Hi Jason, On Fri, Aug 23, 2019 at 11:24:27AM -0300, Jason Gunthorpe wrote: > On Thu, Jul 11, 2019 at 10:30:30AM -0700, Nathan Chancellor wrote: > > On Thu, Jul 11, 2019 at 02:18:08PM -0300, Jason Gunthorpe wrote: > > > On Thu, Jul 11, 2019 at 10:16:44AM -0700, Nick Desaulniers wrote: > > > > On Thu, Jul 11, 2019 at 6:39 AM Jason Gunthorpe <jgg@xxxxxxxx> wrote: > > > > > > > > > > On Thu, Jul 11, 2019 at 01:14:34AM -0700, Nathan Chancellor wrote: > > > > > > Maybe time to start plumbing Clang into your test flow until it can get > > > > > > intergrated with more CI setups? :) It can catch some pretty dodgy > > > > > > behavior that GCC doesn't: > > > > > > > > > > I keep asking how to use clang to build the kernel and last I was told > > > > > it still wasn't ready.. > > > > > > > > > > Is it ready now? Is there some flow that will compile with clang > > > > > warning free, on any arch? (at least the portion of the kernel I check) > > > > > > > > $ make CC=clang ... > > > > > > > > Let us know if you find something we haven't already. > > > > https://clangbuiltlinux.github.io/ > > > > https://github.com/ClangBuiltLinux/linux/issues > > > > > > What clang version? > > > > > > Jason > > > > You'll need clang-9 for x86 because of the asm-goto requirement (or a > > selective set of reverts for clang-8) but everything else should be > > good with clang-8: > > The latest clang-9 packages from apt.llvm.org do seem to build the > kernel, I get one puzzling warning under RDMA: > > drivers/infiniband/hw/hfi1/platform.o: warning: objtool: tune_serdes()+0x1f4: can't find jump dest instruction at .text+0x118a Any particular config that I should use to easily reproduce this? > And a BPF one: > > kernel/bpf/core.o: warning: objtool: ___bpf_prog_run()+0xd: sibling call from callable instruction with modified stack frame I think this might be related to this? https://lore.kernel.org/lkml/cf0273fb-c272-72be-50f9-b25bb7c7f183@xxxxxxxxxxxxx/ Cheers, Nathan
