Might be a decompiler bug Can you show the real bytecode her kotlinlang #announcements

Channels

100daysofkotlin

100daysofkotlin-2021

algolialibraries

android-architecture

android-databinding

androidgithubprojects

androidx-xprocessing

arrow-contributors

beepiz-libraries

cincinnati-user-group

codemash-precompiler

codingconventions

competitive-programming

competitivecoding

compose-android

compose-desktop

compose-ui-showcase

connect-audit-events

cscenter-course-2016

domain-driven-design

effective-kotlin

effectivekotlin

embedded-kotlin

event21-community-content

fb-internal-demo

fluid-libraries

framework-elide

getting-started

github-workflows-kt

graviton-browser

image-processing

intellij-plugins

intellij-tricks

introduce-yourself

java-to-kotlin-refactoring

jhipster-kotlin

jvm-ir-backend-feedback

k2-early-adopters

kotest-contributors

kotlin-by-example

kotlin-data-storage

kotlin-foundation

kotlin-in-action

kotlin-multiplatform-contest

kotlin-pakistan

kotlinforbeginners

kotlingforbeginners

kotlinprogrammers

kotlintest-devs

kotlinultimatechallenge

kotlinx-datetime

lambdaworld_cadiz

language-evolution

language-proposals

leedskotlinusergroup

library-development

machinelearningbawas

myndocs-oauth2-server

navigation-architecture-component

ohio-kotlin-users

pattern-matching

pocketgitclient

practical-functional-programming

prozis-android-backup

python-contributors

realworldkotlin

refactoring-to-kotlin

refreshversions

russian-kotlinasfirst

skia-wasm-interop-temp

sofl-user-group

spring-security

squarelibraries

stayhungrystayfoolish

strife-discord-lib

swagger-gradle-codegen

swiss-user-group

tricity-kotlin-user-group

Title

c

cbruegg

05/19/2019, 7:50 PM

Might be a decompiler bug. Can you show the real bytecode here?

i

itnoles

05/19/2019, 8:20 PM

I am seeing again with while, isNullOrEmpty and other functions

https://pastebin.com/JJnLJ3Hx here for bytecode for isNullOrEmpty that is showing boolean var3 = false; boolean var4 = false;

c

cbruegg

05/19/2019, 8:29 PM

Yeah, that's odd. Looks like it's indeed like this in the bytecode.

i

itnoles

05/19/2019, 8:30 PM

decompiler bug?

c

cbruegg

05/19/2019, 8:32 PM

No, I mean the useless variable indeed seems to be in the bytecode. Unless I'm missing something.

You can see

ICONST_0

twice followed by

ISTORE_3

/

ISTORE_4

to store the value

(=

false

) into the variables 3 and 4, but there's no

ILOAD_3

/

ILOAD_4

to ever load it again.

m

Marko Mitic

05/19/2019, 8:35 PM

I've seen medum article recommending using if instead of let because of extra variable in bytecode.

c

cbruegg

05/19/2019, 8:36 PM

That's not good advice, and generally the article is of low quality. This should be something that's very easy to optimize for the JVM or ART. And even if not, it's a stack allocation, which are really cheap.

For reference, this is the article: https://medium.com/@elye.project/kotlin-dont-just-use-let-7e91f544e27f?sk=6d8d04c5c00613f20e20fe057b3ebd51

m

Marko Mitic

05/19/2019, 8:38 PM

I guess kotlin compiler tests don't check for scenarios like these and this slipped in by accident

I mean, this bytecode is functionally correct, but it's not optimal

c

cbruegg

05/19/2019, 8:40 PM

It's not optimal, but I would be surprised to see an actual performance impact.

i

itnoles

05/19/2019, 8:51 PM

Android Studio has Kotlin Bytecode with Optimizations checked, it still show same as unoptimizations

c

cbruegg

05/19/2019, 9:48 PM

That doesn't include optimizations performed at runtime by ART or the JVM.

i

itnoles

05/19/2019, 9:49 PM

odd

g

gildor

05/19/2019, 11:54 PM

This is not a bug, this variable used for inline functions debugging See: https://kotlinlang.slack.com/archives/C7L3JB43G/p1558022163043700?thread_ts=1557907293.041100&cid=C7L3JB43G

c

cbruegg

05/20/2019, 8:47 AM

Ah, thanks! That's clever.

3 Views

#announcements Join Slack