On Thu, Aug 11, 2022 at 11:31:21AM -0700, Nathan Chancellor wrote:
> On Thu, Aug 11, 2022 at 06:22:54PM +0100, Sudip Mukherjee (Codethink) wrote:
> > Hi All,
> >
> > Not sure if it has been reported, builds of arm64 with clang failed to
> > build next-20220811 with the error:
> >
> > drivers/vdpa/vdpa_sim/vdpa_sim_blk.c:201:3: error: expected expression
> > struct virtio_blk_discard_write_zeroes range;
> > ^
> > drivers/vdpa/vdpa_sim/vdpa_sim_blk.c:204:25: error: use of undeclared identifier 'range'
> > if (to_pull != sizeof(range)) {
> > ^
> > drivers/vdpa/vdpa_sim/vdpa_sim_blk.c:207:21: error: use of undeclared identifier 'range'
> > to_pull, sizeof(range));
> > ^
> > drivers/vdpa/vdpa_sim/vdpa_sim_blk.c:212:60: error: use of undeclared identifier 'range'
> > bytes = vringh_iov_pull_iotlb(&vq->vring, &vq->out_iov, &range,
> > ^
> > drivers/vdpa/vdpa_sim/vdpa_sim_blk.c:222:38: error: use of undeclared identifier 'range'
> > sector = vdpasim64_to_cpu(vdpasim, range.sector);
> > ^
> > drivers/vdpa/vdpa_sim/vdpa_sim_blk.c:224:43: error: use of undeclared identifier 'range'
> > num_sectors = vdpasim32_to_cpu(vdpasim, range.num_sectors);
> > ^
> > drivers/vdpa/vdpa_sim/vdpa_sim_blk.c:225:37: error: use of undeclared identifier 'range'
> > flags = vdpasim32_to_cpu(vdpasim, range.flags);
> > ^
> > drivers/vdpa/vdpa_sim/vdpa_sim_blk.c:202:7: error: mixing declarations and code is incompatible with standards before C99 [-Werror,-Wdeclaration-after-statement]
> > u32 num_sectors, flags;
> > ^
> > 8 errors generated.
> >
> >
> > git bisect pointed to d79b32c2e4a4 ("vdpa_sim_blk: add support for discard and write-zeroes").
> > And, reverting that commit has fixed the build failure.
> >
> > I will be happy to test any patch or provide any extra log if needed.
>
> I am very surprised GCC does not error out in the same way, since as far
> as I understand it, labeled statements have to be followed by a
> statement and a declaration is not a statement in C so braces are
> needed. In fact, it seems like something changed (regressed?) between
> GCC 10.x and 11.x?
>
> https://godbolt.org/z/EYaGa1eE3
>
> I am going to bisect GCC to find out whether or not that was
> intentional.
Just for the record, it was:
https://gcc.gnu.org/git/?p=gcc.git;a=commit;h=8b7a9a249a63e066cff6e95db05a3158b4cc56cc
The clang developers are aware, it seems it will get implemented at some
point:
https://github.com/llvm/llvm-project/issues/56040
Cheers,
Nathan
