We have three competing PRs to address the crashes when -O0 is used: lixiaoy1 https://github.com/ceph/ceph/pull/12796 kefu https://github.com/ceph/ceph/pull/13819 adam https://github.com/ceph/ceph/pull/12931 Adam's is appealing because it will ensure the bounded encodes are in fact constant at compile time. Unfortunately, it requires you to define DENC twice, like so: - DENC_FEATURED(bar_t, v, p, f) { + DENC_FEATURED_BOUNDED(bar_t, v, p, f) { ::denc(v.a, p, f); ::denc(v.b, p, f); } + DENC_FEATURED_BOUND_ENCODE(bar_t, p, f) { + DENC_START(1, 1, p); + ::denc(bounded_t<decltype(bar_t::a)>{}, p, f); + ::denc(bounded_t<decltype(bar_t::b)>{}, p, f); + DENC_FINISH(p); + } which feels like a deal-killer to me... I think I'd rather just drop the check entirely and rely on the developer to be careful. The other two PRs fix up the handful of type(s) that were actually causing failures to manually define bound (again making the single DENC macro unusable). What about instead changing all of the instances of denc(*(const T*)nullptr, elem_size); with a macro that will either do the above "*(const T*)nullptr" -or-, if it looks like this is a debug build or compiler optimizations are off or some other compiler environment indicates it won't work, replace it with "T()"? That avoids crashes in the -O0 case, but still lets us keep some safeguard for other builds? Any other ideas? sage -- To unsubscribe from this list: send the line "unsubscribe ceph-devel" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
