In article <yk1c+omsyd3qm...@internode.on.net>, Brett Lymn <bl...@internode.on.net> wrote: > >Folks, > >This started out innocuously enough, I picked up a PR about fileobj from >pkgsrc not working properly. This was partially an upstream problem but >also an issue with our curses chgat() function. After a bit of digging >I found two things. Firstly, every other colour capable implementation >of curses seems to use colour pair 0 as the default colour - our >implementation >set the default pair to the last pair. Secondly, without looking at our >libcurses code and exploiting access to a static variable there is >simply no way of actually using the default colour pair in our libcurses. > >Given these findings I embarked on changing the default colour pair >number to be 0. This actually turned out to be a lot harder than it >first seems because the refresh code used a colour pair of 0 to mean >"don't apply colour at all" along with a bunch of initialisation issues >where the initial colour set wouldn't happen. After a bit of rework I >have all the curses tests passing (some output is different now of >course) and using the updated code seems to work fine. Certainly, the >original issue with fileobj is gone. > >The reason I am putting this out here is that there are now some visual >differences, mainly with the colour of the background. In some curses >applications the assume_default_colors extension is used and the colours >set to -1 (another curses extension that means "no colour"). Previously >in NetBSD curses this had no effect, the terminal remained in white >foreground and black background but now this request is honoured the >terminal default colour scheme is used. > >So, the point of all this blather is even though there have been no >interface changes as a result of my work there is the potential for a >significant visual change. Should I mark this with a libcurses major >bump? I don't think it necessary but I am putting it out there for >discussion...
1. Is there a way to have a "compatibility mode"? Or is that too complicated or unwanted? 2. Bumping the library version is probably not necessary because there is no ABI change, but I can see that perhaps changing how the programs look when the library is updated is unexpected and perhaps there is a benefit doing so. 3. I noticed that there is no version number in <curses.h>. Perhaps programs could use that to check the version and alter their behavior with respect to colors if it existed? Best, christos
