Currently the perl wrapper around ncurses behaves strangely for UTF-8 locales, and probably for any other multibyte-locales as well. I managed to work around the issue by exporting LD_PRELOAD=/lib/libncursesw.so but it would be nice if the ebuild could link against ncursesw instead of ncurses if the unicode USE flag is enabled. Or would you consider this an upstream issue?
Created attachment 120631 [details] Sample program This is a stripped down version of a program that caused me trouble. It simply allows you to enter one line of text, and provides some cursor movement as well as a 72 characters blue background. You will notice that when you enter some non-ASCII characters like German umlauts, you will get garbage on the screen, the blue area will change in size, and cursor movement will stop working properly. Experiment a bit, it might be that you won't get all these effects immediately.
I filed an upstream RFE as well, and adjusted the URL field here. http://rt.cpan.org/Public/Bug/Display.html?id=27335
Created attachment 121329 [details, diff] Changes to Curses-1.15.ebuild adding unicode USE flag While the upstream bug is concerned with whether libncursesw should be the default, Gentoo users specifying the unicode USE flag clearly state their opinion on this, I believe. I just created a modified ebuild that will use the wide character enabled versions of ncurses if the unicode USE flag is given.
My patch from comment #3 works well here in daily use. I would assume it fit for inclusion in an official ebuild.
Rejected usptream -> WONTFIX.
Seems this got fixed after all, in bug #217096. I'm happy for it. :-)
