On 06/03/17 12:31, Roger Pau Monne wrote: > Clang 4.0 has added a new check that triggers when taking the address of a > field of a packed structure. I've explained our situation to the llvm/clang > folks, but nobody seem to care: > > https://reviews.llvm.org/D20561 > > So simply disable that check if present. > > Signed-off-by: Roger Pau Monné <roger....@citrix.com>
In this case, I'd prefer that we drop the unnecessary packed on segment_attributes and segment_register, perhaps adding some BUILD_BUG_ON()'s in the SVM code to check that the structure still matches hardwares representation. Or were there other structures which ended up in a similar situation? ~Andrew _______________________________________________ Xen-devel mailing list Xen-devel@lists.xen.org https://lists.xen.org/xen-devel
