Re: Graphics Rounding (was Re: Pixel-aligned autoresizing)

[Riccardo Mottola]

Hi,

> Note that with steps of 0.1 you will not see the round-to-nearest-even  
> behavior, as it will not trip on the exact halfway points due to  
> rounding errors. Increment by 0.25 to check whether  
> round-to-nearest-even is in action.

Perhaps indeed, however Eric already has proven rounding to be too system  
depended to be reliable for our use. While reading the documentation  I  
thought it to be more consistent.

I don't know how portable the function to set the rounding direction is.  
Perhaps it makes sense to round, but set the behaviour. I suppose both  
round up or down work, it is round-to-nearest-even which is undesirable  
for graphics.

Riccardo