Re: [libvirt] [PATCH v3 1/3] qemuxml2argvtest: Set more fake drivers

On 07/27/2018 12:06 PM, Michal Privoznik wrote: The above error is only seen when the virSetConnectNWFilter is used... So, let's take the extra step to be more like fakeSecretDriver and fakeStorageDriver. Squash the following in and you can remove that last paragraph: diff --git a/tests/qemuxml2argvtest.c b/tests/qemuxml2argvtest.c index b905cc943c..cedef5c9ad 100644 --- a/tests/qemuxml2argvtest.c +++ b/tests/qemuxml2argvtest.c @@ -262,6 +262,33 @@ static virStorageDriver fakeStorageDriver = { .storagePoolIsActive = fakeStoragePoolIsActive, }; + +/* virNetDevOpenvswitchGetVhostuserIfname mocks a portdev name - handle that */ +static virNWFilterBindingPtr +fakeNWFilterBindingLookupByPortDev(virConnectPtr conn, + const char *portdev) +{ + if (STREQ(portdev, "vhost-user0")) + return virGetNWFilterBinding(conn, "fake_vnet0", "fakeFilterName"); + + virReportError(VIR_ERR_NO_NWFILTER_BINDING, + "no nwfilter binding for port dev '%s'", portdev); + return NULL; +} + + +static int +fakeNWFilterBindingDelete(virNWFilterBindingPtr binding ATTRIBUTE_UNUSED) +{ + return 0; +} + + +static virNWFilterDriver fakeNWFilterDriver = { + .nwfilterBindingLookupByPortDev = fakeNWFilterBindingLookupByPortDev, + .nwfilterBindingDelete = fakeNWFilterBindingDelete, +}; + typedef enum { FLAG_EXPECT_FAILURE = 1 << 0, FLAG_EXPECT_PARSE_ERROR = 1 << 1, @@ -470,6 +497,7 @@ testCompareXMLToArgv(const void *data) conn->secretDriver = &fakeSecretDriver; conn->storageDriver = &fakeStorageDriver; + conn->nwfilterDriver = &fakeNWFilterDriver; virSetConnectInterface(conn); virSetConnectNetwork(conn); ------------ What's really interesting (perhaps to me) is that as long as this patch (and the squashed hunk) are there, the error message you're trying to avoid in the next two patches no longer appears. That is - if just applying this patch, I don't see the Failed to connect socket to '/run/user/1000...'. If I comment out virSetConnectNWFilter, then the message reappears. If I flip-flop the patches (eg last 2 first) and rebuild, I don't see the changed error message. But that could be because some file isn't being rebuilt - I forget I think I tripped across it before, but cannot remember how. I did at least figure out why virSetConnectNWFilter triggers the calls you're trying to avoid from the commit message for/from the net-vhostuser-multiq. If qemuBuildVhostuserCommandLine fails, then we goto cleanup which calls virDomainConfNWFilterTeardown sending us down the path to calling virDomainConfNWFilterTeardownImpl because @conn is no longer NULL. This calls in succession virNWFilterBindingLookupByPortDev and virNWFilterBindingDelete before returning. It's still not clear to me why when using virSetConnectNWFilter does the "Failed to connect socket..." message not appear and I don't have the energy to chase that right now. So of course this is a really long way to say, I'm wondering whether the subsequent patches are really necessary or not. I'll leave that to you to decide. In any case, with the squashed bit applied... Reviewed-by: John Ferlan <jferlan(a)redhat.com&gt; John