[std] Fix unicode passed through stdout/stderr

[tobil4sk]

These patches are needed to restore ipc sys tests in the haxe repo.

[ncannasse]

Does the unit tests actually test writing on stdout/stderr with sys.io.File.stdout().write ? Because the code you are removing is intended for that. Also does it work with a not Unicode terminal like cmd ? You should be able to do hl app.hl >file.txt and get proper unicode this way.

[tobil4sk]

Does the unit tests actually test writing on stdout/stderr with sys.io.File.stdout().write ?

Yes, it is tested in these cases:

https://github.com/HaxeFoundation/haxe/blob/d55a4dd07b15b2c00ddb5d7a62e378ed7369cb01/tests/sys/src/TestUnicode.hx#L297-L299

https://github.com/HaxeFoundation/haxe/blob/d55a4dd07b15b2c00ddb5d7a62e378ed7369cb01/tests/sys/src/UtilityProcess.hx#L184

Without this file_write change, it writes garbled text to stderr/stdout.

Because the code you are removing is intended for that.

Note, hl_sys_print used to use a similar thing but it caused issues so it got disabled already: 9d21b13

Also does it work with a not Unicode terminal like cmd

I don't think this fixes how unicode is displayed on the console, it only makes sure that the correct utf-8 bytes are put into the buffer (so piping to a file or another process should work). For displaying it in cmd or powershell, that is a different issue and will probably require a fix like hxcpp: HaxeFoundation/hxcpp#1307

You should be able to do hl app.hl >file.txt and get proper unicode this way.

This should work, I will double check this.

[tobil4sk]

function main() {
    trace("Hello world!");
    Sys.println("éá🙄🦉");
    Sys.stdout().writeString("éá🙄🦉");
}

Without this patch hl hello.hl > file.txt produces:

HelloWorld.hx:3: Hello world!
��????
éá🙄🦀͍͍ﯮﯮﯝݝ

With this patch hl hello.hl > file.txt produces:

HelloWorld.hx:3: Hello world!
éá🙄🦉
éá🙄🦉

[yuxiaomao]

I need this fix, and I confirm that it fix the described bug. Utf-8 text are displayed correcly in git-bash (not in cwd, as expected).