Hi everyone,

apologies for the confusions. I was doing this in between other things which obviously did not work out too well :)

I will try this again in a few days but with some more structure. My idea is to first get a better idea of the scope of this change by finding all tests relying on the old behavior and writing some new tests capturing the use cases which should be possible with the new behavior. After that I will remove that line again and we (hopefully) have some more insight into the extent of the effects of that change.

Bests 
Patrick

>
>
>On Tue, Oct 16, 2018 at 3:17 PM Bert Freudenberg <bert at freudenbergs.de> wrote:
>
>On Tue, Oct 16, 2018 at 10:29 AM <patrick.rein at hpi.uni-potsdam.de> wrote:
>
>Hi all,
>
>the change is reverted now but I would like to document our findings and our insights for future changes.
>
>From Berts mail I gathered that the 255-0-0-0 to be transparent is required for 16bit -> 32bit conversion to work. As this has been around for quite some time some code relies on it. At the same time this regularly leads to problems when working with images loaded from external sources. For example, loading a PNG and using collectColor: to recolor it does not work if the PNG includes pure rgb black pixels. For this to work we would have to map pure black to Squeak Form black on loading and reconvert it on storing. This would however destroy information as we can not distinguish between 0-0-1 and 0-0-0 anymore. The other way to go would be to change the 16bit -> 32bit conversion to set alpha explicitly, but I do not know whether that is actually possible efficiently.
>
>Are there other aspects which I am missing? Is it likely for external packages to heavily rely on this functionality?
>
>
>I didn't say you need to revert this change ;) 
>I was only explaining where the odd convention comes from.
>Now, early in the dev cycle is a good time to try this. There are certainly other places that need clean up (e.g. isTransparent).
>If we can limit the odd color treatment to the 16 bpp case (and only do it when between 16 and 32) I'd be pretty happy.
>+1. Why don't we reinstitute the change in trunk and fix as required?
>
>- Bert -
>
>
>_,,,^..^,,,_
>best, Eliot --0000000000003e52660578629c63--