Make gm's json results include checksums even if run without --readPath

Make gm's json results include checksums even if run without --readPath
Committed: https://code.google.com/p/skia/source/detail?r=6917

[epoger, 2012-12-19 19:48:31]

https://codereview.appspot.com/6979043/diff/1/gm/tests/outputs/no-readpath/ou...
File gm/tests/outputs/no-readpath/output-expected/json-summary.txt (right):

https://codereview.appspot.com/6979043/diff/1/gm/tests/outputs/no-readpath/ou...
gm/tests/outputs/no-readpath/output-expected/json-summary.txt:11:
"expected-results" : null
Similar to
https://code.google.com/p/skia/source/browse/trunk/gm/tests/outputs/compared-...
, but expected-results is EMPTY rather than showing "ignore-failure" for each
test.

https://codereview.appspot.com/6979043/diff/1/gm/tests/rebaseline.sh
File gm/tests/rebaseline.sh (right):

https://codereview.appspot.com/6979043/diff/1/gm/tests/rebaseline.sh#newcode30
gm/tests/rebaseline.sh:30: # Delete all the "actual" directories, so we can
svn-add any new "expected"
The change in this file allowed me to rebaseline the rest of the CL
automatically.

[rmistry, 2012-12-19 20:01:00]

LGTM

https://codereview.appspot.com/6979043/diff/1/gm/gmmain.cpp
File gm/gmmain.cpp (right):

https://codereview.appspot.com/6979043/diff/1/gm/gmmain.cpp#newcode662
gm/gmmain.cpp:662: } else {
make this an 'else if' ?

[epoger, 2012-12-20 17:28:01]

Per discussion with Eric yesterday, I added a new "no-comparison" status. 
Details below...

I'd like another LG at patchset 2... first LG wins.

https://codereview.appspot.com/6979043/diff/1/gm/gmmain.cpp
File gm/gmmain.cpp (right):

https://codereview.appspot.com/6979043/diff/1/gm/gmmain.cpp#newcode662
gm/gmmain.cpp:662: } else {
On 2012/12/19 20:01:00, rmistry wrote:
> make this an 'else if' ?

Done, and augmented the comment to more explicitly explain the referenceBitmap
check.

https://codereview.appspot.com/6979043/diff/6001/gm/gmmain.cpp
File gm/gmmain.cpp (right):

https://codereview.appspot.com/6979043/diff/6001/gm/gmmain.cpp#newcode627
gm/gmmain.cpp:627: fJsonActualResults_NoComparison[name.c_str()] =
actualResults;
Added new "no-comparison" status for actual results, which we now use in all
cases when we used to use "failure-ignored".

Once we actually support ignored failures (i.e. here's the expected result, but
it's ok if you don't match it) I'll add a test of that status.

https://codereview.appspot.com/6979043/diff/6001/gm/tests/rebaseline.sh
File gm/tests/rebaseline.sh (right):

https://codereview.appspot.com/6979043/diff/6001/gm/tests/rebaseline.sh#newco...
gm/tests/rebaseline.sh:10: function replace_expected_with_actual {
The change in this file at patchset 2 LOOKS big, but I mainly pulled some code
into function blocks.

https://codereview.appspot.com/6979043/diff/6001/gm/tests/rebaseline.sh#newco...
gm/tests/rebaseline.sh:63: svn_add_new_files
Here's the main modification to rebaseline.sh: only run svn commands once all
the self-tests succeed.

[EricB, 2012-12-20 17:56:40]

LGTM.  If the rebaselining script gets much bigger, I'd suggest moving to python
with a thin wrapper.

[epoger, 2012-12-20 18:33:34]

On 2012/12/20 17:56:40, EricB wrote:
> LGTM.  If the rebaselining script gets much bigger, I'd suggest moving to
python
> with a thin wrapper.

Agreed, thanks.