http://gcc.gnu.org/bugzilla/show_bug.cgi?id=50386
--- Comment #5 from dave.anglin at bell dot net 2011-09-23 16:04:18 UTC --- On 9/20/2011 6:32 PM, ramana at gcc dot gnu.org wrote: > http://gcc.gnu.org/bugzilla/show_bug.cgi?id=50386 > > --- Comment #4 from Ramana Radhakrishnan<ramana at gcc dot gnu.org> > 2011-09-20 22:32:49 UTC --- > I think this is a bootstrap issue rather than a problem with libgomp . A > reduced testcase of this form fails with the stage2 compiler but is ok with > the > stage1 compiler. > > > struct gomp_work_share > { > /* This member records the SCHEDULE clause to be used for this construct. > The user specification of "runtime" will already have been resolved. > If this is a SECTIONS construct, this value will always be DYNAMIC. */ > union { > struct { > /* This is the chunk_size argument to the SCHEDULE clause. */ > long chunk_size; > > /* This is the iteration end point. If this is a SECTIONS construct, > this is the number of contained sections. */ > long end; > > /* This is the iteration step. If this is a SECTIONS construct, this > is always 1. */ > long incr; > }u; > > }; > }a; > > Where the stage2 compiler goes : > > /home/40000/t1.i:21:3: warning: unnamed struct/union that defines no instances > [enabled by default] > > > and the stage1 compiler goes quietly. > > Could you check if you see similar behaviour ? > Yes, exactly. I noted in my original report that the error occurred in stage2. While not fully confirmed, it looks like this was introduced by r178692. Dave