[Stackless] Current issues

Bob Ippolito bob at redivi.com
Mon Feb 16 05:50:18 CET 2004


On Feb 15, 2004, at 9:34 PM, Christian Tismer wrote:

>> This is off the top of my head.. we really need an issue tracker.  
>> Should I install/configure roundup on the stackless.com server?
>
> There is just one real bug issue left, so I don't need a need
> for an issue tracker, yet. Unless we weant to use that for
> features.
> But I don't liek that too much, since that makes missing
> features look like bugs...

One real bug left *that we know of* ;)  The test suite we have isn't 
exactly comprehensive.  Every issue tracker I've ever seen makes it 
pretty easy to distinguish between bugs and features, so that is not a 
problem.

>> Missing features:
>> [no test coverage] itertools types can not be pickled (how should 
>> this be implemented, Christian?)
>
> I'm not familiar with itertools, yet.
> I think there are lots of other objects which cannot be
> pickled, yet, and it is not that urgent to invent all of that.
>
> Well, looking at one of these, "takewhile":
> it has the structure
>
> typedef struct {
> 	PyObject_HEAD
> 	PyObject *func;
> 	PyObject *it;
> 	long	 stop;
> } takewhileobject;
>
> which looks like an easy to implement pickle?
> Again, not urgent. What we need is correctness,
> not completeness.

itertools is pretty important, I use imap/izip/etc. a lot more than I 
use the list versions.

>> [module case covered by test_pickle] the 'pricklepit' breaks the 
>> __new__ behavior of a few types, most notably function and module
>
> I don't see how functions break the pickling; this seems
> to be compatible.

Ok, I may have been wrong about functions.. I could've sworn that 
someone else had a problem with them.

> Not so with modules, it is semantically different,
> since the types.ModuleType(name) does not get imported.
>
> code objects seem to have a similar problem.
> What I will do is to try to remove all factories
> which are existent already and adjust them if they
> use a different way to initialize.
>
> For the moment, I changed init_type to enforce
> that tp_new was not already set, but then
> patched the three remaining problems with setting
> tp_new to NULL and a comment like
>
> /* XXX tp_new hack: need to rewrite this XXX */
>
> The implementation must here relly be changed.
>
> I'm especially unhappy that modules are so weird now.
> Modules can just be created without importing.
> How do we pickle that kind of modules ???

Save everything in its dictionary, of course... should work just fine 
so long as you can pickle classes and functions that you don't have the 
code for.  Since we can do lamba, there is no reason we shouldn't be 
able to do this kind of module.

-bob


_______________________________________________
Stackless mailing list
Stackless at stackless.com
http://www.stackless.com/mailman/listinfo/stackless



More information about the Stackless mailing list