a few thoughts. I'm not in love with changing CFLAGS if hostname is raccoon. i would make a choice to either include no-format or not.
I would not run the regress tests as part of the main build but would disconnect it. this is how most ports work as far as I know. while my ulimit patch will let the regress tests work, I think a note might be desirable in pkg-readme. otherwise someone with a low ulimit -n will have to hunt for the same problem when they try to run. ok daniel@ if these changes are made. > On Nov 20, 2014, at 8:14 AM, Jonathan Gray <j...@jsg.id.au> wrote: > >> On Thu, Nov 20, 2014 at 11:44:08PM +1100, Jonathan Gray wrote: >>> On Wed, Nov 19, 2014 at 02:08:32PM +1100, Jonathan Gray wrote: >>> Here is a quick port of lcamtuf/Michal Zalewski's instrumented fuzzer >>> 'American fuzzy lop'. Only tested on amd64 where it requires the binutils >>> change I just committed to allow sahf/lahf instructions. >>> >>> http://lcamtuf.coredump.cx/afl/ for more details >> >> Updated port attached for version 0.60b that includes >> various changes made by Michal Zalewski upstream for OpenBSD. >> In particular afl can now handle instrumenting OpenBSD binaries >> without having to disable pie. >> >> Also adds a change to the Makefile to raise the fd ulimit to >> ensure the regress test passes from Daniel Dickman. > > And here is another version of the port as sthen@ points > out the distfile was rerolled. Apparently for a workaround > for lahf / sahf on older releases of OpenBSD/amd64 before > http://marc.info/?l=openbsd-cvs&m=141636589924400 > <afl.tgz>