Make serializer non-static.

Make serializer non-static.

R=svenpanne@chromium.org

Committed: https://code.google.com/p/v8/source/detail?r=21430

[Sven Panne, 2014-05-22 08:04:15]

Nice cleanup. Tested with --no-snapshot, valgrind etc., I hope... ;-) The only
thing which is unclear to me is the change in the use_crankshaft() predicate.

https://codereview.chromium.org/296853007/diff/1/src/isolate.h
File src/isolate.h (right):

https://codereview.chromium.org/296853007/diff/1/src/isolate.h#newcode968
src/isolate.h:968: bool serializer_enabled() { return serializer_enabled_; }
const

https://codereview.chromium.org/296853007/diff/1/src/isolate.h#newcode974
src/isolate.h:974: return use_crankshaft_ && !serializer_enabled_;
Hmmm, why do we need this additional term?

[Yang, 2014-05-22 08:46:06]

Changed slightly. What I can say is that at there are at least no new leaks
added by this CL. I'll work on reducing the leaks though.

https://codereview.chromium.org/296853007/diff/1/src/isolate.h
File src/isolate.h (right):

https://codereview.chromium.org/296853007/diff/1/src/isolate.h#newcode968
src/isolate.h:968: bool serializer_enabled() { return serializer_enabled_; }
On 2014/05/22 08:04:15, Sven Panne wrote:
> const

Done.

https://codereview.chromium.org/296853007/diff/1/src/isolate.h#newcode974
src/isolate.h:974: return use_crankshaft_ && !serializer_enabled_;
On 2014/05/22 08:04:15, Sven Panne wrote:
> Hmmm, why do we need this additional term?

My mistake. Some point during refactoring I thought it was a good idea to be
able to disable serializer.

I now changed this so that the serializer state has to be set before
Isolate::Init, and use_crankshaft_ is set based on the serializer state.

[Yang, 2014-05-22 08:54:27]

On 2014/05/22 08:46:06, Yang wrote:
> Changed slightly. What I can say is that at there are at least no new leaks
> added by this CL. I'll work on reducing the leaks though.
> 


Nvm my comment. It seems that if d8 exits normally, no leaks occur. If d8 exits
for example due to uncaught exception, we leave a mess.

> https://codereview.chromium.org/296853007/diff/1/src/isolate.h
> File src/isolate.h (right):
> 
> https://codereview.chromium.org/296853007/diff/1/src/isolate.h#newcode968
> src/isolate.h:968: bool serializer_enabled() { return serializer_enabled_; }
> On 2014/05/22 08:04:15, Sven Panne wrote:
> > const
> 
> Done.
> 
> https://codereview.chromium.org/296853007/diff/1/src/isolate.h#newcode974
> src/isolate.h:974: return use_crankshaft_ && !serializer_enabled_;
> On 2014/05/22 08:04:15, Sven Panne wrote:
> > Hmmm, why do we need this additional term?
> 
> My mistake. Some point during refactoring I thought it was a good idea to be
> able to disable serializer.
> 
> I now changed this so that the serializer state has to be set before
> Isolate::Init, and use_crankshaft_ is set based on the serializer state.

[Sven Panne, 2014-05-22 08:56:49]

On 2014/05/22 08:46:06, Yang wrote:
> [...] I now changed this so that the serializer state has to be set before
> Isolate::Init, and use_crankshaft_ is set based on the serializer state.

Hmmm, I think that is still not completely right: It would really simplify
things quite a lot when we can merge the Isolate constructor and Isolate::Init
into just the constructor. Previously this split was necessary because of the
default Isolate which was half-constructed before main(), but this reason has
gone.

use_crankshaft_ seems to be just a tiny cache, can't we use inline the stuff
into the use_crankshaft() predicate and kill use_crankshaft_?

[Yang, 2014-05-22 09:22:50]

On 2014/05/22 08:56:49, Sven Panne wrote:
> On 2014/05/22 08:46:06, Yang wrote:
> > [...] I now changed this so that the serializer state has to be set before
> > Isolate::Init, and use_crankshaft_ is set based on the serializer state.
> 
> Hmmm, I think that is still not completely right: It would really simplify
> things quite a lot when we can merge the Isolate constructor and Isolate::Init
> into just the constructor. Previously this split was necessary because of the
> default Isolate which was half-constructed before main(), but this reason has
> gone.
> 
> use_crankshaft_ seems to be just a tiny cache, can't we use inline the stuff
> into the use_crankshaft() predicate and kill use_crankshaft_?

Done.

[Sven Panne, 2014-05-22 09:26:54]

lgtm

[Yang, 2014-05-22 09:36:33]

Committed patchset #3 manually as r21430 (presubmit successful).