In message: <4C129ECE.8040709_at_FreeBSD.org> Doug Barton <dougb_at_FreeBSD.org> writes: : On 06/11/10 13:35, Ed Schouten wrote: : > * M. Warner Losh<imp_at_bsdimp.com> wrote: : >> Except that clang isn't quite disabled when cross-building, due to the : >> issue I pointed out when the commit went in wrt bsd.own.mk. : >> MACHINE_ARCH is still amd64 until we start to build the sparc64 : >> binaries, so anything in the bootstrapping part of the build will not : >> be disabled. that's one of many reasons disabling things in : >> bsd.own.mk based on architecture is going to fail. : > : > Exactly. I already mentioned this problem on arch_at_ back in May. That's : > why a small portion of LLVM, namely tblgen, is always built. Even : > though : > this will slightly increase the build time, it shouldn't be excessive. : > If someone knows of a nice way to fix this, I am more than willing to : > get it fixed. Ed: Sorry I missed it back in may on arch_at_. I'm usually pretty good about catching things like that... : What is "this" that needs to be fixed? Or, asked another way, why are : we not optimizing for the common case (building world on the machine : that it will be run on)? "This" is building the proper set of tools for the target. It is easy to do, and only a couple lines of Makefile foo in Makefile.inc1 instead of in bsd.own.mk. It is a fairly natural consequence of the tbemd stuff I have been working on and have started merging. The consequences today are that you build some extra tools that are only needed to build clang when in fact you aren't really going to be building clang. The "cost" is however long it takes to do this on the platform you are building on. This can range from a minute or two to tens of minutes depending on the power of your build system. WarnerReceived on Fri Jun 11 2010 - 19:31:33 UTC
This archive was generated by hypermail 2.4.0 : Wed May 19 2021 - 11:40:04 UTC