On Thu, 2019-03-28 at 11:53 +0000, Daniel P. Berrangé wrote:
On Thu, Mar 28, 2019 at 12:30:30PM +0100, Michal Privoznik wrote:
Quite frankly, I hate this patch. But the problem is in our mocking, not in the actual code its testing. And I've tried everything I was able to come up with on how to fix this. So if you have any idea, I'm all ears.
I'm happy to take a look at this. I wouldn't be surprised if there is an entirely different syscall API that needs mocking on 32bit to deal with the largefile stuff.
Thanks in advance for looking at this :) Just one more note:
+#if defined(__x86_64__) || defined(__amd64__) || defined(__aarch64__) +/* XXX Dirty hack, but mocking stat on 32bits is above my skills */
If we actually end up doing something like this after all, we should use a blacklist approach, excluding architectures where we know the test fails, rather than a whitelist approach such as the one presented here. Case in point, the above excludes a bunch of architectures where the tests work fine on account of their 64-bitness: ppc64, riscv64 and s390x. -- Andrea Bolognani / Red Hat / Virtualization