Re: g77 3.4.(2|3): Failure when calling gcc 3.4.(2|3) functions

[Tim Prince <tprince at computer dot org>]

At 03:46 PM 11/22/2004, Samuel Ziegler wrote:

> The semantic change seems to deal with the short-circuiting of if
> statements.  The G77 3.4.x behavior appears to be correct.  According to
> this document:
> http://www.ibiblio.org/pub/languages/fortran/ch1-8.html
>
> I poked through the release notes, but didn't see anything about this
> change.  Does this describe a change which occurred?
>
> Here is a test program from my coworker:
>
> g77  -fno-underscoring -g -fugly-logint fort2.f -o testcase
>
> ------fort2.f-------------------------
>
>       integer function child()
>         write (*, *) 'Gahh'
>         child = 1
>       end
>
>       subroutine parent
>       integer a
>       integer child
>
>       a = 0
>
> c     Is the second expression evaluated ?
> c     G77 3.3 - Says no
> c     G77 3.4 - Says yes
>       if(a.eq.1.and.child().eq.1) then
> c     do nothing
>       endif
>
>       return
>       end
>
>
>       program testcase
>       call parent
>       end
>
> -------- end fort2.f ---------------------

As your example shows, if you depend on undefined behavior, a change in the 
back end could change your results, even though no change was likely to 
have been made specifically to g77.


Tim Prince