Do not call preDraw for NOOP.

Do not call preDraw for NOOP.

Fixes a bug where playing back SKPs generated by chrome using
--enable-deferred-image-decoding hit an assert in SkDebuggerGUI.
SkOffsetPicturePlayback adds to an array in preDraw, and expects
the length of that array to be the same as the list of commands
in its SkDebugCanvas. In the asserting case, preDraw was called
but then SkPicturePlayback skipped drawing to the SkDebugCanvas,
so the command was not propagated to the SkDebugCanvas, and the
counts did not match.

Closing in favor of https://codereview.chromium.org/15746003/

[scroggo, 2013-04-15 23:08:04]

Robert,

It seems odd to me that Chrome is generating a NOOP (only with
--enable-deferred-image-decoding; maybe that's intended? I'll investigate
tomorrow). That said, it seems logical to me that preDraw should not be called
in the case where there was a NOOP. Let me know if you think this is an
appropriate fix.

[scroggo, 2013-04-16 13:29:19]

On 2013/04/15 23:08:04, scroggo wrote:
> Robert,
> 
> It seems odd to me that Chrome is generating a NOOP (only with
> --enable-deferred-image-decoding; maybe that's intended? I'll investigate
> tomorrow). That said, it seems logical to me that preDraw should not be called
> in the case where there was a NOOP. Let me know if you think this is an
> appropriate fix.

Okay, it looks like the NOOPs are part of peephole optimizations, which of
course you knew.

[scroggo, 2013-05-21 18:18:53]

On 2013/04/16 13:29:19, scroggo wrote:
> On 2013/04/15 23:08:04, scroggo wrote:
> > Robert,
> > 
> > It seems odd to me that Chrome is generating a NOOP (only with
> > --enable-deferred-image-decoding; maybe that's intended? I'll investigate
> > tomorrow). That said, it seems logical to me that preDraw should not be
called
> > in the case where there was a NOOP. Let me know if you think this is an
> > appropriate fix.
> 
> Okay, it looks like the NOOPs are part of peephole optimizations, which of
> course you knew.

Robert, how would you feel if I changed the ASSERT in SkDebuggerGUI.cpp to a
printf (instead of this change)? I still hit this when I try to view skps.

[robertphillips, 2013-05-21 19:47:11]

I think this change could work with some tweaking.

https://codereview.chromium.org/14103024/diff/1/src/core/SkPicturePlayback.cpp
File src/core/SkPicturePlayback.cpp (right):

https://codereview.chromium.org/14103024/diff/1/src/core/SkPicturePlayback.cp...
src/core/SkPicturePlayback.cpp:718: size_t skipTo = 0;
remove "0 == skipTo" check.

https://codereview.chromium.org/14103024/diff/1/src/core/SkPicturePlayback.cp...
src/core/SkPicturePlayback.cpp:722: }
add "else { /* put the code you moved here */ }"