Re: lib/58282: Sysinst terminal output size increased because curses

[Valery Ushakov <uwe%stderr.spb.ru@localhost>]

The following reply was made to PR lib/58282; it has been noted by GNATS.

From: Valery Ushakov <uwe%stderr.spb.ru@localhost>
To: gnats-bugs%netbsd.org@localhost
Cc: Brett Lymn <blymn%internode.on.net@localhost>
Subject: Re: lib/58282: Sysinst terminal output size increased because curses
Date: Tue, 28 May 2024 14:42:06 +0300

 On Tue, May 28, 2024 at 16:48:48 +0930, Brett Lymn wrote:
 
 > On Sun, May 26, 2024 at 02:20:02AM +0000, Valery Ushakov wrote:
 >
 > >  makech() in refresh.c has a block of code with the comment:
 > >
 > >    Work out if we can use a clear to end of line.  If we are using
 > >    color then we can only erase the line if the terminal can erase to
 > >    the background color.
 > >
 > >  it uses (for the HAVE_WIDE case, which is the defautl) the helper
 > >  function _cursesi_celleq() that also checks if "attr" is the same, but
 > >  the next test in the condition narrows the attr comparison to
 > >  attr_mask.  That looks suspicious.
 > >
 >
 > Having a look at the output all the spam comes from the foreground and
 > background colour being set for each cell.  Running from memory the
 > narrowing to attr in that section of code is deliberate to force the
 > update of the background colour because the actual screen will not have
 > changed but the virtual screen image will match the update screen image
 > so the background will not be set.
 
 I don't understand offhand what that code tries to do, but here are
 two observations:
 
 1.  the wide case and non-wide case have different semantic (modulo
     wideness).
 
 2.  the narrowing is a no-op in the wide case b/c celleq already tests
     if cp->attr == win->battr, and then the narrowed tests redundantly
     checks if
 
       (cp->attr & attr_mask) == (win->battr & attr_mask)
 
     if the intention was (as non-wide case seems to suggest) to test
     the cell's text and only parts of "attr", then using
     _cursesi_celleq is wrong.
 
 -uwe