On Sun, Oct 21, 2018 at 03:45:20PM +0000, Glen Barber wrote: > On Sun, Oct 21, 2018 at 11:55:20AM +0200, Dimitry Andric wrote: > > On 21 Oct 2018, at 04:08, Jan Beich <jbe...@freebsd.org> wrote: > > > > > > Glen Barber <g...@freebsd.org> writes: > > > > > >> Modified: head/lib/clang/llvm.build.mk > > >> ============================================================================== > > >> --- head/lib/clang/llvm.build.mk Fri Oct 19 00:24:23 2018 > > >> (r339435) > > >> +++ head/lib/clang/llvm.build.mk Fri Oct 19 00:37:47 2018 > > >> (r339436) > > >> @@ -17,7 +17,7 @@ CFLAGS+= -I${LLVM_SRCS}/include > > >> CFLAGS+= -DLLVM_BUILD_GLOBAL_ISEL > > >> CFLAGS+= -D__STDC_LIMIT_MACROS > > >> CFLAGS+= -D__STDC_CONSTANT_MACROS > > >> -#CFLAGS+= -DNDEBUG > > >> +CFLAGS+= -DNDEBUG > > > > > > Would Clang still hit asserts instead of crashing with cryptic messages? > > > > In llvm, assertions (and various other internal consistency checks) are > > turned off by defining NDEBUG. So indeed, if something in the internal > > state goes in an unexpected direction, you could get crashes or other > > "interesting" effects. > > > > That said, the resulting executable(s) should run somewhat faster, and > > be somewhat smaller. I think we have always done this for stable > > branches and releases. > > > > I think I screwed up, and changed this in the wrong branch. I'm looking > into the "why", but the commit to head in this case should be reverted > (I'll take care of it if it was not already done), and this particular > change should be applied to stable/12 (which I will also take care of). > I first want to look through our docs to understand how I messed this > up. >
Ok, reverted in head as of r339541, and the commit to apply the change to stable/12 is sent to re@ for approval. Glen
signature.asc
Description: PGP signature