Rendering failure with combined DCH + line scroll

[Paul "LeoNerd" Evans]

$ echo -en "\e[25H\r\nABCDE\b\b\b\e[2P"; echo

renders incorrectly

$ echo -en "\e[25H\r\nABCDE\b\b\b\e[2P"; sleep 2; echo

renders correctly.

Seems just to be a surface GTK problem, because the underlying VTermScreen buffer is fine; using mouse-selection or scrolling away+back re-renders it and now displays correctly.

[Paul "LeoNerd" Evans (leonerd)]

A little bit of analysis on these (by stderr printfs() in pangoterm.c) reveals some interesting differences between these.

OK case:

BEGIN{
INPUT bytes: 1B 5B 32 35 48 0D 0D 0A 41 42 43 44 45 08 08 08 1B 5B 32 50
MOVERECT 1..25,0..80 -> 0..24,0..80
  blit dirty
DAMAGE 24..25,0..80
  putglyph E
MOVERECT 24..25,4..80 -> 24..25,2..78
  blit dirty
DAMAGE 24..25,78..80
  putglyph E
  blit dirty
}END
BEGIN{
  putglyph E
INPUT bytes: 0D 0A
MOVERECT 1..25,0..80 -> 0..24,0..80
  blit dirty
DAMAGE 24..25,0..80
  blit dirty
}END

FAIL case:

BEGIN{
INPUT bytes: 1B 5B 32 35 48 0D 0D 0A 41 42 43 44 45 08 08 08 1B 5B 32 50 0D 0D 0A
MOVERECT 1..25,0..80 -> 0..24,0..80
  blit dirty
DAMAGE 24..25,0..80
  putglyph E
MOVERECT 24..25,4..80 -> 24..25,2..78
  blit dirty
DAMAGE 24..25,78..80
MOVERECT 1..25,0..80 -> 0..24,0..80
  blit dirty
DAMAGE 24..25,0..80
  blit dirty
}END

What's notable is the number of times a "putglyph E" appears.

[Paul "LeoNerd" Evans (leonerd)]

UGH. Total red-herring.

Turns out the "OK" case is all because the cursor ends up over that E character while sleeping; the putglyph that forms part of cursor rendering is the one that makes it appear. By fixing the sequence to

  echo -en "\e[25H\r\nABCDE\b\b\b\e[2P\e[C"

It renders just as broken with or without the sleep. So it doesn't seem to be timing related after all.

[Paul "LeoNerd" Evans (leonerd)]

Progress! By adding even more verbose debugging, I have got it down to the following:

BEGIN{
INPUT bytes: 1B 5B 32 35 48 0D 0D 0A 41 42 43 44 45 08 08 08 1B 5B 32 50 1B 5B 43
MOVERECT 1..25,0..80 -> 0..24,0..80
DAMAGE 24..25,0..80
  putglyph E at (2,24)
MOVERECT 24..25,4..80 -> 24..25,2..78
DAMAGE 24..25,78..80
}END

The 'E' glyph is correctly printed at its final destination position of (2,24), but that happens before the MOVERECT is flushed into it, moving the (currently-blank) (4,24) cell over the top of it, thus making it appear blank.

[Paul "LeoNerd" Evans (leonerd)]

Fixed by -r654