[Cython] Bug in Cython producing incorrect C code

[Vitja Makarov]

2012/2/4 Vitja Makarov <vitja.makarov at gmail.com>:
> 2012/1/26 mark florisson <markflorisson88 at gmail.com>:
>> On 26 January 2012 19:27, Stefan Behnel <stefan_ml at behnel.de> wrote:
>>> mark florisson, 26.01.2012 20:15:
>>>> On 26 January 2012 18:53, Stefan Behnel wrote:
>>>>> mark florisson, 26.01.2012 16:20:
>>>>>> I think this problem can trivially be solved by creating a ProxyNode
>>>>>> that should never be replaced by any transform, but it's argument may
>>>>>> be replaced. So you wrap self.rhs in a ProxyNode and use that to
>>>>>> create your CloneNodes.
>>>>>
>>>>> I can't see what a ProxyNode would do that a CloneNode shouldn't do anyway.
>>>>
>>>> It wouldn't be a replacement, merely an addition (an extra indirection).
>>>
>>> What I was trying to say was that a ProxyNode would always be required by a
>>> CloneNode, but I don't see where a ProxyNode would be needed outside of a
>>> CloneNode. So it seems rather redundant and I don't know if we need a
>>> separate node for it.
>>
>> Yes it would be needed only for that, but I think the only real
>> alternative is to not use CloneNode at all, i.e. make the
>> transformation Dag mentioned, where you create new rhs (NameNode?)
>> references to the temporary result.
>>
>
> Now it seems to be the only case when we got problem like this. It
> means that clones may be safely created at very late stage.
> So transforming CascadeAssignment into SingleAssignments doesn't solve
> generic problem.
>
> I tried to implement conditional inlining the same problem may happen
> there (ConditionalCallNode owns arguments and replaces
> SimpleCallNode's args with clones). Splitting analyse_expressions()
> would help. On the other hand moving this optimization after
> OptimizeBuiltinCalls() would help too.
>

I tried to introduce finalize_expressions() here:

https://github.com/vitek/cython/tree/_finalize_expressions

I moved arg_tuple creation logic from SimpleCallNode's analyse_types()
to finalize_expressions() so few tests are broken now.

Now inlining is done right before AnalyseExpressions before arg_tuple
is created (before pyobject coercion nodes are created). It must be
run after expression analysis. So I'm completely sure that
analyse_types() must be split.

-- 
vitja.