If I recall correctly it has to do with LLVM’s new license—Apache 2.0. OpenBSD isn’t okay with this, I believe. According to Wikipedia:
The OpenBSD project does not consider the Apache License 2.0 to be an acceptable free license because of its patent provisions. The OpenBSD policy argues that when the license forces one to give up a legal right that one otherwise has, that license is no longer free.
OpenBSD 6.6 made lots of progress on the replacement of GCC4 with Clang. Somewhat hilarious / sad that the replacement has a similar license hurdle that keeps it, too, pinned at a max version. I do not think the project has the resources to fork and maintain LibreLLVM… :)
But for the base system, they are probably only interested in maintaining enough to run clang (and not clang++). Since there are probably no huge performance leaps for UNIX-type of work in newer compilers, I can imagine that just maintaining (as in do bugfixes, fix CVEs, adding OpenBSD-specific extensions) the last LLVM/clang with a favorable license is good enough for them for a while.
For third-party software, people can always install a newer LLVM/clang from ports.
Based on this argument, why leave GCC4 behind? Why shouldn’t base be able to rely on new sanitizers, and innovations moving forward? Or bug fixes? Or new optimizations?
I get that this is a philosophical choice and that they could change their stance on Apache 2.0 if they wanted. But, their objection is valid, and I wish other projects would follow suit and consider publicly speaking out against shitty “open source” licenses that restrict freedoms, instead of blindly accepting them, and allowing for the proliferation of subtly nonfree licensing. See also the new Open Core related licenses that keep popping up.
you need clang++ to compile clang (just like you need g++ these days for gcc). “Just maintaining” will work a few years until new architectures (or architecture extensions) come along that need to be integrated.
Very cool artwork on this one! Here’s to another great release!
Any particular reason for not using version 10?
If I recall correctly it has to do with LLVM’s new license—Apache 2.0. OpenBSD isn’t okay with this, I believe. According to Wikipedia:
OpenBSD 6.6 made lots of progress on the replacement of GCC4 with Clang. Somewhat hilarious / sad that the replacement has a similar license hurdle that keeps it, too, pinned at a max version. I do not think the project has the resources to fork and maintain LibreLLVM… :)
But for the base system, they are probably only interested in maintaining enough to run clang (and not clang++). Since there are probably no huge performance leaps for UNIX-type of work in newer compilers, I can imagine that just maintaining (as in do bugfixes, fix CVEs, adding OpenBSD-specific extensions) the last LLVM/clang with a favorable license is good enough for them for a while.
For third-party software, people can always install a newer LLVM/clang from ports.
Based on this argument, why leave GCC4 behind? Why shouldn’t base be able to rely on new sanitizers, and innovations moving forward? Or bug fixes? Or new optimizations?
I get that this is a philosophical choice and that they could change their stance on Apache 2.0 if they wanted. But, their objection is valid, and I wish other projects would follow suit and consider publicly speaking out against shitty “open source” licenses that restrict freedoms, instead of blindly accepting them, and allowing for the proliferation of subtly nonfree licensing. See also the new Open Core related licenses that keep popping up.
you need clang++ to compile clang (just like you need g++ these days for gcc). “Just maintaining” will work a few years until new architectures (or architecture extensions) come along that need to be integrated.