use macro SkFixedRoundToInt where necessary

use macro SkFixedRoundToInt where necessary

This makes developers easy to grasp what's going on.

BUG=
TEST=

[guanqun, 2012-02-02 01:01:04]

Please review.
Thanks!

[reed1, 2012-02-02 13:17:54]

looks good. Have you run the gm tool against known images, to confirm that there
was no change due to this CL?

out/Release/gm -r gm/base-<your platform>

Inspecting the code looks fine, but its always good to run the tool as well.
Good practice for larger changes in the future.

[guanqun, 2012-02-02 15:13:03]

On 2012/02/02 13:17:54, reed1 wrote:
> looks good. Have you run the gm tool against known images, to confirm that
there
> was no change due to this CL?
> 
> out/Release/gm -r gm/base-<your platform>
> 
> Inspecting the code looks fine, but its always good to run the tool as well.
> Good practice for larger changes in the future.

Thanks for your advice, I've tried gm on macpro on top of revision 3125. Without
this CL,
the results are 80 passed, 90 failed, 93 missing reference images.
and with this CL, the result is the same.
So I assume this CL is safe.

BTW: why there are so many fails? Is it worthwhile for me to take a look? Or are
these fails a known issue?

Thanks!

[reed1, 2012-02-02 15:44:18]

Hmmm, lots of failures (w/o) this patch sounds bad. I also run on a macpro
(using 'gm -r gm/base-macpro') and I see no failures. Could be differences in
font rendering.

Can you send me one image output from a failing test? You can generate images by
just passing -w <dir> to the gm tool. You can also (when you want to limit to a
single test) pass --match substring_of_test_name to gm.

[guanqun.lu_gmail.com, 2012-02-02 16:24:37]

Hi,

Here are some output with my patch on top of revision 3125:
...
drawing... bitmapcopy [540 330]
-----  pixel mismatch for bitmapcopy_8888 at [39 17] base 0xFFD2D2D2
current 0xFFDADADA
-----  pixel mismatch for bitmapcopy_4444 at [38 18] base 0xFF444444
current 0xFF555555
-----  pixel mismatch for bitmapcopy_565 at [39 17] base 0xFFCECFCE
current 0xFFD6D7D6
-----  pixel mismatch for bitmapcopy_gpu at [39 17] base 0xFFD3D3D3
current 0xFFDADADA
FAILED to read gm/base-macpro/bitmapcopy_pdf.png
drawing... arithmode [640 480]
-----  pixel mismatch for arithmode_8888 at [365 6] base 0xFFC2C2C2
current 0xFFB9B9B9
-----  pixel mismatch for arithmode_4444 at [365 6] base 0xFFDDDDDD
current 0xFFCCCCCC
-----  pixel mismatch for arithmode_565 at [365 6] base 0xFFBDBEBD
current 0xFFB5B6B5
-----  pixel mismatch for arithmode_gpu at [365 6] base 0xFFC2C2C2
current 0xFFB9B9B9
-----  pixel mismatch for arithmode_pdf at [365 6] base 0xFFC2C2C2
current 0xFFB9B9B9
drawing... aarectmodes [640 480]

And I attach bitmapcopy_8888.png and arithmode_888.png . Looking at
the diff of arithmode_888.png, it seems these number fonts differs..

It's late for my local time. I'll check mails tomorrow. If there's
anything I can do, I'll be glad to.

Thanks!

On Thu, Feb 2, 2012 at 11:44 PM,  <reed@google.com> wrote:
> Hmmm, lots of failures (w/o) this patch sounds bad. I also run on a
> macpro (using 'gm -r gm/base-macpro') and I see no failures. Could be
> differences in font rendering.
>
> Can you send me one image output from a failing test? You can generate
> images by just passing -w <dir> to the gm tool. You can also (when you
> want to limit to a single test) pass --match substring_of_test_name to
> gm.
>
> http://codereview.appspot.com/5617047/



-- 
Guanqun

[guanqun, 2012-02-06 02:20:03]

Hi Mike,

Sorry to bother you but have you got my pictures? Any comments on the errors?

Thanks!

[reed1, 2012-02-06 19:06:30]

The differences are not even LCD specific, so I can only guess 10.6 -vs- 10.7
font rendering differences in CG? Who knows.

landed in rev. 3142.

[guanqun, 2012-02-07 00:33:51]

On 2012/02/06 19:06:30, reed1 wrote:
> The differences are not even LCD specific, so I can only guess 10.6 -vs- 10.7
> font rendering differences in CG? Who knows.

Thanks for the info, this is a strong reason I should upgrade to 10.7 for my
MacBookPro. :)

[guanqun, 2012-02-08 02:48:37]

On 2012/02/07 00:33:51, guanqun wrote:
> On 2012/02/06 19:06:30, reed1 wrote:
> > The differences are not even LCD specific, so I can only guess 10.6 -vs-
10.7
> > font rendering differences in CG? Who knows.
> 
> Thanks for the info, this is a strong reason I should upgrade to 10.7 for my
> MacBookPro. :)

FYI. I've upgraded to Lion, but still, on rev 3125, I'm still having mismatched
pixels. Don't know how to fix it...