Windows unicode output doesn't seem to work

[cbm755]

In Octave 3.8.1 (MXE), I can't even see UTF-8 chars in the editor.

pretty(sin(sym(x)/2)) just shows garbage chars.

[cbm755]

https://savannah.gnu.org/bugs/?43099

[oheim]

UTF-8 output works for me in version 2.3.0, at least in octave-cli (version 4.0.2). In octave-gui I could not set the terminal to UTF-8 mode. However, unicode output of the symbolic package works only once. Then, any output from Octave is disabled.

To reproduce:

• Run octave.bat --no-gui
• Set terminal font to Lucida Console
• Run system ("chcp 65001"); in Octave
• Run pkg load symbolic and sympref display unicode

After these steps the first output with unicode from the symbolic package seems to break stdout (disp ("foo") doesn't work anymore). Stderr seems to be working (error ("foo") prints error: foo).

[cbm755]

If the unicode output does not come from symbolic, does the same happen?

[oheim]

Yes, this is definitely an upstream problem.

I have printed some unicode characters coming from __makeinfo__ (using @documentencoding UTF-8, e. g. help infsup with the interval package). There are several issues:

1. The graphical Terminal (octave-gui) does not respect the code page and interprets every byte as a (ANSI?) character.
2. The console Terminal (octave-cli via cmd.exe) is limited to the characters supported by the font. Unsupported characters, e. g. subset (⊂) or right arrow (⇒) are printed as squares.
3. I get a warning message during the call to makeinfo: Use of uninitialized value in lc at /usr/lib/perl5/5.8/utf8_heavy.pl line 123.
4. By default the help command produces a lot of chunked output and resets the screen several times. The pager interprets some unicode characters as control sequences (I have also seen this behavior on android with GNURoot Octave). A lot of information is not printed, but the console output is still working in my example afterwards.
5. When i disable the pager (more off), the full help text is printed correctly until output stops and is broken after that point. In help.m the command printf ("\n%s\n%s", text, __additional_help_message__ ()); prints only the text, but no additional help message—not even the additional \n character. Some unicode characters break Octave's stdout inside printf.

[cbm755]

Thanks for all that. Should you re-post all that upstream? Or at least ping there with a link to your sleuthing.

FYI: In the upstream bug https://savannah.gnu.org/bugs/?43099, I posted a simple m-file which defines a bunch of utf-8 strings. Might be easier than messing with __makeinfo__.

[cbm755]

BTW, re-reading your earlier post, I'm not clear on something: you say "works for me with octave-cli" but then your "steps to reproduce" include octave.bat --no-gui. So which is it?

And are 3., 4., and 5. above on gui only? Do they work fine in cmd.exe (other than font issues)?