On 02/03/22 14:49, Eric Blake wrote:
On Thu, Feb 03, 2022 at 11:52:43AM +0100, Laszlo Ersek wrote:
What's the reason to test writing if the read test fails first?
When running the testsuite, it's nice to uncover as many cases as possible in a single run, rather than patching one early failure just to encounter another failure on the next test run that was previously skipped because of the first failure.
This is a great idea, and I wish it worked consistently *between* separate test cases too. I distinctly remember connect-tcp6 failing for me (earlier), and myself dismissing it, and assuming everything else was OK. Then, in the RHEL9 Brew environment, where connect-tcp6 was specifically disabled, I was hit by the actual test failures :/ (And there was no easy way to skip connect-tcp6.)
One way to do this is to write two separate tests (one that tests failed reads from NBD as source, a second that tests failed writes to NBD as destination), but writing two nearly-identical .sh files is more boilerplate overhead than just jamming related tests into the same .sh script.
Right, but this is exactly what I wish worked -- but IME doesn't always work.
Looks OK to me, but should be definitely reviewed by another reviewer.
Looks like Rich has given ACK as well.
Thanks! Laszlo