Re-enable imageresizetiled, but skip it in tilegrid and rtree.

Re-enable imageresizetiled, but skip it in tilegrid and rtree.

This change makes kSkipTiled_Flag also skip --tileGrid and --rtree.
All GMs which were passing kSkipTiled_Flag before were also passing
kSkipPicture_Flag, which also skips tilegrid and rtree, so this should
have no effect on them, but provides a smaller hammer for GMs which
still want to test picture playback, but not tiling.

The exception is magnifier, which was passing only kSkipTiled_Flag,
but magnifier is an odd beast and not web-exposed, so I'm not worried
about reducing its coverage slightly.

R=scroggo@google.com
BUG=skia:

Committed: https://code.google.com/p/skia/source/detail?r=13514

[Stephen White, 2014-02-20 16:48:52]

scroggo: PTAL. Thanks!

[scroggo, 2014-02-20 16:53:00]

lgtm

cc'ing Mike Klein, who owns DM (current/eventual GM replacement)

[mtklein, 2014-02-20 17:05:20]

On 2014/02/20 16:53:00, scroggo wrote:
> lgtm
> 
> cc'ing Mike Klein, who owns DM (current/eventual GM replacement)

Looks to me we'd want to update dm/DMReplayTask.cpp to replace
    if (FLAGS_rtree && fUseRTree) {
        return false;
    }
with
    if (FLAGS_rtree && fUseRTree) {
        return fGM->getFlags() & skiagm::GM::kSkipTiled_Flag;
    }

I'd have thought we'd need to update dm/DMTileGridTask.cpp too, but it appears
to already have the right logic.

Should I be defaulting DM to --tileGrid true?  Last I checked tileGrid had tons
of failures.

[Stephen White, 2014-02-20 17:11:52]

On 2014/02/20 17:05:20, mtklein wrote:
> On 2014/02/20 16:53:00, scroggo wrote:
> > lgtm
> > 
> > cc'ing Mike Klein, who owns DM (current/eventual GM replacement)
> 
> Looks to me we'd want to update dm/DMReplayTask.cpp to replace
>     if (FLAGS_rtree && fUseRTree) {
>         return false;
>     }
> with
>     if (FLAGS_rtree && fUseRTree) {
>         return fGM->getFlags() & skiagm::GM::kSkipTiled_Flag;
>     }
> 
> I'd have thought we'd need to update dm/DMTileGridTask.cpp too, but it appears
> to already have the right logic.
> 
> Should I be defaulting DM to --tileGrid true?  Last I checked tileGrid had
tons
> of failures.

The bots seem to be running it in GM (hence the failure that led to this CL).
Tilegrid is also used in Chrome, and has been for a few mstones now.

[Stephen White, 2014-02-20 17:18:02]

Committed patchset #1 manually as r13514 (presubmit successful).

[mtklein, 2014-02-21 20:14:05]

On 2014/02/20 17:11:52, Stephen White wrote:
> On 2014/02/20 17:05:20, mtklein wrote:
> > On 2014/02/20 16:53:00, scroggo wrote:
> > > lgtm
> > > 
> > > cc'ing Mike Klein, who owns DM (current/eventual GM replacement)
> > 
> > Looks to me we'd want to update dm/DMReplayTask.cpp to replace
> >     if (FLAGS_rtree && fUseRTree) {
> >         return false;
> >     }
> > with
> >     if (FLAGS_rtree && fUseRTree) {
> >         return fGM->getFlags() & skiagm::GM::kSkipTiled_Flag;
> >     }
> > 
> > I'd have thought we'd need to update dm/DMTileGridTask.cpp too, but it
appears
> > to already have the right logic.
> > 
> > Should I be defaulting DM to --tileGrid true?  Last I checked tileGrid had
> tons
> > of failures.
> 
> The bots seem to be running it in GM (hence the failure that led to this CL).
> Tilegrid is also used in Chrome, and has been for a few mstones now.

Ah, for the record, GM is recording using a tilegrid bounding box, but not
actually rendering the tiles.
This is line ~1670:
                // We cannot yet pass 'true' to generate_image_from_picture to
                // perform actual tiled rendering (see Issue 1198 -
                // https://code.google.com/p/skia/issues/detail?id=1198)
So, GM is effectively testing that SkTileGridPicture works the same as
SkPicture, but not that drawing tiles work.  Indeed, if you swap in an SkPicture
for an SkTileGridPicture, you get the very same failures (about half of GMs have
at least one pixel wrong) when you try to draw and restitch the grid of 16x16
tiles.