On Thu, Jul 06, 2017 at 07:28:21PM +0200, Antoine Jacoutot wrote: > On Thu, Jul 06, 2017 at 07:55:04AM -0700, Josh Elsasser wrote: > > On Thu, Jul 06, 2017 at 04:17:54PM +0200, Marc Espie wrote: > > > Regularly, my bulks barf on lang/sbcl,threads > > > > > > > > > This is a problem because then the lock remains, which I have to manually > > > remove... depending on build order (somewhat random) lang/sbcl may have > > > built already, or not. > > > > > > There are about 3 choices here: > > > > > > - fix lang/sbcl,threads > > > - disable the self-test which is failing randomly > > > - remove lang/sbcl,threads from the list of build ports. > > > > > > This failure has been around since the threads FLAVOR was introduced. > > > > > > At some point, it's got to get some kind of resolution. > > > > I think the best thing to do here is to not build -threads by > > default. I think the small post-build test suite should be kept, as > > sbcl sticks its nose into all sorts of private interfaces which have > > been changed in the past and the bulk builds may very be the only time > > the small test suite is run on openbsd. > > > > I'd be happy to work with someone who wants to start digging into sbcl > > to fix the sb-concurrency test failures on amd64 or implement -threads > > on other architectures, or fix sbcl with randomized mmap(), PIE, W^X, > > and whatever else is coming in the future. > > > > ok MAINTAINER to remove SUBDIR += sbcl,threads from lang/Makefile > > I'd like an XXX in the port Makefile then. > Because not building a FLAVOR is not sth we want to do without a very good > reason.
Yeah, it's always a good idea, obviously. :)