Path: utzoo!attcan!uunet!clyde.concordia.ca!mcgill-vision!bloom-beacon!snorkelwacker!usc!brutus.cs.uiuc.edu!zaphod.mps.ohio-state.edu!swrinde!ucsd!ucbvax!RAND.ORG!salzman From: salzman@RAND.ORG (Isaac Salzman) Newsgroups: comp.soft-sys.andrew Subject: Re: .amsalien Message-ID: <9003051818.AA02465@gaucho.rand.org> Date: 5 Mar 90 18:18:34 GMT References: Sender: usenet@ucbvax.BERKELEY.EDU Reply-To: salzman@rand.org Organization: The Internet Lines: 25 >Nathaniel Borenstein writes: >No, that's a Sun compiler bug, not an Andrew bug. But I think that at >least -O is no longer the default! Well yeah, that makes sense. I recall detecting a Sun compiler bug with the beta version in the whitepages code. I don't know if I'd *totally* blame it on Sun's compiler (though there's no excuse for different behavior with or w/o -O). These bugs tend to crop up in some "not so kosher" (and not so portable) code (e.g. passing structs or unions and expecting pointers - that sort of thing), where assumptions are made about what the compiler is generating. I don't think I've seen any mention of this on the list. Has anyone tried building Andrew on a SPARC machine using GCC (GNU C Compiler, where using -O tends not to break your code)? Thanks! -- * Isaac J. Salzman ---- * The RAND Corporation - Information Sciences Dept. /o o/ / * 1700 Main St., PO Box 2138, Santa Monica, CA 90406-2138 | v | | * AT&T : +1 213-393-0411 x6421 or x7923 (ISL lab) _| |_/ * Internet : salzman@rand.org / | | * UUCP : !uunet!rand.org!salzman | | |