From: khan AT xraylith DOT wisc DOT edu (Mumit Khan) Subject: Re: cp/mv/rm problems with .exe extension filenames 27 Sep 1998 21:39:28 -0700 Message-ID: References: <19980928003353 DOT 20720 AT mundook DOT cs DOT mu DOT OZ DOT AU> Mime-Version: 1.0 Content-Type: TEXT/PLAIN; charset=US-ASCII To: Fergus Henderson Cc: "Parker, Ron" , "'gnu-win32 AT cygnus DOT com'" On Mon, 28 Sep 1998, Fergus Henderson wrote: > On 25-Sep-1998, Mumit Khan wrote: > > What is the general consensus, if any, here? Here're the choices: > > > > 1. add .exe never. > > 2. add .exe always, unless .exe is already there. > > 3. add .exe only if no suffix is provided. > > My vote is for #1. That option allows Unix software to build and > install without modifications, which is after all the whole idea behind > gnu-win32. Options 2 and 3 require modifications to Makefiles. Mine was too until recently when others using our software started complaining that ShellExecute (a very useful WIN32 API routine) doesn't seem to work when no extension is provided. Anders Norlander also points this out in a recent message (related thread). I guess I'm somewhat torn between #1 and #3; all our config tools and makefiles already can switch automatically (this was to facilitate cross-builds, which I use almost exclusively), so I can live with either one. Now I use --with-executable-suffix=SUFFIX option to configure to pick a suffix, which defaults to .exe for ix86-*win32* targets. > Of course, for mingw32 the answer should be different, because mingw32 has > different aims. But if I type `gcc' at a bash prompt, it should give > me the same results that it does under Unix. For mingw32, #3 IMO is the only reasonable option at the present time. When NT takes over and "command.com" on W9x dies, the Unix behaviour will make sense Regards, Mumit - For help on using this list (especially unsubscribing), send a message to "gnu-win32-request AT cygnus DOT com" with one line of text: "help".