[Bug 49378] New: kernel32.WriteFile with multi byte characters reports incorrect number of written bytes

[WineHQ Bugzilla]

https://bugs.winehq.org/show_bug.cgi?id=49378

Bug ID: 49378
           Summary: kernel32.WriteFile with multi byte characters reports
                    incorrect number of written bytes
           Product: Wine
           Version: 5.10
          Hardware: x86-64
                OS: Linux
            Status: UNCONFIRMED
          Severity: minor
          Priority: P2
         Component: kernel32
          Assignee: [e-mail]
          Reporter: [e-mail]
      Distribution: ---

Created attachment 67438
  --> https://bugs.winehq.org/attachment.cgi?id=67438
Wine test case to trigger the bug

Writing utf-8 multibyte characters to stdout using stdio led to uncovering this
bug. Attached test case is a simple way to trigger the underlying issue.

Minimal steps leading to issue:
1. Set console output to CP_UTF8
2. Use WriteFile with multibyte characters

Expect outcome:
WriteFile reports number of bytes written.

Actual outcome:
WriteFile reports number of characters written.

Other observations while writing test case:
I first run same test case using kernelbase.WriteFile. Base version seems to
behavior differently because my test case passed. I only then noticed the
important detail that there is multiple WriteFile entry points.

[Bug 49378] kernel32.WriteFile with multi byte characters reports incorrect number of written bytes

[WineHQ Bugzilla]

https://bugs.winehq.org/show_bug.cgi?id=49378

--- Comment #1 from Rafał Mużyło  ---
...it might be completely unrelated, but (mentioned this awhile ago on irc)
I've recently ran into an interesting case: a Unity game, that was crashing due
to 'address space exhaustion', that before the crash was printing a stack trace
several times.

The potentially relevant part is that the game was run under LANG=ja_JP.utf8
and the stacktrace contained mojibake...

...but only if run as 'LANG=ja_JP.utf8 wine <executable>', if I redirected the
log via 'LANG=ja_JP.utf8 wine <executable> 2>&1 | tee crash.log', both the
crash log and the console output were mojibake-free.

[Bug 49378] kernel32.WriteFile with multi byte characters reports incorrect number of written bytes

[WineHQ Bugzilla]

https://bugs.winehq.org/show_bug.cgi?id=49378

--- Comment #2 from Pauli Nieminen  ---
(In reply to Rafał Mużyło from comment #1)
> ...it might be completely unrelated, but (mentioned this awhile ago on irc)
> I've recently ran into an interesting case: a Unity game, that was crashing
> due to 'address space exhaustion', that before the crash was printing a
> stack trace several times.

If output uses stdio then https://source.winehq.org/patches/data/186903 is a
potential fix for multiple prints from a buffered stdio output. The patch was a
response to a related issue report at mailing list.

But even with that patch I can see some output errors from buffered stdio. I
think there is more work to be done for stable multibyte output using msvcrt
stdio. But issues are relative minor after patch 186903 and fix for this bug.

It is easy to make a hack fix for this bug. Proper fix would be fairly easy
using mbrlen or mbrtowc. But I think those aren't available in kernel32.dll.

[Bug 49378] kernel32.WriteFile with multi byte characters reports incorrect number of written bytes

[WineHQ Bugzilla]

https://bugs.winehq.org/show_bug.cgi?id=49378

François Gouget  changed:

What    |Removed                     |Added
----------------------------------------------------------------------------
           Keywords|                            |source, testcase
                 CC|                            |[e-mail]