On Tue, 24 Jun 2025 at 08:16:42 +0000, Graham Inggs wrote:
rdflib's autopkgtest is currently regressed in testing on all
architectures [1]. I've copied what I hope is the relevant part of
the log below.
A more relevant part seems to be:
87s # A url that returns content-type text/html.
87s url = "https://github.com/RDFLib/rdflib/issues/2734"
87s with pytest.raises(PluginException):
87s > graph = Graph().parse(url)
...
87s self = <urllib.request.HTTPDefaultErrorHandler object at 0x7f40b9c7cc20>
87s req = <urllib.request.Request object at 0x7f40b80a5860>
87s fp = <http.client.HTTPResponse object at 0x7f40b80ac040>, code = 406
87s msg = 'Not Acceptable'
87s hdrs = <http.client.HTTPMessage object at 0x7f40b50f3250>
87s
87s def http_error_default(self, req, fp, code, msg, hdrs):
87s > raise HTTPError(req.full_url, code, msg, hdrs, fp)
87s E urllib.error.HTTPError: HTTP Error 406: Not Acceptable
87s
87s /usr/lib/python3.13/urllib/request.py:613: HTTPError
It looks as though this test is accessing remote resources (in this
case, a Github issue) in a way that has changed its behaviour, possibly
as a result of website changes on Github.
As well as fixing the regression, this test presumably needs the
needs-internet restriction. Or, the test that needs internet access and
is failing could perhaps be skipped?
Or, if accessing these remote resources can be forced to fail (perhaps
by setting an unreachable http_proxy and https_proxy, like
/usr/share/dh-python/pybuild does during build-time testing) then
perhaps the test would be skipped automatically?
smcv