| Message ID | 20170526211131.12341-1-joe@ovn.org |
|---|---|
| State | Accepted |
| Headers | show |
On Fri, May 26, 2017 at 02:11:31PM -0700, Joe Stringer wrote: > Clang 4.0 complains: > > ../tests/test-hash.c:160:16: error: taking address of packed member 'b' of > class or structure 'offset_ovs_u128' may result in an unaligned pointer value > [-Werror,-Waddress-of-packed-member] > in0 = &in0_data.b; > > Set the bit in the aligned u128 first then copy the contents into the > offset u128 so that we don't have to take the address of the non-aligned > u128 and pass it to set_bit128. > > For the 256byte_hash, fix it up so that it's actually testing the 256B > hash inside a 32-bit offset u128 as well. > > Suggested-by: Ben Pfaff <blp@ovn.org> > Signed-off-by: Joe Stringer <joe@ovn.org> Thanks! Acked-by: Ben Pfaff <blp@ovn.org>
On 6 June 2017 at 16:13, Ben Pfaff <blp@ovn.org> wrote: > On Fri, May 26, 2017 at 02:11:31PM -0700, Joe Stringer wrote: >> Clang 4.0 complains: >> >> ../tests/test-hash.c:160:16: error: taking address of packed member 'b' of >> class or structure 'offset_ovs_u128' may result in an unaligned pointer value >> [-Werror,-Waddress-of-packed-member] >> in0 = &in0_data.b; >> >> Set the bit in the aligned u128 first then copy the contents into the >> offset u128 so that we don't have to take the address of the non-aligned >> u128 and pass it to set_bit128. >> >> For the 256byte_hash, fix it up so that it's actually testing the 256B >> hash inside a 32-bit offset u128 as well. >> >> Suggested-by: Ben Pfaff <blp@ovn.org> >> Signed-off-by: Joe Stringer <joe@ovn.org> > > Thanks! > > Acked-by: Ben Pfaff <blp@ovn.org> Thanks, applied.
diff --git a/tests/test-hash.c b/tests/test-hash.c index d1beead36ed5..5d3f8ea43f65 100644 --- a/tests/test-hash.c +++ b/tests/test-hash.c @@ -153,14 +153,13 @@ check_hash_bytes128(void (*hash)(const void *, size_t, uint32_t, ovs_u128 *), OVS_PACKED(struct offset_ovs_u128 { uint32_t a; ovs_u128 b; - }) in0_data; - ovs_u128 *in0, in1; + }) in0; + ovs_u128 in1; ovs_u128 out0, out1; - in0 = &in0_data.b; - set_bit128(in0, i, n_bits); set_bit128(&in1, i, n_bits); - hash(in0, sizeof(ovs_u128), 0, &out0); + in0.b = in1; + hash(&in0.b, sizeof(ovs_u128), 0, &out0); hash(&in1, sizeof(ovs_u128), 0, &out1); if (!ovs_u128_equals(out0, out1)) { printf("%s hash not the same for non-64 aligned data " @@ -205,14 +204,15 @@ check_256byte_hash(void (*hash)(const void *, size_t, uint32_t, ovs_u128 *), OVS_PACKED(struct offset_ovs_u128 { uint32_t a; ovs_u128 b[16]; - }) in0_data; - ovs_u128 *in0, in1[16]; + }) in0; + ovs_u128 in1[16]; ovs_u128 out0, out1; - in0 = in0_data.b; - set_bit128(in0, i, n_bits); set_bit128(in1, i, n_bits); - hash(in0, sizeof(ovs_u128) * 16, 0, &out0); + for (j = 0; j < 16; j++) { + in0.b[j] = in1[j]; + } + hash(&in0.b, sizeof(ovs_u128) * 16, 0, &out0); hash(in1, sizeof(ovs_u128) * 16, 0, &out1); if (!ovs_u128_equals(out0, out1)) { printf("%s hash not the same for non-64 aligned data "
Clang 4.0 complains: ../tests/test-hash.c:160:16: error: taking address of packed member 'b' of class or structure 'offset_ovs_u128' may result in an unaligned pointer value [-Werror,-Waddress-of-packed-member] in0 = &in0_data.b; Set the bit in the aligned u128 first then copy the contents into the offset u128 so that we don't have to take the address of the non-aligned u128 and pass it to set_bit128. For the 256byte_hash, fix it up so that it's actually testing the 256B hash inside a 32-bit offset u128 as well. Suggested-by: Ben Pfaff <blp@ovn.org> Signed-off-by: Joe Stringer <joe@ovn.org> --- tests/test-hash.c | 20 ++++++++++---------- 1 file changed, 10 insertions(+), 10 deletions(-)