On Tue, Feb 25, 2025 at 12:54:40PM -0500, Nathan Hartman wrote: > On Tue, Feb 25, 2025 at 9:50 AM Daniel Sahlberg <daniel.l.sahlb...@gmail.com> > wrote: > > > Hi, > > > > The following test fails, for obvious[1] reasons, since January last year: > > > > [[[ > > $ make check ALLOW_REMOTE_HTTP_CONNECTION=1 > > TESTS=subversion/tests/cmdline/dav_tests.py#3 > > [1/1] > > dav_tests.py...............................................................................................FAILURE > > At least one test FAILED, checking /home/dsg/svn_trunk2/tests.log > > FAIL: dav_tests.py 3: connect to GitHub's SVN bridge > > Summary of test results: > > 1 test FAILED > > Python version: 3.12.3. > > SUMMARY: Some tests failed > > > > make: *** [Makefile:553: check] Error 1 > > ]]] > > > > The test is explicitly to connect to GitHub. There are other tests in the > > same file connecting to other (local or remote) http servers. I'm proposing > > to simply remove the test. > > > > +1 to removing the test, since it cannot pass (due to 3rd party changes > outside our control). Unless the idea below makes sense: > > There seems to be a test condition (is_remote_http_connection_allowed) and > > corresponding testsuite option (--allow-remote-http-connection) > > in cmdline/svntest/main.py, build/run_tests.py and Makefile.in. I'm not so > > sure about removing these, in case it breaks scripts for someone. On the > > other hand the above test is the only user so we would leave unused code if > > they are not also removed. > > > > > > > Thoughts? > > > > To give a *good* answer here, I would need to actually look at the code. > But my initial reaction is: > > Maybe the failing test can be modified to contact Subversion's Subversion > repository instead? (Still gated by is_remote_http_connection_allowed.)
I believe the circumstances of this test getting added involved an issue in our svn client which prevented connections to Github's former SVN service from working. This was a user-reported problem, if I recall correctly. The test's purpose is to alert us of similar issues before they end up being shipped in a release of Subversion. As far as I know there are no other 3rd party server implementations we could test against. So I am in favour of removing this test. "svn info" calls with URLs against our own server implementation have plenty of test coverage elsewhere in the test suite. I agree we could keep the remote-connection-allowed flag. It does no harm and would just be a no-op for now. Cheers, Stefan
