Path: utzoo!utgpu!news-server.csri.toronto.edu!mailrus!cs.utexas.edu!samsung!zaphod.mps.ohio-state.edu!lavaca.uh.edu!uhnix1!sugar!karl From: karl@sugar.hackercorp.com (Karl Lehenbauer) Newsgroups: comp.sys.amiga.tech Subject: Re: Manx 5.0 state? Summary: I'm using it exclusively Message-ID: <5677@sugar.hackercorp.com> Date: 9 May 90 15:17:20 GMT References: <15891@phoenix.Princeton.EDU> Reply-To: karl@sugar.hackercorp.com (Karl Lehenbauer) Organization: Count Floyd's 3-D House of Unix Lines: 26 A few words in the defense of Manx 5.0. Sure, it has some problems, but by making heavy use of function prototypes (and it has a command-line option to read source and generate prototypes), I believe I've cut my coding errors by about two-thirds. That is to say that with prototypes, the compiler is catching 2/3rds of the bugs that I used to have to debug at runtime. Further, you no longer have to be super-careful to cast all your ints to longs when calling Amiga library routines from the 16-bit int model, because the compiler does it for you, again thanks to prototypes, and eliminating a major source of gurus during the development process. Also, the code it generates is significantly better if you compile with the -so option, and it has this #pragma definition for library functions where it generates the code to call the library routines directly rather than calling glue routines that pop arguments off the stack and move them into registers, then calling the library routine from the glue routine, improving performance. 32-bit integers are the default now, which'll make some programs slower unless you compile with the right switches to get 16-bit ints. (I'm pretty much of a fan of 32-bit ints on 32-bit machines anyway and yes, in my opinion, the 68000 is a 32-bit machine -- certainly as much as the 386SX is) All in all, I would not consider going back to 3.6 for a second. -- -- uunet!sugar!karl -- Usenet access: (713) 438-5018