The reason I didn't do things this way in my patch is that it doesn't give you any reasonable guarantee that things will actually work. The number of colors required is not known by this layer of the program (or by any individual layer), so you always run the risk of exceeding the color-pair limit if you allow more than sqrt(COLOR_PAIRS) colors.
On the other hand, there are only 8 (well, 9) named colors, there's no way for users to even access the higher colors, and no-one has complained about it yet; so I will probably just stick with my minimal correct solution to this bug until I get requests to access arbitrary color values. If there is a desire for that, I can implement some sort of dynamic allocation that just gives you white-on-black if you use too many colors (and a guarantee that using 8 colors is safe). Daniel
signature.asc
Description: Digital signature
