hirchert@uxe.cso.uiuc.edu (03/18/89)
Keith Bierman and Richard Edwin Maine have been engaged in a somewhat speculative discussion about what conformance checking is required by Fortran 8x. Here's the straight scoop: FORTRAN 77 (the current standard) consists almost entirely of a description of standard-conforming programs and what they mean. Standard-conforming processors are addressed primarily by the requirement that they correctly execute standard-conforming programs, with a few secondard requirements stemming from places where the behavior of a standard-conforming program is only partially defined. Shortly after FORTRAN 77 was made an ANSI standard, it was also made a U.S. government purchasing standard (a FIPS). The FIPS additionally required a processor to be able to identify nonstandard usage. (The FIPS is the reason the FORTRAN compilers from most major vendors have some means of checking for standard conformance.) Throughout the Fortran 8x development there has strong pressure to place more requirements on a standard-conforming processor. X3J3 has resisted adding any restrictions that would prohibit extensions or require run-time tests, but has been open to imposing requirements roughly equivalent to that of the FIPS (since a revised FIPS would probably have such requirements anyway). The text Keith and Richard have been discussing is X3J3's attempt to incorporate the FIPS requirement into the standard itself. In other words, processors conforming to Fortran 8x will have to have an option to identify nonstandard usage in Fortran programs. Kurt W. Hirchert hirchert@ncsa.uiuc.edu National Center for Supercomputing Applications
khb@fatcity.Sun.COM (fatcity) (03/21/89)
In article <50500106@uxe.cso.uiuc.edu> hirchert@uxe.cso.uiuc.edu writes: > > >In other words, processors conforming to Fortran 8x will have to have an >option to identify nonstandard usage in Fortran programs. > >Kurt W. Hirchert hirchert@ncsa.uiuc.edu >National Center for Supercomputing Applications Thank you for the clarification. Keith H. Bierman It's Not My Fault ---- I Voted for Bill & Opus
khb@fatcity.Sun.COM (fatcity) (03/21/89)
In case it was not clear from my last posting... I was wrong. The new standard will require vendors to produce warnings about non-compliance. But vendors are free to generate code which does what they want, in the event that the user code is non-standard complying. Thanks to Kurt H. and Gary C. who showed me the error of my ways. Cheers all. khb Keith H. Bierman It's Not My Fault ---- I Voted for Bill & Opus