🚀 go-pugleaf

>>>>> On 28 Jun 91 23:13:12 GMT, jbs@WATSON.IBM.COM said:

jbs>          The DO, ENDDO construct is not in the Fortran 77 standard.
jbs> Nevertheless many Fortran 77 compilers recognize it and many users use
jbs> it.  I would not advise omitting it from your compiler in order to
jbs> "expose" "broken" user code.

The word "broken" is clearly being used in this context to refer to
code whose behavior is undetermined under the standard --- therefore
*logically* "broken".  The examples which have been the topic of
discussion here depend on either subtle issues of floating-point code
re-arrangement or (in the other thread) on a particular
implementation's treatment of functions with side effects.

The use of extensions is an entirely different issue.  The resulting
codes are *syntactically* incorrect FORTRAN-77 --- their *logical*
interpretation under a *superset* of the FORTRAN language is
independent of the fact that extensions are used.

I certainly appreciate any efforts that a compiler makes to tell me my
code is either broken or non-portable.  Most major vendors provide
compiler switches to flag ANSI non-compliance.  Very few provide
robust and comprehensive extensions to detect common *logical* errors,
such as:
	-- use of uninitialized variables
	-- argument mismatch in calling sequences
	-- use of undeclared variables (mispellings)
	-- out-of-bounds array references in called routines
		(often a result of an argument mismatch)
--
John D. McCalpin			mccalpin@perelandra.cms.udel.edu
Assistant Professor			mccalpin@brahms.udel.edu
College of Marine Studies, U. Del.	J.MCCALPIN/OMNET