On 6/19/20 7:53 PM, Bryan Henderson wrote: > I tracked it down to an Ncurses change. > > Bash/readline attempts to move the cursor left by calling Ncurses (libtinfo) > function 'tputs' with ASCII BS (ctl-H) as the argument. The function of > 'tputs' is to write stuff to the terminal with delays added, as required by > early printing terminals. > > In the past, if TERM were set to something undefined, tputs would go ahead and > write the BS to the terminal, but in the libtinfo linked to Bash on my Debian > 10 system, tputs fails (returns failure code without writing anything to the > terminal) if TERM is set to something undefined. (It still works if TERM > isn't defined at all, though). Bash ignores the failure and continues, > writing a space to the terminal, followed by another failing call to 'tputs' > for the final backspace.
This is consistent with what I discovered, which is that tputs does not call the character output function passed as the third argument at all. I don't know why the ncurses maintainers changed this in version 6, though I suspect they had some rationale for it at the time. It looks like a departure from historical versions of curses/ncurses. The historical behavior (what readline expects) of tputs is to set the amount of padding required to 0 if the terminal type is unknown, which basically makes it an indirect per-character function call. From what I've seen, it fails only if the string parameter is NULL. -- ``The lyf so short, the craft so long to lerne.'' - Chaucer ``Ars longa, vita brevis'' - Hippocrates Chet Ramey, UTech, CWRU c...@case.edu http://tiswww.cwru.edu/~chet/