Deallocation of arrays passed as c_ptr

[PierU]

@mEm

Thanks for the reminder of “carefully read the standard!”

I read, and missed the text (I was not looking at the right place)!

[mEm]

I thought the discussion was on the Fortran side only

[ivanpribec]

My misunderstanding in that case. Within Fortran I would avoid such pointer casting tricks anyways.

[everythingfunctional]

@mEm

But, so why is c_loc not made to throw even a warning when a pointer to non-contiguous memory is passed as argument?

That’s normative text, not a numbered constraint. Thus, a processor is not required to diagnose it. In fact, no runtime errors are really “required” to be diagnosed. That said, an implementation would be allowed to, but that would complicate what is probably a very simple implementation.