Josh Elsasser writes: > On Fri, Apr 27, 2018 at 08:56:04PM -0700, Josh Elsasser wrote: >> Patching the i386 and ppc *-arch.c files isn't necessary. However i386 >> didn't build for me (in a VM). I think we need to do this in >> patches/patch-src_runtime_bsd-os_c: > > This is obviously wrong now that I look closer. I'll investigate further. > >> $OpenBSD$ >> >> MAP_TRYFIXED is needed to allow SBCL to relocate its heap if the >> hardcoded virtual address is unavailable. >> >> Index: src/runtime/bsd-os.c >> --- src/runtime/bsd-os.c.orig >> +++ src/runtime/bsd-os.c >> @@ -157,7 +157,11 @@ os_validate(int movable, os_vm_address_t addr, os_vm_s >> * the hint address, and moreover that it "is the default >> behavior") */ >> // FALLTHROUGH_INTENDED >> case NOT_MOVABLE: >> +#ifdef MAP_TRYFIXED >> + flags = MAP_TRYFIXED; >> +#else >> flags = MAP_FIXED; >> +#endif >> } >> #ifdef MAP_EXCL // not defined in OpenBSD, NetBSD, DragonFlyBSD >> if (flags & MAP_FIXED) flags |= MAP_EXCL; >>
Christian Weisgerber reported me that sbcl build was failing on i386 with 1.4.6, build log attached to the mail. Maybe sbcl is allocating too much memory, I don't know... I don't have an i386 system at the moment to debug this. Otherwise, next week I will be able to try sbcl on powerpc arch as I am using it daily.
sbcl.log.gz
Description: sbcl build log
