On Tue, Sep 05, 2017 at 02:37:21PM -0500, Eric Blake wrote: > On 09/04/2017 08:28 AM, Kashyap Chamarthy wrote: > > On Mon, Sep 04, 2017 at 08:55:23PM +0800, Fam Zheng wrote: > >> Hi Kashyap, > >> > >> On Mon, 09/04 13:16, Kashyap Chamarthy wrote: > > >>> The test 192 ("Test NBD export with '-incoming' (non-shared > >>> storage migration use case from libvirt")) is currently using HMP. > >>> Replace the HMP usage with QMP, as the upstream preference seems to be: > >>> "Use QMP where possible, unless you're explicitly testing something > >>> related to HMP". > > Kevin actually argued that keeping some HMP coverage is a GOOD thing: > > https://lists.gnu.org/archive/html/qemu-devel/2017-09/msg00798.html
Yeah, noticed it after I sent my patch. I agree with Kevin's argument, so I'll just simply drop my patch. > >> As an improvement maybe you could rebase to Stefan's "iotests: clean up > >> resources using context managers" series and switch to "with" for the temp > >> file > >> and VM. > > > > Good point. I did notice that thread[*] about resource clean up. And I > > see it's already applied to 'block-next'. Will rebase and change to the > > 'with' statement. > > I'm not sure if this patch helps us any; I personally find the unit-test > style python iotests harder to debug than the ones that produce > diff-able nnn.out files (debugging an nnn.out file that has only a list > of ..... doesn't make it easy to reproduce). If the test already runs > well in shell, what does the conversion to Python actually buy us? While turning it to QMP, thought I'd convert it to Python as the test is _very_ similar to 194 (also in Python). Given Kevin's argument in the above thread you pointed out, just disregard my patch. -- /kashyap