Adam Borowski, on sam. 25 févr. 2017 22:31:57 +0100, wrote: > On Sat, Feb 25, 2017 at 08:05:32PM +0100, Samuel Thibault wrote: > > That's expected: some characters have double-width, others have > > zero-width. > > My test sheet accounts for that: it includes only wcwidth()==1 and 2 > characters (recently updated for unstable's glibc).
Ah, ok, sorry I didn't actually check :) > > But for characters that have single-width, they are really > > aligned with a proper fixed-width font. > > Depends on your software. xterm, libvte, pterm, rxvt-unicode get it right, > mousepad, firefox, chromium and Microsoft Edge don't. Ok, but would one really look at the output of text-gnuplot in such software? Cases where I happened to use the text-gnuplot were always inside an xterm or such. In other words: it's not because in some odd cases things go wrong that one shouldn't implement the often-used case. > > > gnuplot relies on being able to place labels within the image, which works > > > for ASCII and maybe Latin/Greek/Cyrillic but, except for most terminals, > > > not > > > for anything else. > > > > Then gnuplot is missing taking into account the value returned by > > wcwidth() (0, 1, 2, ...), that's the bug. > > I don't know whether gnuplot is doing it correctly, I haven't tested -- but > even if it does, the output will be misrendered by browsers. But if the output is to be rendered in a browser, one would use a png or svg output :) Samuel
