But would modularising GCC basically mean recreating LLVM's design? I'd be curious to see why the GCC developers would think that route to be better than just helping improve LLVM?
And GCC won't accept BSD-like licensed patches also (the FSF prefer the authors to rennouce to their copyright; and the FSF naturally have that newly assigned code licensed under the GPLv3) → http://gcc.gnu.org/contribute.html#legal // http://gcc.gnu.org/gccmission.htmlMS Visual Studio will not accept GPL patches. Neither will LLVM.
¿Your point? I don't see GCC dying ... but it could definitely be thrown out his golden throne if it doesn't adapt to newer expectations when it comes to compilers (and that could require not only a tech work, but also a legal work of at least a partial re-licensing to something like the LGPL)
I do hope that GCC continues to improve and be a healthy compiler project. But because it's good for the project, not because of trying to catch some imaginary lead. If closed-source developers prefer cannibalising LLVM to fit into their products, then they shold continue doing so. Relicensing GCC for that purpose would be suicide, IMHO.
+1True, but I don't see it as a catastrophe
Yes, nobody should really be happy about this ... having competition and compiler diversity is actually something *everyone* benefits from; and hearing the GCC people talking about being fit for being able to compete is a good example of that.GCC was never meant to sit on a golden throne. If we have two top-notch compilers, one GPL and one BSD, that would be amazing for everybody, including the FSF.
Yes and no ... projects like KDevelop, Qt Creator, vim (there's already works on this, and a plugin which invokes clang for completion) and Code::Blocks could nourish from the fact that LLVM/Clang is modular and allows to use their highly advances parser, and that enables them to make their codebase more easier to mantain and more fit (by not entering the swamp of code parsing/analysis), and that could give space to a next-step evolution of IDEs of having really complex parsing (and features building on top of that) done in real time and a paired support of features with the compiler (and not having things like today's where your IDEs code parsing is not supporting something your compiler does, awfully true with C++11) : Without worrying about code tainting the original project decision with the license.If closed-source developers prefer cannibalising LLVM to fit into their products, then they shold continue doing so.
So I find your argument a little bit misleading and visceral.
I was talking about a partial re-license which would make possible to offer the *use* this facility without being viral with the license.Relicensing GCC for that purpose would be suicide, IMHO.
That's one of the newer technical expectations if GCC wants to get an ecosystem building around like LLVM/Clang is having.