On Wed, 2019-03-06 at 16:44 +-0000, Jason Gunthorpe wrote: +AD4 On Wed, Mar 06, 2019 at 02:37:17PM +-0200, Leon Romanovsky wrote: +AD4 +AD4 diff --git a/include/linux/overflow.h b/include/linux/overflow.h +AD4 +AD4 index 40b48e2133cb..a47fb6046c0a 100644 +AD4 +AD4 +-+-+- b/include/linux/overflow.h +AD4 +AD4 +AEAAQA -227,10 +-227,10 +AEAAQA +AD4 +AD4 typeof(d) +AF8-d +AD0 d+ADs +AFw +AD4 +AD4 u64 +AF8-a+AF8-full +AD0 +AF8-a+ADs +AFw +AD4 +AD4 unsigned int +AF8-to+AF8-shift +AD0 +AFw +AD4 +AD4 - +AF8-s +AD4APQ 0 +ACYAJg +AF8-s +ADw 8 +ACo sizeof(+ACo-d) ? +AF8-s : 0+ADs +AFw +AD4 +AD4 +- (+AF8-s +AD0APQ 0 +AHwAfA +AF8-s +AD4 0) +ACYAJg +AF8-s +ADw 8 +ACo sizeof(+ACo-d) ? +AF8-s : 0+ADs +AFw +AD4 +AD4 I think this is just trading a tautological compare gcc happens to +AD4 know about today (+AF8-s +AD4APQ 0) with one it currently doesn't (+AF8-s +AD0APQ 0 +AHwAfA +AD4 +AF8-s +AD4 0) Since it is uncommon to write (+AF8-s +AD4APQ 0) as (+AF8-s +AD0APQ 0 +AHwAfA +AF8-s +AD4 0) my hope is that the gcc maintainers won't consider it interesting to make gcc report a warning about the latter expression. Bart.
