Wrong decl_type for function pointers ?

[teto]

When the argument is a function pointer, there seems to be a few problems:

For instance for the function int atexit(void (*function)(void));, the pygccxml code decl.arguments[0].decl_type returns: void (*)( ) * which has 1 more asterisk then necessary ?

Nb: I made a mistake and ended up posting it first here
CastXML/CastXML#67

[iMichka]

Thanks for reporting.

I will be on holidays for 3 weeks, so I will only able to have a look at this end of August. I will probably not have internet so I won't be able to respond to any questions/mails.

[teto]

No problem, I have a workaround :)
Before you leave ( if you have the time, else no worry ) could you give me pointers (like what file to look into, what do you think is wrong) for #62 which is the only blocking pb for me. I might be able to provide a fix.
Enjoy your holyday to the fullest.

[iMichka]

Hi

I had a look at this problem. I do not think that it is a bug at first.

If you print type(decl.arguments[0].decl_type) you will see that the type is of pointer_t. Normally you would then use declarations.remove_pointer() to remove that type level and access the underlying free_function_type_t. The declarations.remove_pointer() function was broken and I fixed that, make sure you use the latest checkout from the develop branch.

I wrote an example to explain how these nested types work:
http://pygccxml.readthedocs.io/en/develop/examples/nested-types/example.html

And a more specific example for function pointers, which is the case you are treating here:
http://pygccxml.readthedocs.io/en/develop/examples/function-pointer/example.html

Hope that helps. Please close this bug if this works for you.

[teto]

when I ask for "decl.arguments[0].decl_type", do you agree it should return void (*)( ) and not
void (*)( ) * ? This is a function pointer, not a pointer to a function pointer, IMO I shouldn't have to call declarations.remove_pointer() .

[iMichka]

Yes I guess the * is not wanted in that case. I am not using C++ often, so function pointer / pointer things are sometimes a little bit hairy for me.

I made a small change to the develop branch. I special-cased the function pointer decl_type creation, and the * is now gone.

The type is still pointer_t. One thing that may need to be done is to introduce a calldef_pointer_t class, extending pointer_t. Not sure this is needed anyway. It would require some work, for something that can still be checked with the is_calldef_pointer function.

Please tell me if that change is OK for you. I will apply that specific patch to the next 1.8.1 release then.

[iMichka]

I'm closing this bug, as I consider the problem as fixed for now. Do not hesitate to open a new one if this does not work for you.

[teto]

The commit is in develop branch: is that the branch that will become pygccxml1.9 ? or should I cherry pick the commit ?
I am waiting for the struct fix in castxml before relaunching my tests.

[iMichka]

At the moment yes, but 1.9.0 will probably not go out before next year. It took at least 1 year between 1.7.0 and 1.8.0.

I will just make a 1.8.1 release with that specific fix, probably next week. I am currently looking at the struct thing.