[Python-Dev] RE: Introducing memprof (was PyErr_NoMemory)

[Vladimir Marangozov]

Tim Peters wrote:
> 
> My bandwidth is consumed by 2.0 issues, so I won't look at it.  On the
> chance that Guido gets hit by a bus, though, and I have time to kill at his
> funeral, it would be nice to have it available on SourceForge.  Uploading a
> postponed patch sounds fine!

Done. Both patches are updated and relative to current CVS:

Optional object malloc:
http://sourceforge.net/patch/?func=detailpatch&patch_id=101104&group_id=5470

Optional memory profiler:
http://sourceforge.net/patch/?func=detailpatch&patch_id=101229&group_id=5470

Let me insist again that these are totally optional and off by default
(lately, a recurrent wish of mine regarding proposed features). 

Since they're optional, off by default, and consitute a solid base for
further work on mem + GC, and despite the tiny imperfections I see in
the profiler, I think I'm gonna push a bit, given that I'm pretty
confident in the code and that it barely affects anything.

So while I'm out of town, my mailbox would be happy to register any
opinions that the python-dev crowd might have (I'm thinking of Barry
and Neil Schemenauer in particular). Also, when BDFL is back from
Palo Alto, give him a chance to emit a statement (although I know
he's not a memory fan <wink>).

I'll *try* to find some time for docs and test cases, but I'd like to
get some preliminary feedback first (especially if someone care to try
this on a 64 bit machine). That's it for now.

-- 
       Vladimir MARANGOZOV          | Vladimir.Marangozov@inrialpes.fr
http://sirac.inrialpes.fr/~marangoz | tel:(+33-4)76615277 fax:76615252