Thomas Wolff writes:
I guess it's more about the configuration of tmux. There is in fact a
cursor style setting sequence that mintty newly supports.
Please make a terminal log and check whether ^[[34h (ESC [ 3 4 h)
appears during tmux initialization.
I figured as much meanwhile, but I've not had luck in finding a
description of that sequence until much later today (in the teraterm
documentation). I also fell into the trap again that you really can't
recompile a terminfo entry and expect it to work until you've closed all
mintty processes.
If so, however, the assumption is that tmux sends it on purpose, so
the blame is on tmux :/
Not tmux, what you need to blame is the terminfo entry for
screen-256color again, specifically the cursor_normal variable. There
are a bunch of other terminals using the same sequences though. But I
need to set the terminal to this exact string or colors inside screen
(e.g. Emacs) don't work correctly (it falls back to some lower number of
colors and gets completely illegible with the default theme). Anyway,
since I already needed to patch the stupid italics/bold swap in that
terminfo entry, I just patched out this bit of nonsense as well.
About an option to suppress dynamic changes of cursor style, that
might indeed be useful. I assume, though, that it should uniformly
also suppress the DEC sequence (DECSCUSR). Or should different cursor
attributes be addressable separately? (shape, blinking, colour??, even
hiding???) Design proposals welcome.
Well, one simple switchbox to "keep the cursor fixed at these settings"
next to the other cursor settings would have saved me a few hours of
heartburn. Allowing to make each of them separately sticky would be
nicer, but it's probably also a touch too much (besides, there's be a
ton of other options that would then need the same treatment for the
same reason).