On Mon, 2019-04-01 at 12:47 -0400, Cole Robinson wrote:
[...]
> @@ -170,6 +150,11 @@ mymain(void)
> char *fakerootdir;
> struct testQemuInfo info;
> virQEMUDriverConfigPtr cfg = NULL;
> + virHashTablePtr capslatest = NULL;
> +
> + capslatest = testQemuGetCapsLatest();
> + if (!capslatest)
> + abort();
Woah, that's a bit harsh, isn't it? How about a nice and polite
return EXIT_FAILURE;
instead? :)
[...]
> @@ -192,11 +177,14 @@ mymain(void)
>
> # define DO_TEST_FULL(name, when, gic, ...) \
> do { \
> - if (testInfoSetCommon(&info, gic) < 0) { \
> + if (testQemuInfoSetArgs(&info, capslatest, \
> + ARG_GIC, gic, \
> + ARG_QEMU_CAPS, __VA_ARGS__, QEMU_CAPS_LAST, \
> + ARG_END) < 0 || \
> + qemuTestCapsCacheInsert(driver.qemuCapsCache, info.qemuCaps) < 0) { \
I haven't really spent any time digging, but that call to
qemuTestCapsCacheInsert() looks odd to me. What exactly is the point
in caching if we're going to be using a different set of capabilities
pretty much every single time?
Considering it is pre-existing, of course, your change is perfectly
fine. Feel free to investigate further, though :)
Additional note: xml2argv calls this in the test function rather than
in the corresponding macro, and it would be nice if they would be
made consistent. Follow-up material, of course.
With the abort() call replaced with something more gentle,
Reviewed-by: Andrea Bolognani <abologna@xxxxxxxxxx>
--
Andrea Bolognani / Red Hat / Virtualization
--
libvir-list mailing list
libvir-list@xxxxxxxxxx
https://www.redhat.com/mailman/listinfo/libvir-list
