On Thu, 2019-04-11 at 14:15 +0200, Ján Tomko wrote:
On Thu, Apr 11, 2019 at 12:44:16PM +0200, Andrea Bolognani wrote:
On Mon, 2019-04-01 at 12:47 -0400, Cole Robinson wrote:
if (virAsprintf(&info->outfile, - "%s/qemuxml2xmloutdata/%s-%s.xml", + "%s/qemuxml2xmloutdata/%s-%s%s.xml",
I'd definitely put another minus between these suffixes (also, I'd like to see them in use).
Looking at xml2argv, and as I suggested the implementation should be the same in xml2xml, we have # define DO_TEST_CAPS_INTERNAL(name, arch, ver, ...) \ DO_TEST_INTERNAL(name, "." arch "-" ver, \ ARG_CAPS_ARCH, arch, \ ARG_CAPS_VER, ver, \ __VA_ARGS__) # define DO_TEST_CAPS_ARCH_LATEST_FULL(name, arch, ...) \ DO_TEST_CAPS_INTERNAL(name, arch, "latest", __VA_ARGS__) # define DO_TEST_CAPS_ARCH_LATEST(name, arch) \ DO_TEST_CAPS_ARCH_LATEST_FULL(name, arch, ARG_END) # define DO_TEST_CAPS_LATEST(name) \ DO_TEST_CAPS_ARCH_LATEST(name, "x86_64") so when you use DO_TEST_CAPS_LATEST("virtio-transitional"); 'suffix' will be ".x86_64-latest" and the resulting filename will be qemuxml2xmloutdata/virtio-transitional.x86_64-latest.xml If you had a similar test for WHEN_ACTIVE only, then you'd get qemuxml2xmloutdata/virtio-transitional-active.x86_64-latest.xml It doesn't look to me like any extra separators are necessary. [...]
None of the above is really connected to whether or not we should use 'suffix' as I suggested earlier: we should definitely format it, even if it causes test suite churn. Not only that: you should also make sure...
The reason for the suffix in xml2argv is to allow the CAPS_LATEST tests to coexist with the ones with enumerated capabilities.
But it also contains the architecture, so even if -latest would be the prevailing case, I'd rather format it anyway.
Looks like we agree then :) -- Andrea Bolognani / Red Hat / Virtualization