2008/6/5 Erik Trimble <[EMAIL PROTECTED]>: > Andrew John Hughes wrote: >> >> 2008/6/4 Kelly O'Hair <[EMAIL PROTECTED]>: >> >>> >>> Not sure what you mean by the Sun Studio trap. >>> >>> >> >> I'm referring back to the Java trap - before Sun released their JDK >> under the GPL, >> it was possible to have applications under a Free Software license >> written in Java >> which couldn't be run in a Free environment because they would only work >> under >> the proprietary JDK. GNU Classpath was the community's attempt to free >> Java >> from this trap. >> >> Only being able to build the OpenJDK using the proprietary Sun Studio >> compiler on Solaris creates >> a similar issue, though the scope of the problem is fortunately more >> restricted. I'm not sure OpenSolaris >> itself can even be built with GCC, which is an even worse issue - it's >> not truly Free Software if it can only >> be built with proprietary tools. >> >> > > This is NOT a trap. This is a CHOICE OF PREFERENCE made by the FS > Community. It is no more a trap than using GPL'd programs in conjunction > with the Apache http server is. Just because we don't play exactly in your > world doesn't mean there is a trap. Traps are when you cannot replace a > whole component layer easily with something else. So, if your GPL'd Java > program had to run on a Sun JDK under Solaris, that would certainly be true. > However, you can run your Java program on one of about 4 major JDKs now, > under over a dozen OSes, so that hardly qualifies as a trap. > > Remember, not everyone wants everything in one particular format (or, > license). We've got end-users who can't stand the GPL, for some _very_ good > reasons. We've got other folks who are sore at us for not using an MPL-style > license. So, I take license-criticism extremely poorly. >
Andrew Haley already made the point very succinctly. I wasn't discussing specific licenses or Java applications, but the specific issue of building the OpenJDK on OpenSolaris. It's a different and more specific trap, but one nonetheless. On the bright side, I think fixing it is much more tenable. >>> Each release of a compiler requires some kind of work to the Makefiles, >>> happened with gcc4, and will happen with SS12 and VS2008. >>> >> >> While I can understand some changes being necessary for major releases >> (e.g. the move from GCC 3 to 4), >> every single release shouldn't need work; this suggest an issue with >> the build system itself. >> > > Sadly, no, this is an issue with the COMPILERS, not the make system. GCC, > SS, and all other compilers has a nasty tendency to subtly break all sorts > of things without rev-ing the major version number. In GCCs case, this is > often directly related to changes in GLIBC, particularly on Linux. GCC has > some bad (recent) history, as major changes were implemented without obvious > notice, with no major version bump, and occasionally with no minor version > bump either. gcc 3.2 is considerably different than 3.3 or 3.4. And, in > Microsoft's case, their myriad of different compiler 'distros' within the > same general release (i.e. Visual Studio vs Express vs VS Professional) is > even worse, as they support very different library sets and compiler flags. > So, every time we want to support a new compiler, there's some work to be > done to discover these differences, and adjust the makefiles to compensate. > > It would certainly be nice for the JDK to be able to build with more than > one compiler on any given platform. But that's what a community is for - > people interested in using a specific compiler should certainly not be > prevented from doing so. Our (Sun's) interest is primarily in supporting > our own compilers. > > That said, the make system could use some serious streamlining, and the code > itself could _really_ do a much better job of disentangling itself, so that > it could be built in a more modular form. > Exactly; the reason for highlighting such issues openly on a mailing list is so problems can be worked on by the community. I don't think it's Sun's job alone to go and fix it, and likewise I would see Sun as part of the community not a distinct entity from it. >>> Andrew John Hughes wrote: >>> >>>>> >>>>> GCC will NOT work under Solaris/SPARC, and I'm pretty darned sure it >>>>> won't >>>>> work under Solaris/x86 or Solaris/x64. There are some significant >>>>> GCC-isms >>>>> which the JDK does not currently support. >>>>> >>>>> That said, it would not be terribly difficult to modify the source to >>>>> get >>>>> GCC to work, but you'd definitely have to spend a bit of time doing so. >>>>> >>>> >>>> Maybe the next logical campaign is to avoid the Sun Studio trap then... >>>> :) >>>> > > > -- > Erik Trimble > Java System Support > Mailstop: usca22-123 > Phone: x17195 > Santa Clara, CA > Timezone: US/Pacific (GMT-0800) > > -- Andrew :-) Support Free Java! Contribute to GNU Classpath and the OpenJDK http://www.gnu.org/software/classpath http://openjdk.java.net PGP Key: 94EFD9D8 (http://subkeys.pgp.net) Fingerprint: F8EF F1EA 401E 2E60 15FA 7927 142C 2591 94EF D9D8
