Re: Bug/Analysis: Incorrect Input Rendering after Undo_Justification

[Thomas Dickey]

On Thu, Aug 31, 2017 at 03:07:37PM +0200, Leon Winter wrote:
> Hi,
> 
> attached test case shows the bug. Instead of correctly overriding the first
> character in the field, the character is appended at the end of the text.
> 
> The cause is the incorrect state left by:
>  - set_field_back
>    - _nc_Synchronize_Attributes
>      - Undo_Justification
> When changing a field attribute the field needs to be redrawn. If the field
> would normally be justified but no is the current field, justification will be
> dropped and the function "Undo_Justification" will be called.
> Otherwise "Buffer_To_Window" is called. The function "Buffer_To_Window" resets
> the coordinates of the window cursor to that state before it was called. The
> function "Undo_Justification" does not do that.
> Therefore after the call of "set_field_back" the field is in a state of a
> correct curcol (=0) but an incorrect window cursor (=end of text).
> 
> Maybe "Undo_Justification" should also properly reset the window cursor of its
> drawing operation.

hmm - my first inclination would be to compare with Solaris.  I see two
problems (perhaps the latter is what you're talking about :-) :

        + win_form is never initialized, so Solaris dumped core.

        + changing the program to initialize win_form to a new window
          not related to win_body got Solaris to display a menu in
          win_form with "A" at the beginning of the second field,
          while ncurses 5.9 (on my Debian/7) shows the "A"
          at the end of the field.

However, running the test with ncurses6 (current), I get the same picture
as Solaris.

so... it seems that this is a bug that may have been fixed sometime,
and I might have to bisect to find when (I don't recall).

-- 
Thomas E. Dickey <address@hidden>
http://invisible-island.net
ftp://ftp.invisible-island.net

signature.asc
Description: Digital signature