Hi, Again the final patch won't make it into a pull-request but I'm just keeping it around to keep track of failures. So far my numerous re-builds have been mainly plain timeouts.
The alternate co-routine builds are sailing the closest to timeout purgatory which makes me think we should limit the scope of the testing - I'm not sure how much of the unit tests or qapi tests are likely to exercise the co-routine code. Is it likely only the (currently not run) block tests really exercise this area of code? I've also updated the Travis image to be more recent. I noticed while doing so that the Travis clang was newer than our hand crafted tools so I took the opportunity to drop them and just concentrate on exercising the python stuff. It might be worth just limiting the target list for those tests to x86_64 only though? There are also a couple of Philippe's patches which I should have queued ages ago. If we get some quick reviews I can turn around a pull request by the end of the week. Alex Bennée (6): .travis.yml: disable linux-user build for gcov docker: update Travis docker image .travis.yml: rationalise clang testing .travis.yml: make current setup explicit .travis.yml: update GCC sanitizer build to GCC 7 tests/Makefile: mark flakey tests (!UPSTREAM) Philippe Mathieu-Daudé (2): docker: sort images list displayed by 'make docker' docker: do not display deprecated images in 'make docker' help .travis.yml | 82 +++++--------------------- tests/Makefile.include | 4 ++ tests/docker/Makefile.include | 5 +- tests/docker/dockerfiles/travis.docker | 7 ++- 4 files changed, 28 insertions(+), 70 deletions(-) -- 2.17.0
