Path: utzoo!utgpu!jarvis.csri.toronto.edu!mailrus!tut.cis.ohio-state.edu!zaphod.mps.ohio-state.edu!swrinde!ucsd!ucbvax!ACF8.NYU.EDU!mckenney
From: mckenney@ACF8.NYU.EDU (Alan Michael McKenney)
Newsgroups: comp.lang.fortran
Subject: Re: Bug cured by debugging output.
Message-ID: <9001111806.AA01823@acf8.NYU.EDU>
Date: 11 Jan 90 18:06:21 GMT
Sender: daemon@ucbvax.BERKELEY.EDU
Lines: 59

Subject: Re: Bug cured by debugging output.
Newsgroups: comp.lang.fortran
In-Reply-To: article <711@dbrmelb.dbrhi.oz> of 9 Jan 90 22:29 EST

In article <711@dbrmelb.dbrhi.oz>, davidp@dbrmelb.dbrhi.oz (David Paterson)
writes:

# I've been programming in Fortran 77 for ten years now and have seen
# all sorts of things but this is one of the strangest.
# 
# The program below works perfectly ....
...
# 
# 
#       CHARACTER*80 LINE
#       DATA XLO,XHI,YLO,YHI/0.,10.,0.,10./
#  49   WRITE(*,50) XLO,XHI,YLO,YHI
#  50   FORMAT(' Type XLO,XHI,YLO,YHI <',3(G9.3,','),G9.3,'>: ')
#       READ(*,'(A80)') LINE
#       WRITE(*,*) 
#       READ(LINE,*,END=49,ERR=49) XLO,XHI,YLO,YHI
#       STOP
#       END
# 
# However, if the line 'WRITE(*,*)' is removed then the program
# fails miserably (endless loop). Why?
# 
# -------------------------------------------------------------------
# David Paterson
# CSIRO Division of Building, Construction and Engineering
# Highett, Victoria, AUSTRALIA

    The legalistic answer to why your program doesn't work: it violates
the FORTRAN-77 standard: Par 12.2.5.2 says

    Internal File Restrictions. ...

    (1) Reading and writing records is accomplished only by
        sequential access formateed input/output statements
	(12.8.1) that do not specify list-directed formatting.

I can't give a good practical justification for this; the Unix
compilers we use allow it, but my (limited) experience with
microcomputer compilers suggests that you are lucky to have it even run
standard-conforming programs correctly.

    Since it isn't legal FORTRAN 77, it isn't a bug for a compiler/
run-time library to do strange things with it.  Unless the manual for
Microsoft FORTRAN says it allows this as an extension, you are out of
luck.  A user-friendly compiler would flag this usage as an error, or
try to make it behave in some intuitively obvious way.  I don't know if
the END or ERR exit is Microsoft's way of flagging the illegal code, or
if it is just an unplanned result of feeding the I/O library arguments
it doesn't expect; I suspect the latter, since the "WRITE(*,*)" has an
effect on it.


Alan McKenney        E-mail:  mckenney@robalo.nyu.edu          (INTERNET)
Courant Institute,NYU         ...!cmcl2!robalo!mckenney        (UUCP)