On 7/26/07, Alan W. Irwin <[EMAIL PROTECTED]> wrote: > On 2007-07-26 18:35-0400 Brandon Van Every wrote: > > > On 7/26/07, Alan W. Irwin <[EMAIL PROTECTED]> wrote: > >> On 2007-07-26 13:55-0400 Brandon Van Every wrote: > >> > >> > I think it is very important that any experimental releases have no > >> > effect on official CMake installations at all. > >> > >> That's already been stated. I say again, you will not be affected by this > >> proposal at all. > > > > That's not the point. The point is to define the exact mechanism by > > which this is guaranteed. Just having people download "official" > > CMakes or "experimental" CMakes at their whim won't guarantee it. > > People will get confused about which CMake they're using. If the > > world starts getting populated by CMakes of uncertain composition, I > > am indeed affected. > > Hmm. That (keeping users from being confused about which version they have) > is what release numbers are for. Thus, I am not at all convinced by your > argument.
I don't want a non-standard CMake blowing up my build, unless the user is automatically banged over the head with a wooden mallet that they're using a non-standard CMake. People put tools in paths and forget about them. They have no idea what they're using. I don't want to write a bunch of extra query logic to determine what they are using. And I know in the real world, most CMakeLists.txt authors won't write that logic. I've seen the endgames of DirectX capability queries and they ain't pretty. If some builds gotta have experimental module stuff, and other builds gotta *not* have experimental module stuff, the user's got a problem. Is he gonna maintain 2 different CMakes all the time? That's dumb, and he'll screw it up too. I'm a buildmaster, I have 5 different Windows compiler environments hermetically sealed off from each other, which I could permute with CMake stable and CMake CVS if I wanted to. Most people aren't buildmasters, they just want to run CMake and have their software. The disambiguation mechanism needs to function automatically for all versions of CMake produced. An "experimental" version should automatically stick out like a sore thumb. > I think the fundamental reason why you are attempting to be a change blocker > on my suggestion to make module releases is you don't feel comfortable with > the free software culture which encourages releases. Stability is one issue. Lack of faith in volunteers actually doing the work is another. > I do feel extremely > comfortable with that culture (I have been living it for the last decade), > and that is why I am encouraging the CMake developers to start making > module releases. Well then sell Kitware on it. So far it doesn't sound like you take the separation of experimental stuff from official stuff very seriously. I mean geez, entire programming languages like C# have been invented around this kind of "DLL Hell" problem. Cheers, Brandon Van Every _______________________________________________ CMake mailing list [email protected] http://www.cmake.org/mailman/listinfo/cmake
