On Tue, Jun 25, 2013 at 03:45:47PM +0200, Michal Privoznik wrote:
On 25.06.2013 12:00, Daniel P. Berrange wrote:
+#define DO_TEST(file, dev, fial, ...) \
+ do { \ + const char *my_mon[] = { __VA_ARGS__, NULL}; \ + struct qemuHotplugTestData data = \ + {.domain_filename = file, .device_filename = dev, .fail = fial, \ + .mon = my_mon}; \ + if (virtTestRun(#file, 1, testQemuHotplug, &data) < 0) \ + ret = -1; \ + } while (0) What's with the 'fail' parameter you're passing across test cases. AFAICT, no test needs to be aware of the fail status of any earlier test. You're re-creating the fake monitor for each test case so no state is shared between tests. Just setting the 'ret = -1' here is sufficient
The parameter is there to tell the testQemuHotplug if error is expected or not. For instance, changing a listen address is expected to fail. Hence, qemuDomainChangeGraphics() called from the test function must return -1. However, the test function knows the error is expected, so it must return 0. This is controlled by 'fial'. I think we have similar approach elsewhere in the test suite.
Oh I see. So its really an "expectFail" flag. ACK to the original patch. Daniel -- |: http://berrange.com -o- http://www.flickr.com/photos/dberrange/ :| |: http://libvirt.org -o- http://virt-manager.org :| |: http://autobuild.org -o- http://search.cpan.org/~danberr/ :| |: http://entangle-photo.org -o- http://live.gnome.org/gtk-vnc :|