On Wed, Jul 31, 2019 at 07:57:00PM +0200, Sven Joachim wrote: > On 2019-07-31 18:38 +0200, Samuel Hym wrote: > > > Package: ncurses-base > > Version: 6.1+20190713-1 > > Severity: normal > > > > Dear Maintainer, > > > > Since updating ncurses from 6.1+20181013-2 to 6.1+20190713-1, I get > > corruption when I run mutt in a tmux. Since reverting that update > > makes this bug disappear, I report it against ncurses-base. I think I > > saw such glitches in other contexts than in mutt, but that’s where I > > got it to reproduce. > > > > To reproduce, run in a tmux: > > > > mutt -F /dev/null -e 'set pager_index_lines=6' -f some-mbox > > > > select the 5th message from the top of the screen and try to return > > from the pager to the index. Instead of showing the clean index, I see > > some leftovers from the pager view, including the separation line > > between the 6 index lines and the pager. > > Refreshing the screen with Ctrl-L clears the glitches away. > > Thanks for these detailed instructions. I can confirm this glitch, and > also that downgrading ncurses-base to 6.1+20181013-2 makes it go away. > > Since tmux sets TERM=screen by default, and there have not really been > any changes to screen in ncurses' terminfo.src file, I wondered what > happened. Here is what I found out so far: > > ,---- > | $ dpkg-deb -x /var/cache/apt/archives/ncurses-base_6.1+20181013-2_all.deb > buster > | $ dpkg-deb -x /var/cache/apt/archives/ncurses-base_6.1+20190713-1_all.deb > sid > | $ infocmp -x -A buster/lib/terminfo -B sid/lib/terminfo screen screen > | comparing screen to screen. > | comparing booleans. > | comparing numbers. > | comparing strings. > | rin: NULL, '\E[%p1%dT'. > `---- > > So rin=\E[%p1%dT has been added in sid, apparently because of the > following change in the 20190630 patchlevel: > > ,---- > | + add a check in tic for paired indn/rin > `---- > > It looks like tmux does not quite cope with this correctly or so.
Nicolas' followup indicates that tmux didn't implement this.
The problem was that tmux's terminal description inherits from (use=)
the screen description, and that a maintenance update for the latter
exposed the problem.
That in turn was prompted by another developer wanting to inherit
from "xterm", but since testing showed _that_ to be not accurate --
no surprise -- the building block approach was the way to go.
The indn/rin pair hadn't been done (noticeable when doing this),
and since inspection of screen's source-code showed it supported
this
case 'S': /* code from a 97801 term / DEC vt400 */
ScrollRegion(win, a1 ? a1 : 1);
break;
case 'T': /* code from a 97801 term / DEC vt400 */
case '^': /* SD as per ISO 6429 */
ScrollRegion(win, a1 ? -a1 : -1);
break;
}
then this applied:
ecma+index|ECMA-48 scroll up/down,
indn=\E[%p1%dS, rin=\E[%p1%dT,
I overlooked the fact that there's no INPUT_CSI_SD in tmux at the moment:
case INPUT_CSI_SU:
n = input_get(ictx, 0, 1, 1);
if (n != -1)
screen_write_scrollup(sctx, n, bg);
break;
The fix would be add "rin@" to the tmux entry (along with a note).
I'll do that...
--
Thomas E. Dickey <dic...@invisible-island.net>
https://invisible-island.net
ftp://ftp.invisible-island.net
signature.asc
Description: PGP signature
