On Thu, Apr 20, 2023 at 12:08:36PM -0400, Theodore Ts'o wrote:
> Commit b76a6cdb40b5 ("report: derive an xml schema for the xunit
> report") in fstests upstream adds an xmlns attribute to the xunit
> <testcase/> tag. Unfortunately, this breaks the junitparser.py Python
> package, since it uses lxml Python package, and by adding an xmlns
> specifier, junitxml.py would need to know the schema and map that to
> namespace tag.
>
> So edit it out of the xml file using sed, which relies on the fact
> that fstests will add the xmlns file on a single line. The "right"
> way would be to use an XSLT processor, but that would bloat the test
> appliance significantly. So we'll just cheat for now while we discuss
> with fstests upstream whether adding the xmlns attribute is really
> worth the pain and incompatibility that it causes.
>
> Signed-off-by: Theodore Ts'o <tytso@xxxxxxx>
> ---
> test-appliance/files/root/runtests.sh | 1 +
> 1 file changed, 1 insertion(+)
>
> diff --git a/test-appliance/files/root/runtests.sh b/test-appliance/files/root/runtests.sh
> index 9b32d287..c4ddb739 100755
> --- a/test-appliance/files/root/runtests.sh
> +++ b/test-appliance/files/root/runtests.sh
> @@ -23,6 +23,7 @@ function copy_xunit_results()
>
> if test -f "$RESULT"
> then
> + sed -i.orig -e 's/xmlns=\".*\"//' "$RESULT"
FWIW I wouldn't mind commenting out the xmlns bits in common/report.
Some XML parsers have this annoying quirk that as soon as anyone
attaches a namespace to the document, every query against that document
has to reference the given namespace explicitly.
Though really, the *frustrating* thing is that in order to enable
automatic validation via the schema document, one has to embed the
xsi:schemaLocation attribute, which forces you to put in a namespace.
Then un-namespaced queries fail, and yay XML. :?
Thoughts?
--D
> if test -f "$RESULTS"
> then
> merge_xunit "$RESULTS" "$RESULT"
> --
> 2.31.0
>
