fix(prompt): fix prompt output of non-ascii characters on windows

[kt3k]

Deno.stdout.writeSync(new TextEncoder().encode(str)) doesn't seem working for non-ascii characters on windows, but Deno.core.print(str) does seem working. See the comment #8179 (comment) .

So this causes the issue #8179.

This PR replaced the use of Deno.stdout.writeSync in prompt with Deno.core.print and avoided the issue.

I checked the fix on window server 2019 machine.

fixes #8179

[kt3k]

I added flush here because print! macro doesn't seem flushing automatically. rust-lang/rust#23818

[bartlomieju]

@kt3k this is a bit strange - Deno.core.print() works for me without a problem without the flush() and it's one of purposes of Deno.core.print not to include new line by default. Could you explain the reason for this change?

[kt3k]

@bartlomieju print! is line-buffered as discussed in rust-lang/rust#23818. So if we use it like Deno.core.print('hello'), the string doesn't show in stdout until new line is printed or process is finished. (Try the script Deno.core.print('xyz'); setTimeout(() => {}, 4000); for example).

If we don't add flush() here, prompt works like the below:

> prompt('hi!')
hello
hi! "hello"

(The answer and question are reversed here.) I think this isn't what the users expect.

If we add flush(), it works like:

> prompt('hi!')
hi! hello
"hello"

[lucacasonato]

@kt3k I can not reproduce that:

[kt3k]

In repl, the returned value is printed with line break (such as "1\n") and this line break should trigger flushing of stdout.

Try the following:

deno eval "Deno.core.print('xyz'); setTimeout(() => {}, 4000);"

[wperron]

Validated on Windows

PS C:\Users\wperron\github.com\kt3k\deno> .\target\release\deno
Deno 1.5.0
exit using ctrl+d or close()
> prompt("ÈÉÊàààà");
ÈÉÊàààà
null
> prompt("ÈÉÊàààà hello");
ÈÉÊàààà hello
null
> prompt("äÄăåČćçĎʤʣĎĔĤʞĿŁÒÔ")
äÄăåČćçĎʤʣĎĔĤʞĿŁÒÔ
null
>

[piscisaureus]

It'd be nice if stdout (when attached to a terminal) actually behaved as expected and interpreted the byte stream as UTF8-encoded text.

But this fix is also legit, and the thing it fixes is really an eyesore.

So LGTM from me.