviously didn't make much
sense. Although these weren't explicitly mentioned anywhere in the
bytecode, they caused some validation tools to report errors (e.g.
animalsniffer used in arrow).
In particular, the current line numbers could lead to stepping
into the catch handler even when the code in the try did not
throw an exception.
This was caused by the code materializing the final value having
the catch line number. This patch delays the materialization
until the line number of the usage has been emitted.
A mechanism that allows kotlinx.serialization plugin to preserve the
correct (program) order of properties after serializing/deserializing
descriptors to kotlin metadata, which is needed for correct and stable
json serialization of class hierarchies in incremental/multi-module scenario.
It uses protobuf extensions.
Georgy Bronnikov