Wrong char count in ServletOutputStream with non-ASCII characters skips content

[barulheira]

The ServletOutputStream skips the last chars when using print or println to print content with multi-byte characters in UTF-8.
The amount of chars skipped matches the difference between char count and byte count, on each print or println.
It doesn't matter whether it's done in the end of the response.
Changing to another encoding (e. g. ISO-8859-1) doesn't fix this.
No problem when using write instead of print.
No problem when using only ASCII characters.
No problem when using PrintWriter response.getWriter().

Steps to Reproduce
Get a response.getOutputStream() and print some text with non-ASCII content.
On each print or println, some characters will be skipped in the end.
out.println("'ATENÇÃO!'"); // Skips last 2 characters
will print out:
'ATENÇÃO
A minimal web app is attached.

Expected behavior
It should print: 'ATENÇÃO!'

Diagnostic information:
Open Liberty 21.0.0.7

Additional context
This s.length() here may be the cause:

open-liberty/dev/com.ibm.ws.webcontainer/src/com/ibm/ws/webcontainer/osgi/response/WCOutputStream.java

Line 199 in 16559e1

this.output.write(s.getBytes(), 0, s.length());

bug.zip

Thanks @Thihup

[wtlucy]

Thanks for reporting this @barulheira - you're correct, the WCOutputStream should be writing the length of the encoded bytes rather than the length of the original string.

[Thihup]

FYI, there are more two suspicious calls here:

https://github.com/OpenLiberty/open-liberty/blob/integration/dev/com.ibm.ws.webcontainer.servlet.3.1/src/com/ibm/ws/webcontainer31/osgi/response/WCOutputStream31.java#L400

https://github.com/OpenLiberty/open-liberty/blob/integration/dev/com.ibm.ws.webcontainer.servlet.3.1/src/com/ibm/ws/webcontainer31/osgi/response/WCOutputStream31.java#L552

[wtlucy]

This is fixed with #18022. @barulheira if you'd like to verify the behavior you can check out this build, otherwise the fix will be in the 21.0.0.9 release.