On Mar 17, 2008, at 4:05 PM, Aleksandar Donev wrote:
On Monday 17 March 2008 14:40, Jeff Squyres wrote:
Craig mentioned last week that if we have these opaque Fortran types (with their contents unspecified), that we could be guaranteed that their first member would be at the first byte of the instance.
Sorry, I don't recall saying that. It's as Aleks says below, the guarantee is that it will conform to the equivalent C struct for the companion C compiler. Is struct foo { int i; } always equivalent to an int in C? I think the C calling convention treats them differently enough so that defining an MPI_HANDLE as a user defined type (essentially a C struct) with one integer component won't be guaranteed to work. In that case, we need the typealias Aleks so dearly wants :-) That was before my time on the committee so I don't have a horse in that race. Cheers, Craig