On 17.12.2012 15:03, C. Michael Pilato wrote:
> On 12/17/2012 05:19 AM, Branko Čibej wrote:
>> On 17.12.2012 09:47, Bert Huijben wrote:
>>> I think we should suppress this warning on Windows when the CryptoAPI
>>> encryption is enabled (read: +- always) to avoid unneeded user questions.
>> I rather think we should disable plaintext password storage by default,
>> I don't care about Windows, and let people enable it explicitly.
> May I assume you mean "runtime-disable", not "compile-time-disable"?
Yes, of course. Here's the thing, though: the define in
svn_private_config.h doesn't actually do what it says. For example, if
someone should define it on Windows, it would disable storing passwords
on disk, even though they're (almost) always encrypted on that platform.
Another dimension of the problem is related to tests, i.e., how to get
this bit from svn_private_config.h into basic_tests.py:basic_auth_test.
I added the warning to svn --version so that I could remove the parsing
of svn_private_config.h. Now, of course, I find that the warning is not
quite precise, and the lack of the warning doesn't mean the test will
properly XFail.
Clearly, as far as tests are concerned, scanning svn_private_config.h is
better than trying to glean information from svn --version; and the
warning in svn --version cannot rely just on the presence of that
symbol. Sadly, a configuration object is not available to
svn_opt__print_version_info.
So now I'm in a dilemma: the warning can be false-positive, and so can
the (lack of) the config option. And the warning, even if it were made
precise, doesn't actually provide the information that the test suite needs.
-- Brane
--
Branko Čibej
Director of Subversion | WANdisco | www.wandisco.com
Received on 2012-12-17 19:06:17 CET