[J3] Another thought on paper 24-142r2 (ASYNCHRONOUS issue)

[Daniel C Chen]

Thanks for all the explanations from everyone!

If I understand correctly, the rational of disallowing an <expr> actual argument (effective actual argument would be a temp) or in this case an array section with vector subscripts (effective actual argument would also be a temp) corresponding to a dummy argument with the ASYNCHRONOUS attribute is to prevent the temp being destroyed after the subroutine returns while the asynchronous I/O is not done yet.

By the same reason, would the following code have the same issue and be invalid?

"
     program main
       interface
         subroutine asynch_write()
         end subroutine
       end interface
       call asynch_write()
     end

     subroutine asynch_write()
       integer :: local_1(3), local_2(3)
       OPEN(ASYNCHRONOUS='yes', UNIT=4, FORM='unformatted')
       WRITE(ASYNCHRONOUS='yes', UNIT=4) local_1 + local_2
     end subroutine
"

Please note that I intentionally made the <output-item> an <expr> as it is what it is in the standard (i.e. not a variable, thus not definable).

"R1217 output-item is <expr>
                or io-implied-do"

Without a WAIT statement at the end of the subroutine, the <output-item> expression could also be destroyed before the asynchronous I/O finishes.

I can't see the difference between these two cases. Am I missing something simple here?

Any comments are highly appreciated.

Thanks,
Daniel
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://mailman.j3-fortran.org/pipermail/j3/attachments/20240626/bb8e22c4/attachment.htm>