Introduce v8::Exception::GetMessage to find location of an error object.

Introduce v8::Exception::GetMessage to find location of an error object.

API=v8::Exception::GetMessage
BUG=chromium:427954
R=yangguo@chromium.org
LOG=Y

Committed: https://code.google.com/p/v8/source/detail?r=25015

Committed: https://code.google.com/p/v8/source/detail?r=25021

[Yang, 2014-10-29 17:59:01]

can you explain when this would be necessary?

[aandrey, 2014-10-30 09:52:19]

If DevTools is closed there is no stack trace on unhandled rejection messages
and no information about location: line/column numbers, scriptId (unlike on
uncaught exceptions where this info is passed via v8::Message). So, this makes
rejection messages almost useless in this case.

Also I deprecated v8::Exception::GetStackTrace() and
v8::PromiseRejectMessage::GetStackTrace() in favor of this new API.

The reason why I don't add v8::Message to v8::PromiseRejectMessage is that for a
promise rejected with a DOMException we have to extract a hidden V8 Error object
first from the DOMException, and then get the v8::Message (location & stack
trace) for that error object, since DOMException has neither simple stack trace
nor detailed stack trace.

[Yang, 2014-10-30 12:32:38]

https://codereview.chromium.org/687253002/diff/20001/src/isolate.cc
File src/isolate.cc (right):

https://codereview.chromium.org/687253002/diff/20001/src/isolate.cc#newcode1258
src/isolate.cc:1258: ComputeLocation(&potential_computed_location);
If you are going to compute the location in CreateMessage anyways, can we omit
it here and put a call to ComputeLocation in ReportBootstrapingException?

https://codereview.chromium.org/687253002/diff/20001/src/isolate.cc#newcode1271
src/isolate.cc:1271: thread_local_top()->pending_message_script_ =
*location->script();
This can't work, if the location is &potential_computed_location in
CreateMessage, since it's stack allocated and already torn down.

I don't think CreateMessage and ComputeLocation guarantee to successfully
compute the location, so I think the check is still necessary.

[aandrey, 2014-10-30 12:47:11]

https://codereview.chromium.org/687253002/diff/20001/src/isolate.cc
File src/isolate.cc (right):

https://codereview.chromium.org/687253002/diff/20001/src/isolate.cc#newcode1258
src/isolate.cc:1258: ComputeLocation(&potential_computed_location);
On 2014/10/30 12:32:38, Yang wrote:
> If you are going to compute the location in CreateMessage anyways, can we omit
> it here and put a call to ComputeLocation in ReportBootstrapingException?

But we need it below also to assign pending_message_script_ and etc.

https://codereview.chromium.org/687253002/diff/20001/src/isolate.cc#newcode1271
src/isolate.cc:1271: thread_local_top()->pending_message_script_ =
*location->script();
On 2014/10/30 12:32:38, Yang wrote:
> This can't work, if the location is &potential_computed_location in
> CreateMessage, since it's stack allocated and already torn down.
> 
> I don't think CreateMessage and ComputeLocation guarantee to successfully
> compute the location, so I think the check is still necessary.

The "location" variable here is not changed by CreateMessage.

https://codereview.chromium.org/687253002/diff/20001/src/isolate.cc#newcode1271
src/isolate.cc:1271: thread_local_top()->pending_message_script_ =
*location->script();
On 2014/10/30 12:32:38, Yang wrote:
> This can't work, if the location is &potential_computed_location in
> CreateMessage, since it's stack allocated and already torn down.
> 
> I don't think CreateMessage and ComputeLocation guarantee to successfully
> compute the location, so I think the check is still necessary.

The "location" variable here is not altered by CreateMessage().

[Yang, 2014-10-30 12:54:22]

On 2014/10/30 12:47:11, aandrey wrote:
> https://codereview.chromium.org/687253002/diff/20001/src/isolate.cc
> File src/isolate.cc (right):
> 
>
https://codereview.chromium.org/687253002/diff/20001/src/isolate.cc#newcode1258
> src/isolate.cc:1258: ComputeLocation(&potential_computed_location);
> On 2014/10/30 12:32:38, Yang wrote:
> > If you are going to compute the location in CreateMessage anyways, can we
omit
> > it here and put a call to ComputeLocation in ReportBootstrapingException?
> 
> But we need it below also to assign pending_message_script_ and etc.
> 
>
https://codereview.chromium.org/687253002/diff/20001/src/isolate.cc#newcode1271
> src/isolate.cc:1271: thread_local_top()->pending_message_script_ =
> *location->script();
> On 2014/10/30 12:32:38, Yang wrote:
> > This can't work, if the location is &potential_computed_location in
> > CreateMessage, since it's stack allocated and already torn down.
> > 
> > I don't think CreateMessage and ComputeLocation guarantee to successfully
> > compute the location, so I think the check is still necessary.
> 
> The "location" variable here is not changed by CreateMessage.
> 
>
https://codereview.chromium.org/687253002/diff/20001/src/isolate.cc#newcode1271
> src/isolate.cc:1271: thread_local_top()->pending_message_script_ =
> *location->script();
> On 2014/10/30 12:32:38, Yang wrote:
> > This can't work, if the location is &potential_computed_location in
> > CreateMessage, since it's stack allocated and already torn down.
> > 
> > I don't think CreateMessage and ComputeLocation guarantee to successfully
> > compute the location, so I think the check is still necessary.
> 
> The "location" variable here is not altered by CreateMessage().

I see. I was a bit confused. LGTM.

[aandrey, 2014-10-30 13:18:07]

Committed patchset #2 (id:20001) manually as 25015 (presubmit successful).

[aandrey, 2014-10-30 14:51:29]

Committed patchset #3 (id:40001) manually as 25021 (presubmit successful).