Add VALIDATE_FAILURE_IS_A_TOOL_FAILURE to specify whether a failure in validation

Add VALIDATE_FAILURE_IS_A_TOOL_FAILURE to specify whether a failure in validation
means the tool should return failure or not. For now it is not defined, which
means any failed pixels are still reported to stdout, but the tool does not
return an error, allowing the bots to go green (until we can fix these failures).
Committed: https://code.google.com/p/skia/source/detail?r=7211

[EricB, 2013-01-15 21:42:09]

https://codereview.appspot.com/7105056/diff/1/trunk/tools/render_pictures_mai...
File trunk/tools/render_pictures_main.cpp (right):

https://codereview.appspot.com/7105056/diff/1/trunk/tools/render_pictures_mai...
trunk/tools/render_pictures_main.cpp:26: //#define
VALIDATE_FAILURE_IS_A_TOOL_FAILURE
I know this is a quick fix, but maybe this could be useful as a command-line
flag?

https://codereview.appspot.com/7105056/diff/1/trunk/tools/render_pictures_mai...
trunk/tools/render_pictures_main.cpp:216: return false;
Should we also switch on VALIDATE_FAILURE_IS_A_TOOL_FAILURE here?

https://codereview.appspot.com/7105056/diff/1/trunk/tools/render_pictures_mai...
trunk/tools/render_pictures_main.cpp:223: return false;
Or here?

https://codereview.appspot.com/7105056/diff/1/trunk/tools/render_pictures_mai...
trunk/tools/render_pictures_main.cpp:238: goto DONE;
Is there any benefit in leaving the #else block out and therefore printing out
all of the pixel mismatches?

[reed1, 2013-01-15 21:49:46]

https://codereview.appspot.com/7105056/diff/1/trunk/tools/render_pictures_mai...
File trunk/tools/render_pictures_main.cpp (right):

https://codereview.appspot.com/7105056/diff/1/trunk/tools/render_pictures_mai...
trunk/tools/render_pictures_main.cpp:26: //#define
VALIDATE_FAILURE_IS_A_TOOL_FAILURE
On 2013/01/15 21:42:09, EricB wrote:
> I know this is a quick fix, but maybe this could be useful as a command-line
> flag?

Hmmm, do we think it will live beyond tomorrow, when Edison can do a cleaner
fix?

https://codereview.appspot.com/7105056/diff/1/trunk/tools/render_pictures_mai...
trunk/tools/render_pictures_main.cpp:216: return false;
On 2013/01/15 21:42:09, EricB wrote:
> Should we also switch on VALIDATE_FAILURE_IS_A_TOOL_FAILURE here?

Possibly, though I really only think of this as a stop-gap, and not a long-lived
feature. Also, these are not currently failing (afaik), and perhaps they are
significant enough that we *should* return failure.

https://codereview.appspot.com/7105056/diff/1/trunk/tools/render_pictures_mai...
trunk/tools/render_pictures_main.cpp:238: goto DONE;
On 2013/01/15 21:42:09, EricB wrote:
> Is there any benefit in leaving the #else block out and therefore printing out
> all of the pixel mismatches?

That would be (possibly very) different behavior, in that we might print out
1000s of lines per tile (if they were different). I was trying for a surgical
strike, minimizing changes.

[EricB, 2013-01-15 21:52:20]

LGTM

https://codereview.appspot.com/7105056/diff/1/trunk/tools/render_pictures_mai...
File trunk/tools/render_pictures_main.cpp (right):

https://codereview.appspot.com/7105056/diff/1/trunk/tools/render_pictures_mai...
trunk/tools/render_pictures_main.cpp:26: //#define
VALIDATE_FAILURE_IS_A_TOOL_FAILURE
On 2013/01/15 21:49:47, reed1 wrote:
> On 2013/01/15 21:42:09, EricB wrote:
> > I know this is a quick fix, but maybe this could be useful as a command-line
> > flag?
> 
> Hmmm, do we think it will live beyond tomorrow, when Edison can do a cleaner
> fix?

Hopefully not...

https://codereview.appspot.com/7105056/diff/1/trunk/tools/render_pictures_mai...
trunk/tools/render_pictures_main.cpp:216: return false;
On 2013/01/15 21:49:47, reed1 wrote:
> On 2013/01/15 21:42:09, EricB wrote:
> > Should we also switch on VALIDATE_FAILURE_IS_A_TOOL_FAILURE here?
> 
> Possibly, though I really only think of this as a stop-gap, and not a
long-lived
> feature. Also, these are not currently failing (afaik), and perhaps they are
> significant enough that we *should* return failure.

Fair enough.

https://codereview.appspot.com/7105056/diff/1/trunk/tools/render_pictures_mai...
trunk/tools/render_pictures_main.cpp:238: goto DONE;
On 2013/01/15 21:49:47, reed1 wrote:
> On 2013/01/15 21:42:09, EricB wrote:
> > Is there any benefit in leaving the #else block out and therefore printing
out
> > all of the pixel mismatches?
> 
> That would be (possibly very) different behavior, in that we might print out
> 1000s of lines per tile (if they were different). I was trying for a surgical
> strike, minimizing changes.

Makes sense.