On Mon, Jan 28, 2019 at 10:34:54AM -0800, Steve Langasek wrote: >On Mon, Jan 28, 2019 at 12:45:22PM +0000, Steve McIntyre wrote: >> Hi Alastair, > >> On Mon, Jan 28, 2019 at 09:50:30AM +0000, Alastair McKinstry wrote: > >> >Yes, thanka for this patch. I'm reluctant to consider that binaries built on >> >armhf failing on arm64 is an _RC Bug_, but its definitely undesirable. > >> It's not *yet* RC, but it's going to go that way soon. Our plan is to >> switch to building for armel and armhf using arm64 machines before too >> long, which would reliably break all packages with bugs like this. > >If it were only about the buildds, then Alastair's solution (disabling the >testsuite on this arch), though suboptimal, would probably be acceptable. >But I think it is increasingly likely that people who are running armhf >Debian binaries will be doing so on 64-bit chips with 64-bit kernels, where >unaligned fixups will not happen. > >This is also the configuration that android 32-bit kernels shipped in since >forever. > >So there is a small but significant and increasing set of configurations >where armhf binaries need to avoid unalign access in order to be useful.
Yup, agreed also. People should also *not* be disabling test suites that are showing real issues being found! -- Steve McIntyre, Cambridge, UK. st...@einval.com < liw> everything I know about UK hotels I learned from "Fawlty Towers"