On 11/8/18 1:48 PM, Jakub Jelinek wrote: > On Thu, Nov 08, 2018 at 01:43:29PM -0700, Jeff Law wrote: >> On 11/8/18 1:27 AM, Martin Liška wrote: >>> libsanitizer/ChangeLog: >>> >>> 2018-11-08 Martin Liska <mli...@suse.cz> >>> >>> PR sanitizer/87892 >>> * (all files): Revert upstream r318802. >> Is it causing a build failure or somesuch? ie, why specifically are you >> wanting to remove it? > > Yes. But perhaps it would be enough to just guard the CPU_COUNT use with > #ifdef CPU_COUNT and have some fallback, including say returning just 1. > If we care about Scudo or whatever is (what would be needed for that on the > compiler side?), then we'd need a proper implementation, one that doesn't > fail if a machine has more CPUs than fit into cpu_set_t, or if old glibc is > used and CPU_COUNT isn't defined, or even if the kernel doesn't have > affinity stuff at all. The obvious idea being to disable it with a more minimal patch making future merges easier -- if there's a less intrusive way to disable the bits, then that's fine with me.
jeff
