[pypy-dev] cpyext: Detecting pypy and other issues

[Maciej Fijalkowski]

On Thu, May 12, 2011 at 5:34 PM, Roger Binns <rogerb at rogerbinns.com> wrote:
> -----BEGIN PGP SIGNED MESSAGE-----
> Hash: SHA1
>
> On 05/12/2011 01:34 AM, Maciej Fijalkowski wrote:
>>>> 32bit pypy works out of the box on valgrind. No idea why 64bit does not
>
> Are you sure it plays nice with the standard memory allocation functions
> (malloc/free)?  That is unlikely due to use of a GC.  The doc I wrote shows
> what you have to do to get CPython to play nice.

Are you using callgrind or using valgrind for memory checks? The
former should work, the latter is rather pointless I think, because
RPython manages memory on it's own.

>
> Additionally I need pypy's behaviour to be deterministic.  I was getting
> different behaviour from just running a test versus running under gdb with a
> breakpoint in one of my C functions and just hitting continue once it hit!

If you have a bug in cpyext somehow (or pypy does, which is also a
possibility) you won't get deterministic behavior due to way memory
works in C.

>
> The most promising thing looked to me to be the refcounting GC but that
> doesn't compile - the error message implied it had something to do with not
> supporting weakrefs.

No, never use refcounting GC for anything. What's the problem with say minimark?

>
> Roger
> -----BEGIN PGP SIGNATURE-----
> Version: GnuPG v1.4.11 (GNU/Linux)
>
> iEYEARECAAYFAk3L/hQACgkQmOOfHg372QQVRwCdEFJKlDHJzW7V5bvGD57tNyfy
> ISMAoL2j4+zO3fM5wxcYTSgoBE63JFU1
> =u/ST
> -----END PGP SIGNATURE-----
> _______________________________________________
> pypy-dev mailing list
> pypy-dev at python.org
> http://mail.python.org/mailman/listinfo/pypy-dev
>