kfogel@collab.net writes:
> Thanks for solving this long-running annoyance, Philip.
I've been thinking that my change may not be enough. I assumed that a
user would want '--config-dir arg' to behave as if the directory arg
was ~/.subversion, so --config-dir doesn't stop the system config
being read. Thus the regression tests are immune to ~/.subversion
effects but not to /etc/subversion. I think we may need to have the
regression test suite generate a config file that actively specifies
things like diff-cmd, global-ignores, etc. That would also allow
--config-dir to be passed to svnadmin create, which isn't done at the
moment since svnadmin will error rather than create a missing config.
> Small question:
>
>> Modified: trunk/subversion/tests/clients/cmdline/prop_tests.py
>> ==============================================================================
>> --- trunk/subversion/tests/clients/cmdline/prop_tests.py (original)
>> +++ trunk/subversion/tests/clients/cmdline/prop_tests.py Thu Aug 14 19:44:52 2003
>> @@ -33,6 +33,7 @@
>> def check_prop(name, path, exp_out):
>> """Verify that property NAME on PATH has a value of EXP_OUT"""
>> out, err = svntest.main.run_command(svntest.main.svn_binary, None, 1,
>> + '--config-dir', 'config',
>> 'pg', '--strict', name, path)
>> if out != exp_out:
>> print "svn pg --strict", name, "output does not match expected."
>
> (Huh, any idea why this test isn't just using run_svn() in the first
> place?)
No.
> Separately, why is the value to '--config-dir' the hardcoded string
> 'config' instead of the new variable svntest.main.config_dir?
I wrote that before I introduced the global variable, and forgot to
change it. I think using run_svn would be the best way to go.
--
Philip Martin
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Received on Fri Aug 15 18:58:09 2003