Configure synthetic delays through TraceLog

Configure synthetic delays through TraceLog

Make it possible to configure synthetic delays using trace categories.
For example, the category filter "DELAY(gpu.SwapBuffers;16)" would make
swap buffers take at least 16 ms to complete.
    
BUG=307841
Committed: https://src.chromium.org/viewvc/chrome?view=rev&revision=244207

[Sami, 2013-12-03 17:45:08]

Nat, is this along the lines of what you had in mind about using category
filters? Here essentially each page would set a custom category like
"gpu.SwapBuffers=0.1;static".

[Sami, 2013-12-11 18:11:52]

Friendly ping? I'll do another pass to clean this up if you think this approach
makes sense.

[Sami, 2013-12-12 21:00:40]

Okay, cleaned this up like we chatted. PTAL, thanks!

[ernstm, 2013-12-17 19:14:11]

Why do you add the synthetic delay configuration to the event filter? Those are
two very different concepts.

[Sami, 2013-12-17 19:23:56]

On 2013/12/17 19:14:11, Manfred Ernst wrote:
> Why do you add the synthetic delay configuration to the event filter? Those
are
> two very different concepts.

We tried various IPC approaches for hooking up the delays and in the end we
thought it's most practical to reuse the existing infrastructure by making the
category filter more expressive. Here's the patch for the TracingController
side: https://codereview.chromium.org/98953002/.

[ernstm, 2013-12-17 20:10:12]

On 2013/12/17 19:23:56, Sami wrote:
> On 2013/12/17 19:14:11, Manfred Ernst wrote:
> > Why do you add the synthetic delay configuration to the event filter? Those
> are
> > two very different concepts.
> 
> We tried various IPC approaches for hooking up the delays and in the end we
> thought it's most practical to reuse the existing infrastructure by making the
> category filter more expressive. Here's the patch for the TracingController
> side: https://codereview.chromium.org/98953002/.

But a synthetic delay isn't a category filter. It is something completely
different. Merging the two concepts sounds like a shortcut we shouldn't take.
Why can't we create a separate mechanism for setting the delays?

[Sami, 2013-12-18 16:12:30]

On 2013/12/17 20:10:12, Manfred Ernst wrote:
> But a synthetic delay isn't a category filter. It is something completely
> different. Merging the two concepts sounds like a shortcut we shouldn't take.
> Why can't we create a separate mechanism for setting the delays?

I agree that it's not ideal. Maybe there's a way to formalize the more
expressive category filters better. Something like a trace configuration object.
Nat, any thoughts?

For reference, here are the two alternate approaches I tried: through the gpu
benchmarking extension (https://codereview.chromium.org/83483007/#ps70001) and
devtools (https://codereview.chromium.org/99543002/).

[nduca, 2014-01-07 19:22:32]

lgtm but lets prefix the delay=x with a keyword so its like SET_DELAY(x,y) or
DELAY(x)=y or etc

[nduca, 2014-01-07 19:24:56]

My rationale regarding manfred's (valid) concerns:
- This is good enough to start experimenting
- Piggybacking on categoryfilter lets us try this idea out until we see if its
good enough to really make official
- We should do renaming of CategoryFilter to something more general as a
followup patch to reflect its new role

[dsinclair1, 2014-01-07 19:52:40]

https://codereview.chromium.org/98953002/diff/60001/base/debug/trace_event_im...
File base/debug/trace_event_impl.cc (right):

https://codereview.chromium.org/98953002/diff/60001/base/debug/trace_event_im...
base/debug/trace_event_impl.cc:2320: }
Do we need to merge the delay lists as well?

[Sami, 2014-01-09 19:20:42]

> lgtm but lets prefix the delay=x with a keyword so its like SET_DELAY(x,y) or
> DELAY(x)=y or etc

Done. I went with DELAY(delay_name;option;option;...).

On 2014/01/07 19:52:40, dsinclair1 wrote:
>
https://codereview.chromium.org/98953002/diff/60001/base/debug/trace_event_im...
> File base/debug/trace_event_impl.cc (right):
> 
>
https://codereview.chromium.org/98953002/diff/60001/base/debug/trace_event_im...
> base/debug/trace_event_impl.cc:2320: }
> Do we need to merge the delay lists as well?

Yes, thanks for pointing this out. I've implemented merging and added a test for
it.

[dsinclair1, 2014-01-09 19:23:02]

lgtm

[commit-bot: I haz the power, 2014-01-09 19:27:04]

CQ is trying da patch. Follow status at
https://chromium-status.appspot.com/cq/skyostil@chromium.org/98953002/170001

[commit-bot: I haz the power, 2014-01-09 20:42:28]

Retried try job too often on ios_dbg_simulator for step(s) base_unittests
http://build.chromium.org/p/tryserver.chromium/buildstatus?builder=ios_dbg_si...

[commit-bot: I haz the power, 2014-01-10 10:35:48]

CQ is trying da patch. Follow status at
https://chromium-status.appspot.com/cq/skyostil@chromium.org/98953002/170001

[commit-bot: I haz the power, 2014-01-10 13:53:45]

CQ is trying da patch. Follow status at
https://chromium-status.appspot.com/cq/skyostil@chromium.org/98953002/450001

[commit-bot: I haz the power, 2014-01-10 18:48:07]

Change committed as 244207