Re: [libvirt] [PATCH] tests: Add DO_TEST_PARSE_ERROR() to qemuxml2xml
On Fri, Jul 07, 2017 at 13:00:10 +0200, Andrea Bolognani wrote:
On Fri, 2017-07-07 at 12:48 +0200, Peter Krempa wrote:
> > qemuxml2argv already supports the ability to include test
> > cases that are known not to make it past
XML parsing, and
> > since we want to keep qemuxml2xml in sync with it as much
> > as possible, we need to implement this missing feature.
>
> I must say that I
don't really see the point to have them fail twice ...
>
> I'd understand if this would be added to the genericxml2xmltest (if it's
> not there)
We have a lot of test cases with xml2argv coverage but no
xml2xml coverage, and vice versa. The obvious solution to
Well. If the file fails to parse, I don't see why it would make sense to
make it fail to parse twice.
The both tests in their success paths have value. Failing to parse the
same document twice does not.
that is to, in due time, make it so you only need to list
input file and capabilities once to have them tested both
for xml2argv and xml2xml, and this is a step towards that
still pretty far away goal.
This is certainly a good worthwhile goal. I just don't really see how
splitting the testing of unparsable documents into two places or
duplicating it will help your effort to simplify it.
The combined test will have to be renamed anyways and the xml2argv test
has the capability flags, so that seems as a better point to start the
rename.
> > @@ -217,6 +231,17 @@ testCompareStatusXMLToXMLFiles(const
void *opaque)
> >
> > ret = 0;
> >
> > + ok:
>
> This is not an okay name for a label.
It's clearly an "ok" name though :P
I borrowed it from testCompareXMLToArgv() in xml2argv,
but I don't have a problem changing it to something you
think is more suitable.
I'd prefer 'done'. 'ok' seems too short. Also the test still may fail
at
that point, so 'done' should not imply success (hopefully).
Attachments:
- signature.asc (application/pgp-signature — 833 bytes)