Screen not cleared after quitting F1 man

[YaLTeR]

fish, version 3.2.1, same issue with default config.

On VTE, xterm, urxvt, Konsole (but not Alacritty) exiting the F1 man page with q doesn't move the prompt down and leaves the text on screen, so subsequent commands will overlap with this text.

asciinema recording

I made a VTE issue where the maintainer came to the conclusion it is an issue with the sequence fish uses:

Hmm. So this exits the alternate screen via RM_DEC 1049 which does not clear the screen. So unless the result differs between vte and xterm, this is a problem in alacrity.

[zanchey]

fish doesn't send this sequence, it is most likely your pager. The output of echo $MANPAGER $PAGER might be helpful.

[YaLTeR]

fish doesn't send this sequence, it is most likely your pager. The output of echo $MANPAGER $PAGER might be helpful.

Empty in both Alacritty and VTE.

[faho]

I made a VTE issue where the maintainer came to the conclusion it is an issue with the sequence fish uses:

fish does not use the sequence here, your pager does.

It appears man uses less if $PAGER and $MANPAGER are unset, so I'd look at $LESS for any options you've set.

Either way, this isn't a fish issue, we just call man.

[YaLTeR]

It appears man uses less if $PAGER and $MANPAGER are unset, so I'd look at $LESS for any options you've set.

Also empty.

Either way, this isn't a fish issue, we just call man.

Executing the man command regularly and quitting out clears the screen, however using fish's F1 doesn't for some reason.

[YaLTeR]

Here's a script(1) recording of opening and closing man by itself, then opening and closing man using F1.

regular-then-fish.log

[faho]

Lovely: Apparently less won't use/clear the alternate screen if stderr isn't connected to a terminal.

See e.g. less file 2>/dev/null.

In this case we want stderr to be silenced because we don't actually know if the man page exists.

I'm not sure what the alternative is.

[YaLTeR]

See e.g. less file 2>/dev/null.

Yep, that also reproduces the issue.

[faho]

Turns out the alternative is awful:

    if man thing &>/dev/null
       man thing

But still less awful than us e.g. forcing the alternate screen with tput smcup, tput rmcup.

[YaLTeR]

Seems like it works fine after that commit, thanks!

[faho]

This could be caused by gwsw/less@4d4e6f4 checking fd 2 for the tty to connect to. That means if you redirect fd 2 less won't believe it's in a terminal in some parts (but it obviously handles input and normal output, so it seems to be inconsistent).

I believe this is a bug in less (it's checking a different fd than it is writing to!), if anyone wants feel free to report it.