Path: utzoo!news-server.csri.toronto.edu!cs.utexas.edu!swrinde!zaphod.mps.ohio-state.edu!think.com!hsdndev!cmcl2!lanl!jlg From: jlg@lanl.gov (Jim Giles) Newsgroups: comp.lang.fortran Subject: Re: LABEL data type (Was: integer variable format address) Message-ID: <17077@lanl.gov> Date: 8 Mar 91 16:46:21 GMT References: <31132@shamash.cdc.com> <214949.21497@timbuk.cray.com> Organization: Los Alamos Natl Lab, Los Alamos, N.M. Lines: 18 From article <214949.21497@timbuk.cray.com>, by wws@raphael.cray.com (Walter Spector): > [...] Maybe what's needed is a pair of > setjmp/longjmp intrinsic routines. No. That leads to as many problems as it solves. What's really needed is a well designed exception handling mechanism. Even alternate return is a better solution than setjmp/longjmp (at least alternate return maintains proper call chain protocols; setjmp/longjmp are _supposed_ to be properly nested, but nothing in either the syntax or semantics makes this constraint easy to apply - and many implementations _don't_ apply it). Frankly, I think that the trend to avoid labels and jumps altogether is a better way to go. I don't think that any language is really complete without a GOTO, but as time goes on it may be less and less of an issue. J. Giles