gen-postmortem-metadata.py: fix code generator

gen-postmortem-metadata.py: fix code generator

* update tools/gen-postmortem-metadata.py
* add postmortem=on switch to the Makefile

Patch by Ben Noordhuis

[piscisaureus, 2012-08-16 15:29:14]

PTAL

[piscisaureus, 2012-08-16 16:17:30]

Added Dave Pacheco as a reviewer

[dave.pacheco, 2012-08-16 16:25:56]

On 2012/08/16 16:17:30, piscisaureus wrote:
> Added Dave Pacheco as a reviewer

Thanks for looping me in, and thanks Ben for doing this change.  Has this been
in Node 0.8 already, or has someone validated that they still work with mdb_v8?

[noordhuis, 2012-08-16 20:13:13]

On 2012/08/16 16:25:56, dave.pacheco wrote:
> Thanks for looping me in, and thanks Ben for doing this change.  Has this been
> in Node 0.8 already, or has someone validated that they still work with
mdb_v8?

No and no. It's against V8's bleeding_edge HEAD which has a considerable delta
with the V8 that ships with node 0.8 (or node master, for that matter).

[Michael Starzinger, 2012-08-16 23:33:21]

Looping in Toon, who did the changes that this CL is adapting against.

[Toon Verwaest, 2012-08-21 13:29:46]

Can someone clarify the context of this CL? Michael is currently on holidays and
I do not really know what I'm looking at.

The proposed changes seem to alter the semantics of what the script originally
intended to do. Is this intentional?

http://codereview.chromium.org/10827379/diff/1/tools/gen-postmortem-metadata.py
File tools/gen-postmortem-metadata.py (right):

http://codereview.chromium.org/10827379/diff/1/tools/gen-postmortem-metadata....
tools/gen-postmortem-metadata.py:86: 'value': 'TransitionArray::kFirstIndex' },
All these changes change the semantics of the fields. Is this on purpose? Do we
care about transitions rather than descriptors now?

http://codereview.chromium.org/10827379/diff/1/tools/gen-postmortem-metadata....
tools/gen-postmortem-metadata.py:113: 'Map, instance_descriptors, uintptr_t,
kTransitionsOrBackPointerOffset',
This field does not contain the instance_descriptors but rather either the
transition array or the back pointer. If it contains a transition array, that
array will store the descriptors.

[noordhuis, 2012-08-28 20:19:22]

On 2012/08/21 13:29:46, Toon Verwaest wrote:
> Can someone clarify the context of this CL? Michael is currently on holidays
and
> I do not really know what I'm looking at.
> 
> The proposed changes seem to alter the semantics of what the script originally
> intended to do. Is this intentional?
> 
>
http://codereview.chromium.org/10827379/diff/1/tools/gen-postmortem-metadata.py
> File tools/gen-postmortem-metadata.py (right):
> 
>
http://codereview.chromium.org/10827379/diff/1/tools/gen-postmortem-metadata....
> tools/gen-postmortem-metadata.py:86: 'value': 'TransitionArray::kFirstIndex'
},
> All these changes change the semantics of the fields. Is this on purpose? Do
we
> care about transitions rather than descriptors now?
> 
>
http://codereview.chromium.org/10827379/diff/1/tools/gen-postmortem-metadata....
> tools/gen-postmortem-metadata.py:113: 'Map, instance_descriptors, uintptr_t,
> kTransitionsOrBackPointerOffset',
> This field does not contain the instance_descriptors but rather either the
> transition array or the back pointer. If it contains a transition array, that
> array will store the descriptors.

I think it's best if I retract this patchset and let someone with a vested
interest in postmortem support (*hint* Joyent *hint*) address it.

The issue is that node.js won't build with V8 HEAD right now but I can work
around that by disabling postmortem support for now.

[Toon Verwaest, 2013-05-31 12:01:33]

Marked as closed to due inactivity. Just file a new issue if there's a proper
fix (and if it's still relevant).