Test msaa on bots using dm. 4 samples for Android, 16 elsewhere

Test msaa on bots using dm. 4 samples for Android, 16 elsewhere

Committed: https://skia.googlesource.com/skia/+/5abf584bcc21a488e04548e5a102265b17c11f83

[bsalomon, 2015-02-26 16:07:42]

bsalomon@google.com changed reviewers:
+ mtklein@google.com, senorblanco@chromium.org

[bsalomon, 2015-02-26 16:07:42]

May have to restrict this but casting a wide net initially with lots of try
bots.

[Stephen White, 2015-02-26 16:10:18]

On 2015/02/26 16:07:42, bsalomon wrote:
> May have to restrict this but casting a wide net initially with lots of try
> bots.

Out of curiosity, is this testing using backbuffer MSAA, or FBO MSAA?

[bsalomon, 2015-02-26 16:14:54]

On 2015/02/26 16:10:18, Stephen White wrote:
> On 2015/02/26 16:07:42, bsalomon wrote:
> > May have to restrict this but casting a wide net initially with lots of try
> > bots.
> 
> Out of curiosity, is this testing using backbuffer MSAA, or FBO MSAA?

FBO, dm/nanobench never use the default framebuffer. This is just dm. Once we
actually have the bots running dm with msaa I'll add nanobench.

[mtklein, 2015-02-26 16:15:13]

https://codereview.chromium.org/958033004/diff/20001/tools/dm_flags.json
File tools/dm_flags.json (left):

https://codereview.chromium.org/958033004/diff/20001/tools/dm_flags.json#oldc...
tools/dm_flags.json:7: "nvprmsaa4",
Huh, this looks unexpected.  I'd expect to see us adding "msaa4", not removing
nvprmsaa4.

re-run tools/dm_flags.py test?

[Stephen White, 2015-02-26 16:16:00]

On 2015/02/26 16:14:54, bsalomon wrote:
> On 2015/02/26 16:10:18, Stephen White wrote:
> > On 2015/02/26 16:07:42, bsalomon wrote:
> > > May have to restrict this but casting a wide net initially with lots of
try
> > > bots.
> > 
> > Out of curiosity, is this testing using backbuffer MSAA, or FBO MSAA?
> 
> FBO, dm/nanobench never use the default framebuffer. This is just dm. Once we
> actually have the bots running dm with msaa I'll add nanobench.

LGTM, but will leave for mtklein@

[bsalomon, 2015-02-26 16:16:44]

https://codereview.chromium.org/958033004/diff/20001/tools/dm_flags.json
File tools/dm_flags.json (left):

https://codereview.chromium.org/958033004/diff/20001/tools/dm_flags.json#oldc...
tools/dm_flags.json:7: "nvprmsaa4",
On 2015/02/26 16:15:12, mtklein wrote:
> Huh, this looks unexpected.  I'd expect to see us adding "msaa4", not removing
> nvprmsaa4.
> 
> re-run tools/dm_flags.py test?

I think I guessed wrong about what this file is for. I assumed it was overrides
for given bots, not test output.... Will run the test :)

[mtklein, 2015-02-26 16:21:22]

> I think I guessed wrong about what this file is for. I assumed it was
overrides
> for given bots, not test output.... Will run the test :)

The whole dm_args.py file is for bot overrides, and then the .json file shows
the current outputs for the bots listed in self_test()'s cases list.
When run in test mode, dm_args.py insists that those test cases give 100% line
coverage inside get_args(), then dumps out an updated .json file so you can `git
diff` them.

I wouldn't worry about passing nvprmsaa4 when it's not supported.  DM will
figure that out up front and ignore it.

[bsalomon, 2015-02-26 16:26:40]

On 2015/02/26 16:21:22, mtklein wrote:
> > I think I guessed wrong about what this file is for. I assumed it was
> overrides
> > for given bots, not test output.... Will run the test :)
> 
> The whole dm_args.py file is for bot overrides, and then the .json file shows
> the current outputs for the bots listed in self_test()'s cases list.
> When run in test mode, dm_args.py insists that those test cases give 100% line
> coverage inside get_args(), then dumps out an updated .json file so you can
`git
> diff` them.

Ok, I'm going down a rat hole trying to install the necessary python packages on
windows to run dm_flags.py. I'll use this patchset for the trys and then
tomorrow when I'm at my linux machine straighten out the json file.

> 
> I wouldn't worry about passing nvprmsaa4 when it's not supported.  DM will
> figure that out up front and ignore it.

Yeah, I think I'll update the CL to run nvprmsaa4 on Android and nvprmsaa16 on
desktop before landing.

[mtklein, 2015-02-26 16:34:51]

Yeah, annoyingly, the coverage package need to run in test mode isn't part of
stock Python.  It's `pip install coverage`, if you've got that.

[bsalomon, 2015-02-26 16:46:46]

On 2015/02/26 16:34:51, mtklein wrote:
> Yeah, annoyingly, the coverage package need to run in test mode isn't part of
> stock Python.  It's `pip install coverage`, if you've got that.

yeah, started to install pip but then hit some permissions issues and threw up
my hands.

[bsalomon, 2015-02-27 18:06:06]

Mike, I couldn't upload to your CL, but I took your patch and added exclusions
based on the try bot failures.

[bsalomon, 2015-02-27 18:06:12]

The CQ bit was checked by bsalomon@google.com

[bsalomon, 2015-02-27 18:06:12]

The patchset sent to the CQ was uploaded after l-g-t-m from
senorblanco@chromium.org
Link to the patchset: https://codereview.chromium.org/958033004/#ps40001 (title:
"Rebase off Mike's patch and add exclusions")

[commit-bot: I haz the power, 2015-02-27 18:06:22]

CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/958033004/40001

[mtklein, 2015-02-27 18:08:32]

lgtm

[commit-bot: I haz the power, 2015-02-27 18:13:41]

Committed patchset #3 (id:40001) as
https://skia.googlesource.com/skia/+/5abf584bcc21a488e04548e5a102265b17c11f83