Use counter for legacy const.

Use counter for legacy const.

We only report the usages when full parse is happening, i.e. only when the function that declares a legacy const is compiled. This is an approximation that is easy to implement, but still should reflect the real-world usage.

BUG=v8:3942
LOG=N

Committed: https://crrev.com/50f4964085e065807254761783bb8e5dcd1aa23c
Cr-Commit-Position: refs/heads/master@{#27526}

[Dmitry Lomov (no reviews), 2015-03-30 13:39:00]

ptal

[Dmitry Lomov (no reviews), 2015-03-30 13:39:55]

dslomov@chromium.org changed reviewers:
+ rossberg@chromium.org, vogelheim@chromium.org

[vogelheim, 2015-03-30 15:32:33]

lgtm

[Dmitry Lomov (no reviews), 2015-03-30 15:35:09]

The CQ bit was checked by dslomov@chromium.org

[commit-bot: I haz the power, 2015-03-30 15:35:20]

CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/1041863002/20001

[commit-bot: I haz the power, 2015-03-30 16:20:59]

Committed patchset #2 (id:20001)

[commit-bot: I haz the power, 2015-03-30 16:21:16]

Patchset 2 (id:??) landed as
https://crrev.com/50f4964085e065807254761783bb8e5dcd1aa23c
Cr-Commit-Position: refs/heads/master@{#27526}

[adamk, 2015-03-31 18:19:31]

adamk@chromium.org changed reviewers:
+ adamk@chromium.org

[adamk, 2015-03-31 18:19:32]

Where's the Blink side of this change?

https://codereview.chromium.org/1041863002/diff/20001/src/parser.cc
File src/parser.cc (right):

https://codereview.chromium.org/1041863002/diff/20001/src/parser.cc#newcode2275
src/parser.cc:2275: ++use_counts_[v8::Isolate::kLegacyConst];
Counting each const in a script separately seems a bit odd, especially since
Blink will collapse all these into one. But I guess that's not particularly
problematic. Just wondering if it's intentional that the V8 API for UseCounters
doesn't match how they're used in Blink.

[Dmitry Lomov (no reviews), 2015-03-31 18:47:48]

On 2015/03/31 18:19:32, adamk wrote:
> Where's the Blink side of this change?

I'll get to it once this rolls.

> 
> https://codereview.chromium.org/1041863002/diff/20001/src/parser.cc
> File src/parser.cc (right):
> 
>
https://codereview.chromium.org/1041863002/diff/20001/src/parser.cc#newcode2275
> src/parser.cc:2275: ++use_counts_[v8::Isolate::kLegacyConst];
> Counting each const in a script separately seems a bit odd, especially since
> Blink will collapse all these into one. But I guess that's not particularly
> problematic. Just wondering if it's intentional that the V8 API for
UseCounters
> doesn't match how they're used in Blink.

Yes, it is intentional. Embedder is free to collapse use counters in any way it
wants, we still report every usage.

[adamk, 2015-03-31 18:49:02]

On 2015/03/31 18:47:48, Dmitry Lomov (chromium) wrote:
> On 2015/03/31 18:19:32, adamk wrote:
> > Where's the Blink side of this change?
> 
> I'll get to it once this rolls.
> 
> > 
> > https://codereview.chromium.org/1041863002/diff/20001/src/parser.cc
> > File src/parser.cc (right):
> > 
> >
>
https://codereview.chromium.org/1041863002/diff/20001/src/parser.cc#newcode2275
> > src/parser.cc:2275: ++use_counts_[v8::Isolate::kLegacyConst];
> > Counting each const in a script separately seems a bit odd, especially since
> > Blink will collapse all these into one. But I guess that's not particularly
> > problematic. Just wondering if it's intentional that the V8 API for
> UseCounters
> > doesn't match how they're used in Blink.
> 
> Yes, it is intentional. Embedder is free to collapse use counters in any way
it
> wants, we still report every usage.

Okay, thanks for the confirmation.