Windows (native build) codepage issues

[Nik-mmzd]

Example code:

# main.kt
import okio.*
import okio.Path.Companion.toPath

fun main(args: Array<String>) {
    FileSystem.SYSTEM.createDirectory("тестовая строка".toPath())
}

# build.gradle.kts
plugins {
    kotlin("multiplatform") version "1.9.23"
}

repositories {
    mavenCentral()
}

kotlin {
    mingwX64 {
        binaries {
            executable()
        }
    }

    sourceSets {
        commonMain {
            dependencies {
                implementation("com.squareup.okio:okio:3.9.0")
            }
        }
    }
}

Provided code creates directory named тестовая строка (cp1251 represenation of utf-8 string) instead of тестовая строка.

This applies to other functions as well: FileSystem.SYSTEM.exists(), FileSystem.SYSTEM.write(), etc.

Tested on Windows 11 and Windows 10. "Unicode (beta)" in "Administrative Locale Settings" should be off.
When "Unicode beta" in "Administrative Locale Settings" is on, the issue is not reproducible.

[swankjesse]

Thanks for reporting this.

I think there’s likely two possible fixes; one where we transcode the native path to UTF-8 first, and one where we don’t do that.

I’m curious what happens for Java, which uses UTF-16 strings for paths. It might just work because it does string encoding before it makes system calls.

[Nik-mmzd]

Windows, same code, JVM target, JVM 17, all seems to be OK

[Nik-mmzd]

I think I should note that cp1251 is locale-dependent codepage and may be changed by user in "Administrative Locale Settings". Current codepage can be retrieved using GetACP call.

[fzhinkin]

I’m curious what happens for Java, which uses UTF-16 strings for paths. It might just work because it does string encoding before it makes system calls.

Java uses API that accepts UTF-16 strings (LPCWSTR) as paths (CreateFileW, DeleteFileW, etc.): https://github.com/openjdk/jdk/blob/master/src/java.base/windows/native/libnio/fs/WindowsNativeDispatcher.c