gm self-test: update expected dashing2.png images (in inputs/ dir) to reflect new behavior

gm self-test: update expected dashing2.png images (in inputs/ dir) to reflect new behavior

TBR=reed1
Committed: https://code.google.com/p/skia/source/detail?r=7116

[epoger, 2013-01-10 15:56:12]

Mike- thanks for rebaselining the gm self-tests in
https://code.google.com/p/skia/source/detail?r=6995 .

But there was a problem that rebaseline.sh didn't take care of for you...

Because your prior change (https://code.google.com/p/skia/source/detail?r=6984)
affected the expected dashing2.png output, the "golden master" images in these
directories needed to be updated:
- tests/inputs/identical-bytes: exact copy of the new dashing2.png
- tests/inputs/identical-pixels: a PNG file with different bytes but the same
pixels (which I currently create by hand using gimp)

This is a weakness in rebaseline.sh, which I had previously used to rebaseline
after changes to *gm*, not changes to *the rendering code*.  At some point we
should look into improving this part of the tool's behavior.

But for now, here's a manual fix.

[epoger, 2013-01-10 16:00:32]

Filed https://code.google.com/p/skia/issues/detail?id=1049 ('make
gm/tests/rebaseline.sh update input images when rendering code changes') to
track resolution

[reed1, 2013-01-10 16:19:17]

lgtm

do we expect to face this everytime we make a skia change that affects dashing2?

[epoger, 2013-01-10 16:23:28]

On 2013/01/10 16:19:17, reed1 wrote:
> lgtm
> 
> do we expect to face this everytime we make a skia change that affects
dashing2?

Unfortunately yes... to the extent that we want a continuously running gm
self-test, we will run into this in some form.  We *could* create a
pathologically simple gm test (just renders a rectangle?) whose output would
almost never change... do you think that would be a good idea?