Merge lp://qastaging/~kdub/mir/fix-1373698 into lp://qastaging/mir

FAILED: Continuous integration, rev:1944
http://jenkins.qa.ubuntu.com/job/mir-0.8-ci/10/
Executed test runs:
    FAILURE: http://jenkins.qa.ubuntu.com/job/mir-0.8-utopic-amd64-ci/10/console
    SUCCESS: http://jenkins.qa.ubuntu.com/job/mir-android-utopic-i386-build/1914
    FAILURE: http://jenkins.qa.ubuntu.com/job/mir-clang-utopic-amd64-build/1921/console
    FAILURE: http://jenkins.qa.ubuntu.com/job/mir-mediumtests-utopic-touch/1876/console
    FAILURE: http://jenkins.qa.ubuntu.com/job/mir-mediumtests-builder-utopic-armhf/813/console

Click here to trigger a rebuild:
http://s-jenkins.ubuntu-ci:8080/job/mir-0.8-ci/10/rebuild

FAILED: Continuous integration, rev:1944
http://jenkins.qa.ubuntu.com/job/mir-team-mir-development-branch-ci/2812/
Executed test runs:
    SUCCESS: http://jenkins.qa.ubuntu.com/job/mir-android-utopic-i386-build/1915
    FAILURE: http://jenkins.qa.ubuntu.com/job/mir-clang-utopic-amd64-build/1922/console
    FAILURE: http://jenkins.qa.ubuntu.com/job/mir-mediumtests-utopic-touch/1877/console
    FAILURE: http://jenkins.qa.ubuntu.com/job/mir-team-mir-development-branch-utopic-amd64-ci/1334/console
    FAILURE: http://jenkins.qa.ubuntu.com/job/mir-mediumtests-builder-utopic-armhf/814/console

Click here to trigger a rebuild:
http://s-jenkins.ubuntu-ci:8080/job/mir-team-mir-development-branch-ci/2812/rebuild

FAILED: Continuous integration, rev:1944
http://jenkins.qa.ubuntu.com/job/mir-ci/1706/
Executed test runs:
    SUCCESS: http://jenkins.qa.ubuntu.com/job/mir-android-utopic-i386-build/1913
    FAILURE: http://jenkins.qa.ubuntu.com/job/mir-clang-utopic-amd64-build/1920/console
    FAILURE: http://jenkins.qa.ubuntu.com/job/mir-utopic-amd64-ci/44/console
    FAILURE: http://jenkins.qa.ubuntu.com/job/mir-utopic-armhf-ci/44/console

Click here to trigger a rebuild:
http://s-jenkins.ubuntu-ci:8080/job/mir-ci/1706/rebuild

Sounds like "alpha_enabled" is a redundant form of "shaped" (shaped means to use the renderable's own alpha channel). We should get back to two alpha-related attributes instead of having three. Or is that too big a job for this fix?

See also bug 1236224 :)

OK I have a nicer solution rather than building on the "alpha_enabled" attribute which should not exist...

https://code.launchpad.net/~vanvugt/mir/simplify-alpha/+merge/236075

I think that the alpha settings on Renderable are a bit confusing too. They got this way because we melded the 'original' Renderable class to the BasicSurface class before separating them out again, and we got this jumble of alpha functions.

But now we have 2 competing MP's to address the bug/ambiguity. I was trying to avoid changing the Renderable interface because I knew it would be a more protracted discussion :) But I guess if we're going to break ABI, the important thing is to fix the ambiguity in Renderable.

In trunk now we have:
1) bool Renderable::alpha_enabled()
alpha_enabled() means that blending is enabled for the renderable
2) bool Renderable::shaped()
shaped() means "the pixel format has alpha"
3) float Renderable::alpha()
This is "plane alpha" or "global alpha", that is, it is an additional alpha value that is blended.

So in trunk now, Renderable::alpha_enabled does have some sort of place. It is useful for the renderable to explicitly designate whether it wants itself blended.

Renderable::alpha() doesn't really indicate what's going on here at first glance, I'd suggest Renderable::plane_alpha()

Renderable::shaped() seems to be a mash-up of information already available via the other functions. If the alpha is to be explicitly disabled, mg::Renderable::alpha_enabled() can be checked. If the pixel format doesn't have an alpha channel, mg::Buffer::pixel_format() can be checked. If the plane alpha is less than 1.0f, then mg::Renderable::alpha() can be checked.

So, how about:
mg::Renderable::blend_required();
This is false if blending is 1) explicitly disabled, or 2) no alpha channel in the pixel format. In the great-and-glorious-future, I'd like to see this function just represent 1), and 2) can be encoded into the pixel format details (lp: #1236254)
mg::Renderable::plane_alpha();
This will require an alpha blend if less than 1.0.

I agree with "blend_required()" and was thinking the same thing, although don't like that particular name. Basically it exists already in "shaped()". We just need to rename "shaped" and document the fact that implementers of the interface do have the option of returning false even if the pixel format is RGBA.

The alternate fix has landed. So rejecting.