Attachment Project1.exe added by nishimotz on 2014-02-24 12:44
Description:
Project1.exe developed with VB6

Comment 1 by jteh on 2014-02-24 21:26
I haven't looked into this yet, but it may be that this is a bug in the Edit API that we can't work around. We rely on the Edit API to give us correct line offsets. Calculating them ourselves isn't an option; it would be way too slow with a lot of content.

Comment 2 by nishimotz on 2014-02-24 22:43
It seems that source/NVDAObjects/window/edit.py uses EM_LINELENGTH for retrieving the length of line.

Microsoft document says "For ANSI text, this is the number of bytes".

http://msdn.microsoft.com/en-us/library/bb761613.aspx

Is that related to this issue?

Comment 3 by jteh on 2014-02-24 23:22
If it's an ANSI Edit control, yes. However, it's quite tricky to detect the encoding for an ANSI control. We'd probably have to go in-process (in-thread actually) and call GetThreadLocale.

Comment 4 by nishimotz on 2014-02-25 00:55
It seems that NVDAObjects/window/init.py already uses IsWindowUnicode() to setup isWindowUnicode property.

How about using this property?

Comment 5 by jteh on 2014-02-26 04:44
That only tells us whether the window is Unicode or ANSI. If ANSI, it doesn't tell us what encoding is being used, so we don't know anything about the characters that will occur.

@jcsteh While I am not too sure about the accuracy of my interpretation, your comments seem to suggest that this ticket might be a cantfix. Could you please clarify/elaborate?

It needs investigation, but it's possibly fixable. See #3918 (comment) regarding using GetThreadLocale.

@nishimotz, is this still something that's relevant for Japanese users?

Project1.zip

Original project1.exe file is attached as the zip file.

Tested with the following setup and confirmed that the issue is not resolved.

• nvda_snapshot_master-14295,890528bd.exe
• Windows 10 Version 1703 64bit Japanese

Original reporter who provided the executable file has migrated to the modern development tools.
However, some NVDA users in Japan still prefer old applications which use ANSI APIs.

@nishimotz testing with NVDA 2018.4 on Windows 10 1809 64bit german, I cannot reproduce this issue.
Are you able to reproduce it still?

It still occurs with NVDA 2018.4 and Windows 10 1803 64bit Japanese.

I think it relates to CP 932 (Shift_JIS) of Windows default encoding.