Message ID | 20191118100719.7968-1-david@redhat.com (mailing list archive) |
---|---|
State | New, archived |
Headers | show |
On 18/11/19 11:07, David Hildenbrand wrote:
> https://github.com/davidhildenbrand/kvm-unit-tests.git tags/s390x-2019-11-18
Pulled, thanks.
Paolo
For what its worth (maybe a future patch) travis has now alpha s390x build support https://docs.travis-ci.com/user/multi-cpu-architectures I think the number of instances for s390 is still limited, but it works for open source software. On 18.11.19 11:07, David Hildenbrand wrote: > Hi Paolo, > > The following changes since commit af8dfe97b47d03876041506e8b38f718914aeea4: > > x86/unittests.cfg: Increase the timeout of the sieve test to 180s (2019-11-15 15:56:19 +0100) > > are available in the Git repository at: > > https://github.com/davidhildenbrand/kvm-unit-tests.git tags/s390x-2019-11-18 > > for you to fetch changes up to c71d8a9fab524269785d688eb0db7f6391b5510c: > > travis.yml: Expect that at least one test succeeds (2019-11-18 10:47:27 +0100) > > ---------------------------------------------------------------- > Cleanups and bugfixes (especially to make the tests run natively under > LPAR). Travis CI improvements (e.g., x86 support, KVM support on x86_64). > > ---------------------------------------------------------------- > Claudio Imbrenda (3): > s390x: remove redundant defines > s390x: improve error reporting for interrupts > s390x: sclp: expose ram_size and max_ram_size > > Janosch Frank (4): > s390x: Use loop to save and restore fprs > s390x: Fix initial cr0 load comments > s390x: Add CR save area > s390x: Load reset psw on diag308 reset > > Thomas Huth (5): > travis.yml: Re-arrange the test matrix > travis.yml: Install only the required packages for each entry in the > matrix > travis.yml: Test with KVM instead of TCG (on x86) > travis.yml: Test the i386 build, too > travis.yml: Expect that at least one test succeeds > > .travis.yml | 156 ++++++++++++++++++++++++++------------- > lib/s390x/asm-offsets.c | 3 +- > lib/s390x/asm/arch_def.h | 6 +- > lib/s390x/interrupt.c | 8 +- > lib/s390x/sclp.c | 10 +++ > lib/s390x/sclp.h | 4 +- > lib/s390x/smp.c | 2 +- > s390x/cstart64.S | 76 ++++++++----------- > 8 files changed, 158 insertions(+), 107 deletions(-) >
On 18/11/2019 12.11, Christian Borntraeger wrote: > For what its worth (maybe a future patch) travis has now alpha s390x build support > https://docs.travis-ci.com/user/multi-cpu-architectures > > I think the number of instances for s390 is still limited, but it works for open > source software. I already tried that... Unfortunately the LXD containers that they use on s390x (and ppc64 and arm64) are rather limited - you can't use KVM here on Travis, unlike with x86. So the only advantage of adding a s390x build here is that we'd check native builds instead of using the cross compiler... not sure whether that gives us that much more of test coverage for the kvm-unit-tests... Thomas
On 18.11.19 12:21, Thomas Huth wrote: > On 18/11/2019 12.11, Christian Borntraeger wrote: >> For what its worth (maybe a future patch) travis has now alpha s390x build support >> https://docs.travis-ci.com/user/multi-cpu-architectures >> >> I think the number of instances for s390 is still limited, but it works for open >> source software. > > I already tried that... Unfortunately the LXD containers that they use > on s390x (and ppc64 and arm64) are rather limited - you can't use KVM > here on Travis, unlike with x86. > > So the only advantage of adding a s390x build here is that we'd check > native builds instead of using the cross compiler... not sure whether > that gives us that much more of test coverage for the kvm-unit-tests... Right. It might be more useful for the qemu travis build?
On 18/11/2019 12.32, Christian Borntraeger wrote: > > > On 18.11.19 12:21, Thomas Huth wrote: >> On 18/11/2019 12.11, Christian Borntraeger wrote: >>> For what its worth (maybe a future patch) travis has now alpha s390x build support >>> https://docs.travis-ci.com/user/multi-cpu-architectures >>> >>> I think the number of instances for s390 is still limited, but it works for open >>> source software. >> >> I already tried that... Unfortunately the LXD containers that they use >> on s390x (and ppc64 and arm64) are rather limited - you can't use KVM >> here on Travis, unlike with x86. >> >> So the only advantage of adding a s390x build here is that we'd check >> native builds instead of using the cross compiler... not sure whether >> that gives us that much more of test coverage for the kvm-unit-tests... > > Right. It might be more useful for the qemu travis build? Yes, it might be useful for QEMU. But there are some issues to be solved first, e.g. some tests of QEMU are failing there in very weird ways (I guess it's because some system calls are blocked in the LXD container). Thomas