Move reusable handles from isolate to thread.

Move reusable handles from isolate to thread.

BUG=
R=koda@google.com

Committed: https://github.com/dart-lang/sdk/commit/53e814f0ddbd2959a4229a28fb202ae14ae417e3

[srdjan, 2015-10-07 20:53:48]

srdjan@google.com changed reviewers:
+ asiva@google.com, koda@google.com

[koda, 2015-10-07 21:02:54]

https://codereview.chromium.org/1394673002/diff/20001/runtime/vm/thread.cc
File runtime/vm/thread.cc (right):

https://codereview.chromium.org/1394673002/diff/20001/runtime/vm/thread.cc#ne...
runtime/vm/thread.cc:237: thread->isolate_ = NULL;
In debug mode, please zap all the reusable handles when exiting an isolate.

As previously mentioned, we could also add debug mode checks that the isolate is
never exited during a reusable handle scope.

https://codereview.chromium.org/1394673002/diff/20001/runtime/vm/thread.h
File runtime/vm/thread.h (right):

https://codereview.chromium.org/1394673002/diff/20001/runtime/vm/thread.h#new...
runtime/vm/thread.h:404: 
Excessive blank lines.

[srdjan, 2015-10-08 16:22:01]

https://codereview.chromium.org/1394673002/diff/20001/runtime/vm/thread.cc
File runtime/vm/thread.cc (right):

https://codereview.chromium.org/1394673002/diff/20001/runtime/vm/thread.cc#ne...
runtime/vm/thread.cc:237: thread->isolate_ = NULL;
On 2015/10/07 21:02:54, koda wrote:
> In debug mode, please zap all the reusable handles when exiting an isolate.

ReusableHandleScope already makes sure that nobody can access the handles
outside the scope. I could clear them by reallocating them anew in
Thread::ExitIsolate, but I am not sure if it would improve reliability.

> 
> As previously mentioned, we could also add debug mode checks that the isolate
is
> never exited during a reusable handle scope.

I could add a check that sees that none of the reusable handle scopes are
active. (not yet uploaded)

https://codereview.chromium.org/1394673002/diff/20001/runtime/vm/thread.h
File runtime/vm/thread.h (right):

https://codereview.chromium.org/1394673002/diff/20001/runtime/vm/thread.h#new...
runtime/vm/thread.h:404: 
On 2015/10/07 21:02:54, koda wrote:
> Excessive blank lines.

Done.

[koda, 2015-10-08 17:13:02]

On 2015/10/08 16:22:01, srdjan wrote:
> https://codereview.chromium.org/1394673002/diff/20001/runtime/vm/thread.cc
> File runtime/vm/thread.cc (right):
> 
>
https://codereview.chromium.org/1394673002/diff/20001/runtime/vm/thread.cc#ne...
> runtime/vm/thread.cc:237: thread->isolate_ = NULL;
> On 2015/10/07 21:02:54, koda wrote:
> > In debug mode, please zap all the reusable handles when exiting an isolate.
> 
> ReusableHandleScope already makes sure that nobody can access the handles
> outside the scope. I could clear them by reallocating them anew in
> Thread::ExitIsolate, but I am not sure if it would improve reliability.

Ack.

> 
> > 
> > As previously mentioned, we could also add debug mode checks that the
isolate
> is
> > never exited during a reusable handle scope.
> 
> I could add a check that sees that none of the reusable handle scopes are
> active. (not yet uploaded)

Yes, that should be sufficient to prevent the patterns I'm worried out.

> 
> https://codereview.chromium.org/1394673002/diff/20001/runtime/vm/thread.h
> File runtime/vm/thread.h (right):
> 
>
https://codereview.chromium.org/1394673002/diff/20001/runtime/vm/thread.h#new...
> runtime/vm/thread.h:404: 
> On 2015/10/07 21:02:54, koda wrote:
> > Excessive blank lines.
> 
> Done.

[koda, 2015-10-08 17:13:15]

lgtm

[srdjan, 2015-10-09 17:13:46]

Committed patchset #5 (id:80001) manually as
53e814f0ddbd2959a4229a28fb202ae14ae417e3 (presubmit successful).