Give isolates names to be used during debugging.

Give isolates names to be used during debugging.

Implement --trace_isolates.

-------------

Sample output of --trace_isolates for the CrossIsolateMessageTest:

[+] Starting isolate:
    isolate:    CrossIsolateMessageTest.dart/main-7112
[+] Starting isolate:
    isolate:    CrossIsolateMessageTest.dart/CrossIsolate1.main-7114
[+] Starting isolate:
    isolate:    CrossIsolateMessageTest.dart/CrossIsolate2.main-7115
[>] Posting message:
    source:     CrossIsolateMessageTest.dart/main-7112
    reply_port: 7117
    dest:       CrossIsolateMessageTest.dart/CrossIsolate2.main-7115
    dest_port:  7115
[>] Posting message:
    source:     CrossIsolateMessageTest.dart/CrossIsolate2.main-7115
    reply_port: 7118
    dest:       CrossIsolateMessageTest.dart/main-7112
    dest_port:  7117
[>] Posting message:
    source:     CrossIsolateMessageTest.dart/main-7112
    reply_port: 7119
    dest:       CrossIsolateMessageTest.dart/CrossIsolate1.main-7114
    dest_port:  7114
[>] Posting message:
    source:     CrossIsolateMessageTest.dart/CrossIsolate1.main-7114
    reply_port: 7120
    dest:       CrossIsolateMessageTest.dart/main-7112
    dest_port:  7119
[>] Posting message:
    source:     CrossIsolateMessageTest.dart/main-7112
    reply_port: 0
    dest:       CrossIsolateMessageTest.dart/CrossIsolate1.main-7114
    dest_port:  7120
[>] Posting message:
    source:     CrossIsolateMessageTest.dart/CrossIsolate1.main-7114
    reply_port: 0
    dest:       CrossIsolateMessageTest.dart/CrossIsolate2.main-7115
    dest_port:  7118
[>] Posting message:
    source:     CrossIsolateMessageTest.dart/CrossIsolate2.main-7115
    reply_port: 0
    dest:       CrossIsolateMessageTest.dart/main-7112
    dest_port:  7116
[>] Posting message:
    source:     CrossIsolateMessageTest.dart/main-7112
    reply_port: 0
    dest:       CrossIsolateMessageTest.dart/main-7112
    dest_port:  7113
[-] Stopping isolate:
    isolate:    CrossIsolateMessageTest.dart/CrossIsolate2.main-7115
[-] Stopping isolate:
    isolate:    CrossIsolateMessageTest.dart/CrossIsolate1.main-7114
[-] Stopping isolate:
    isolate:    CrossIsolateMessageTest.dart/main-7112
Committed: https://code.google.com/p/dart/source/detail?r=3416

[siva, 2012-01-18 06:26:03]

lgtm

https://chromiumcodereview.appspot.com/9242035/diff/1/runtime/bin/main.cc
File runtime/bin/main.cc (right):

https://chromiumcodereview.appspot.com/9242035/diff/1/runtime/bin/main.cc#new...
runtime/bin/main.cc:370: }
Would skipping past slashes cause the name to be non unique?

https://chromiumcodereview.appspot.com/9242035/diff/1/runtime/bin/main.cc#new...
runtime/bin/main.cc:374: char* buffer = new char[len];
ASSERT(buffer != NULL);

https://chromiumcodereview.appspot.com/9242035/diff/1/runtime/lib/isolate.cc
File runtime/lib/isolate.cc (right):

https://chromiumcodereview.appspot.com/9242035/diff/1/runtime/lib/isolate.cc#...
runtime/lib/isolate.cc:258: }
Ditto comment regarding stripping script name.

https://chromiumcodereview.appspot.com/9242035/diff/1/runtime/lib/isolate.cc#...
runtime/lib/isolate.cc:285: const char* isolate_name =
BuildIsolateName(library_url, class_name, "main");
Is the default function "main" or "run" I forget.

https://chromiumcodereview.appspot.com/9242035/diff/1/runtime/vm/isolate.cc
File runtime/vm/isolate.cc (right):

https://chromiumcodereview.appspot.com/9242035/diff/1/runtime/vm/isolate.cc#n...
runtime/vm/isolate.cc:140: result->BuildName(name_prefix);
Should the name be built only if tracing is enabled?

https://chromiumcodereview.appspot.com/9242035/diff/1/runtime/vm/isolate.cc#n...
runtime/vm/isolate.cc:158: }
Under what circumstances would name_ be non null over here?

https://chromiumcodereview.appspot.com/9242035/diff/1/runtime/vm/isolate.cc#n...
runtime/vm/isolate.cc:162: const char* kFormat = "%s-%lld";
Maybe add a TODO here that generation issues in the name will be addressed
later.

[turnidge, 2012-01-18 21:29:20]

https://chromiumcodereview.appspot.com/9242035/diff/1/runtime/bin/main.cc
File runtime/bin/main.cc (right):

https://chromiumcodereview.appspot.com/9242035/diff/1/runtime/bin/main.cc#new...
runtime/bin/main.cc:370: }
On 2012/01/18 06:26:03, asiva wrote:
> Would skipping past slashes cause the name to be non unique?

Yes and no.  If there are two dart scripts with the same file name but different
paths, then the name prefixes will be the same for both of them.  If we were to
include the full path of the script, it would make the name awkwardly long, so I
was trying to strike a balance of providing a prefix that would be usually
unique but still easy to read.

The name prefix here will get a number appended to it inside of isolate.cc, so
in the end then name will be unique regardless of the name prefix generated by
this code.

https://chromiumcodereview.appspot.com/9242035/diff/1/runtime/bin/main.cc#new...
runtime/bin/main.cc:374: char* buffer = new char[len];
On 2012/01/18 06:26:03, asiva wrote:
> ASSERT(buffer != NULL);

Done.

https://chromiumcodereview.appspot.com/9242035/diff/1/runtime/lib/isolate.cc
File runtime/lib/isolate.cc (right):

https://chromiumcodereview.appspot.com/9242035/diff/1/runtime/lib/isolate.cc#...
runtime/lib/isolate.cc:258: }
On 2012/01/18 06:26:03, asiva wrote:
> Ditto comment regarding stripping script name.

Yeah, ditto reply.

https://chromiumcodereview.appspot.com/9242035/diff/1/runtime/lib/isolate.cc#...
runtime/lib/isolate.cc:285: const char* isolate_name =
BuildIsolateName(library_url, class_name, "main");
On 2012/01/18 06:26:03, asiva wrote:
> Is the default function "main" or "run" I forget.

Well, the vm invokes a hidden _run() function on the Isolate base class which
then invokes the abstract main() method which is where the user actually writes
the code.  I figured that since main was more familiar from the user's
standpoint, I would use that to name the isolate.  I could also leave off the
method name entirely.

https://chromiumcodereview.appspot.com/9242035/diff/1/runtime/vm/isolate.cc
File runtime/vm/isolate.cc (right):

https://chromiumcodereview.appspot.com/9242035/diff/1/runtime/vm/isolate.cc#n...
runtime/vm/isolate.cc:140: result->BuildName(name_prefix);
On 2012/01/18 06:26:03, asiva wrote:
> Should the name be built only if tracing is enabled?

I plan to also use it when reporting top-level uncaught exceptions and the like.

https://chromiumcodereview.appspot.com/9242035/diff/1/runtime/vm/isolate.cc#n...
runtime/vm/isolate.cc:158: }
On 2012/01/18 06:26:03, asiva wrote:
> Under what circumstances would name_ be non null over here?

It can't happen.  It used to happen in an earlier version of my change. 
Changing this to an ASSERT.

https://chromiumcodereview.appspot.com/9242035/diff/1/runtime/vm/isolate.cc#n...
runtime/vm/isolate.cc:162: const char* kFormat = "%s-%lld";
On 2012/01/18 06:26:03, asiva wrote:
> Maybe add a TODO here that generation issues in the name will be addressed
> later.

What generation issues?

[siva, 2012-01-18 22:06:08]

lgtm

https://chromiumcodereview.appspot.com/9242035/diff/1/runtime/vm/isolate.cc
File runtime/vm/isolate.cc (right):

https://chromiumcodereview.appspot.com/9242035/diff/1/runtime/vm/isolate.cc#n...
runtime/vm/isolate.cc:140: result->BuildName(name_prefix);
Can this be done lazily when it is actually needed?
It would be nice to keep isolate creation and init times low in order to have
fast startup times.

On 2012/01/18 21:29:21, turnidge wrote:
> On 2012/01/18 06:26:03, asiva wrote:
> > Should the name be built only if tracing is enabled?
> 
> I plan to also use it when reporting top-level uncaught exceptions and the
like.

https://chromiumcodereview.appspot.com/9242035/diff/1/runtime/vm/isolate.cc#n...
runtime/vm/isolate.cc:162: const char* kFormat = "%s-%lld";
i.e if an isolate dies and another isolate is created and it happens to get the
same main port number. In the trace the name would be same but it is actually a
new isolate.

On 2012/01/18 21:29:21, turnidge wrote:
> On 2012/01/18 06:26:03, asiva wrote:
> > Maybe add a TODO here that generation issues in the name will be addressed
> > later.
> 
> What generation issues?

[Ivan Posva, 2012-01-20 16:28:29]

https://chromiumcodereview.appspot.com/9242035/diff/9001/runtime/include/dart...
File runtime/include/dart_api.h (right):

https://chromiumcodereview.appspot.com/9242035/diff/9001/runtime/include/dart...
runtime/include/dart_api.h:248: typedef bool (*Dart_IsolateCreateCallback)(const
char* name_prefix,
name_prefix and callback_data are not documented anywhere.

https://chromiumcodereview.appspot.com/9242035/diff/9001/runtime/include/dart...
runtime/include/dart_api.h:320: * Requires there to be no current isolate.
I don't think this is the case. Please check with Siva, but as far as I remember
it just overrides the current isolate.

https://chromiumcodereview.appspot.com/9242035/diff/9001/runtime/include/dart...
runtime/include/dart_api.h:328: DART_EXPORT Dart_Isolate
Dart_CreateIsolate(const char* name_prefix,
name_prefix and callback_data are not documented.

[turnidge, 2012-01-26 18:07:11]

I thought I had published these earlier.  This has been checked in at this
point.  Let me know if you want me to go back and change things.

:-/

Todd

http://codereview.chromium.org/9242035/diff/1/runtime/vm/isolate.cc
File runtime/vm/isolate.cc (right):

http://codereview.chromium.org/9242035/diff/1/runtime/vm/isolate.cc#newcode140
runtime/vm/isolate.cc:140: result->BuildName(name_prefix);
Well, the information being passed to this function, the name prefix, is
somewhat ephemeral.  I can't rely on it living past this function call.  So I
would have to copy it in any case, right?

On 2012/01/18 22:06:08, asiva wrote:
> Can this be done lazily when it is actually needed?
> It would be nice to keep isolate creation and init times low in order to have
> fast startup times.
> 
> On 2012/01/18 21:29:21, turnidge wrote:
> > On 2012/01/18 06:26:03, asiva wrote:
> > > Should the name be built only if tracing is enabled?
> > 
> > I plan to also use it when reporting top-level uncaught exceptions and the
> like.
>

http://codereview.chromium.org/9242035/diff/1/runtime/vm/isolate.cc#newcode162
runtime/vm/isolate.cc:162: const char* kFormat = "%s-%lld";
I do not believe that we reuse port numbers currently.

On 2012/01/18 22:06:08, asiva wrote:
> i.e if an isolate dies and another isolate is created and it happens to get
the
> same main port number. In the trace the name would be same but it is actually
a
> new isolate.
> 
> On 2012/01/18 21:29:21, turnidge wrote:
> > On 2012/01/18 06:26:03, asiva wrote:
> > > Maybe add a TODO here that generation issues in the name will be addressed
> > > later.
> > 
> > What generation issues?
>