Issue #22 resolved

It makes no sense to supply images at various resolutions that are powers of 2 and then again with scale 2

Scott Palmer
created an issue

I got warnings like this:

.... No image found for point size: 16 at scale 2. .... No image found for point size: 32 at scale 2. ... etc.

Well guess what? That image you found at point size 32 scale 1 IS the image for point size 16 at scale 2!

Basically all of the "scale 2" stuff should go away (and be automated by the plugin by choosing the next point size up if it is available) and the top point size should go one level further.

Comments (5)

  1. shemnon repo owner

    These are errors coming out of the mac tool, not ones I generate. This is related to retina display support, and very mac specific.

  2. Scott Palmer reporter

    Sure, but why make the cross platform java dev do the platform specific bits. If you have the images at the correct resolution, can't the plugin manage things so the mac tool sees what it needs?

  3. shemnon repo owner

    The iconutil program on the mac is restrictive on what it accepts, the plugin already filters out the icons provided to the tool to the subset of the ones it accepts, specifically it wants 2^nx2^n with the exception of 64x64 (I don't know why).

    The tool is also obnoxious in that if it does not get every icon it can use it complains about the icons it is missing. The down side is that these notifications are shared in a fashion that is indistinguishable from other, more serious errors the tool may present. So I provide the output of the tool, unfiltered.

    I don't have the code with me at the moment, but does this text show up with the --info option passed into gradle or not? I could stash the output and print it out at the --info level if I don't do it curently, and only give the fill dump when a non-zero return is given from the iconutil program (assuming it gives error code output).

  4. shemnon repo owner

    Also, why not automatically scale it before feeding it to iconutil? Because OSX does a better job of interpolating based on various screen geometries if we provide the pixel-to-pixel varients only. It gets more convoluted when retina gets involved.

  5. Scott Palmer reporter

    I wasn't using --info.

    I'm not suggesting to scale anything. I'm only suggesting that the retina variants aren't special as far as the source images go and the 2x versions can be copied from the 2^(n+1) version for cases when they are not provided explicitly.

  6. Log in to comment