Use only valid TERM environment with screen ("screen-256color" might fall back to "vt100")

[Daniel Hahler]

bin/byobu uses "screen-256color" if the terminal supports 256 colors (like xterm-256color does).

However, if there's no terminfo installed for screen-256color, screen will fallback to vt100, which means no colors and a lot less goodies.

Therefore, before byobu does SCREEN_TERM="-T screen-256color" it should check if that's a valid value, e.g by testing with "TERM=$to_test tput longname 2>/dev/null".

[Dustin Kirkland  (kirkland)]

Hmm, I don't quite understand what situation we're trying to protect against here, that we're not already handling.

What's the actual problem, and how do I reproduce it?

We currently do this:
                # Check if our terminfo supports 256 colors
                if command -v tput >/dev/null; then
                        if [ "$(tput colors 2>/dev/null || echo 0)" = "256" ]; then
                                SCREEN_TERM="-T screen-256color"
                        fi
                fi

What situations are not handled/detected by that?

[Daniel Hahler (blueyed)]

There's a problem when the current terminal supports 256 colors, but there's no terminfo file for screen-256color.

To reproduce the case of a missing terminfo (e.g. old system etc), just set SCREEN_TERM="-T screen-256color-notfound" in the snippet above.
Screen will then fallback to using vt100 (at least the screen I was using when encountering this).

[Dustin Kirkland  (kirkland)]

Hopefully this should be fixed by the commit for Bug #908766, which changes TERM=screen. From everything I can tell, this doesn't break 256 colors (I can see 256 colors in various applications like vim, mutt, and 256colors2.pl).

[Daniel Hahler (blueyed)]

Unfortunately it is not fixed for me (showing only 8 colors now in Vim). I've commented on bug #908766.

[Daniel Hahler (blueyed)]

Re-opening this bug: the issue still would hold if "screen" was not available via terminfo/termcap.

While that is far more unlikely, it still holds, especially given that we might use "screen-256color(-bce)" again.

[Dustin Kirkland  (kirkland)]

From what I can tell, Bug #900309 and Bug #908766 are in opposition, and I have no solution yet that solves these two requests and causes no regressions.

For that reason, I'm downgrading this bug from Triaged to Incomplete.

I'm open to merge proposals or patches. But in my heavy, daily usage of byobu, I'm not seeing any of the problems in either of these bugs. That doesn't mean they don't exist, they're just off my beaten path.

[Launchpad Janitor (janitor)]

[Expired for byobu (Ubuntu) because there has been no activity for 60 days.]