On Tue, Apr 14, 2020 at 06:34:24PM +0200, Tom Turelinckx wrote: > Hi, > > On Tue, Apr 14, 2020, at 8:25 AM, David Gibson wrote: > > I've just merged and pushed a change which I think will fix this to > > the master branch. Could you build and test that please? > > It seems to work, thanks! > > I've verified on debian sparc64 that make and make check succeed when I checkout b28464a, while make check fails with bus error when I checkout 87a656a. > > I've also verified that the debian src package version 1.6.0-1 builds successfully on sparc64 with patch b28464a applied locally, and that the same package with the same patch applied still builds successfully on amd64. Those are the only archs I have access to. > > > If you can make a minimal testcase which triggers this problem and > > submit it as an addition to the dtc testsuite, that would be terrific. > > I have limited experience with C development, and zero experience > with alignment issues, DTC or its test suite, so I can't be of much > help here... The relevant part of the test suite is nearly all shell. But I realized this isn't actually relevant anyway. I thought the problem was that we didn't have a test which exercised this unaligned case - but you said it tripped on the type preservation test, so there is a testcase there already. I assume it wasn't caught before just because sparc64 is more sensitive to misaligned loads than on the archs I generally build for. I would have thought ARM would have caught it, but apparently not. I can't think of any easy way to extend the testsuite to handle that, so there's not really anything to be done here. -- David Gibson | I'll have my music baroque, and my code david AT gibson.dropbear.id.au | minimalist, thank you. NOT _the_ _other_ | _way_ _around_! http://www.ozlabs.org/~dgibson
Attachment:
signature.asc
Description: PGP signature
