On Sat, 26 Feb 2011 17:38:43 +0000 (GMT), Michael Drake wrote: > In article > <[email protected]>, > Chris Young <[email protected]> wrote: > > > http://homepage.ntlworld.com/cdyoung/tmp/ns-screenshots/ > > Thanks! Actually PNG (or other lossless format I can decode) would be > better, if possible! > > > Erm, my Wikipedia "tabs" are bright blue??! Is this a colour > > endianness problem in libcss? > > That's odd. Looks like cyan, and it should be white. So red component > is wrongly zero. The only thing to change recently is libcss added > opacity support to colours. Maybe something is not right on big endian > systems, but if that was the problem I'd expect the colours to be wrong > all over the place.
I suspect that the alpha component is overwriting R instead of A (NetSurf colours are ABGR so if the endianness is being ignored or assumed, this is exactly what will happen). As you say, I'd expect it to be wrong everywhere. The fact it isn't suggests there is something particular to that colour which may be calling some different code. I looked at the libcss opacity commit but I can't see where the opacity is being converted into the alpha byte of a css_color. The alpha channel doesn't appear to get as far as NetSurf's plotters - I'm only getting three bytes at that point. Is this correct? Chris
