bpo-37738: Fix curses addch(str, color_pair)#15071
Conversation
Fix the implementation of curses addch(str, color_pair): pass the color pair to setcchar(), instead of always passing 0 as the color pair.
vstinner
added
needs backport to 3.6
and removed
awaiting core review
needs backport to 3.6
labels
| if (type == 2) { | ||
| funcname = "add_wch"; | ||
| wstr[1] = L'\0'; | ||
| setcchar(&wcval, wstr, attr, 0, NULL); |
There was a problem hiding this comment.
I don't know if we are supposed to remove the color pair from attr passed to setcchar()?
There was a problem hiding this comment.
According to http://pubs.opengroup.org/onlinepubs/7908799/xcurses/setcchar.html, setcchar() takes an attr_t as the third argument.
According to http://pubs.opengroup.org/onlinepubs/7908799/xcurses/curses.h.html, attr_t is a bit set that can have these constants ORed:
WA_ALTCHARSET Alternate character set
WA_BLINK Blinking
WA_BOLD Extra bright or bold
WA_DIM Half bright
WA_HORIZONTAL Horizontal highlight
WA_INVIS Invisible
WA_LEFT Left highlight
WA_LOW Low highlight
WA_PROTECT Protected
WA_REVERSE Reverse video
WA_RIGHT Right highlight
WA_STANDOUT Best highlighting mode of the terminal
WA_TOP Top highlight
WA_UNDERLINE Underlining
WA_VERTICAL Vertical highlight
Meanwhile, waddch() takes a chtype, which contains a character, a color pair, and a bunch of attributes from this set:
A_ALTCHARSET Alternate character set
A_BLINK Blinking
A_BOLD Extra bright or bold
A_DIM Half bright
A_INVIS Invisible
A_PROTECT Protected
A_REVERSE Reverse video
A_STANDOUT Best highlighting mode of the terminal
A_UNDERLINE Underlining
There's no guarantee that, for example, A_BOLD == WA_BOLD. I'm afraid a more elaborate translation may be necessary here to be 100% correct.
There was a problem hiding this comment.
curses.h on my system has
/*
* X/Open attributes. In the ncurses implementation, they are identical to the
* A_ attributes.
*/
#define WA_ATTRIBUTES A_ATTRIBUTES
#define WA_NORMAL A_NORMAL
#define WA_STANDOUT A_STANDOUT
#define WA_UNDERLINE A_UNDERLINE
#define WA_REVERSE A_REVERSE
#define WA_BLINK A_BLINK
#define WA_DIM A_DIM
#define WA_BOLD A_BOLD
#define WA_ALTCHARSET A_ALTCHARSET
#define WA_INVIS A_INVIS
#define WA_PROTECT A_PROTECT
#define WA_HORIZONTAL A_HORIZONTAL
#define WA_LEFT A_LEFT
#define WA_LOW A_LOW
#define WA_RIGHT A_RIGHT
#define WA_TOP A_TOP
#define WA_VERTICAL A_VERTICAL
#if 1
#define WA_ITALIC A_ITALIC /* ncurses extension */
#endifso maybe we're fine? As long as we know we're using ncurses, and not some other curses library?
There was a problem hiding this comment.
I googled around to see whether there are curses implementations that use different bit values for these constants, and yes:
change ncurses-examples to use attr_t vs chtype to follow X/Open documentation more closely since Solaris xpg4-curses uses different values for WA_xxx vs A_xxx that rely on attr_t being an unsigned short. Tru64 aka OSF1, HPUX, AIX did as ncurses does, equating the two sets.
|
Python not really supports Solaris: last time I checked, the compilation of the _curses module failed with a compiler error anyway. |
|
Thanks @vstinner for the PR 🌮🎉.. I'm working now to backport this PR to: 3.7, 3.8. |
|
GH-15272 is a backport of this pull request to the 3.8 branch. |
|
Sorry, @vstinner, I could not cleanly backport this to |
Fix the implementation of curses addch(str, color_pair): pass the color pair to setcchar(), instead of always passing 0 as the color pair. (cherry picked from commit 077af8c) Co-authored-by: Victor Stinner <vstinner@redhat.com>
|
GH-15273 is a backport of this pull request to the 3.7 branch. |
Fix the implementation of curses addch(str, color_pair): pass the color pair to setcchar(), instead of always passing 0 as the color pair.
Fix the implementation of curses addch(str, color_pair): pass the color pair to setcchar(), instead of always passing 0 as the color pair.
Fix the implementation of curses addch(str, color_pair): pass the color pair to setcchar(), instead of always passing 0 as the color pair.
5 participants
Fix the implementation of curses addch(str, color_pair): pass the
color pair to setcchar(), instead of always passing 0 as the color
pair.
https://bugs.python.org/issue37738