Community
Participate
Working Groups
User-Agent: Mozilla/5.0 (Windows; U; Windows NT 6.1; en-US) AppleWebKit/532.0 (KHTML, like Gecko) Chrome/3.0.195.10 Safari/532.0 Build Identifier: 20090621-0832 As I remember this bug appeared first in the Ganymede release. Reproducible: Always Steps to Reproduce: 1. Add a new CVS repository with "extssh" connection type 2. Configure the host name and user of the repository 3. Start browsing the repository => keyboard interactive authentication dialog will open ("Password Required") 4. Type the correct password, check "Save password" and click "OK" 5. Restart you eclipse 6. Try browsing the newly created CVS repository => keyboard interactive dialog opened again Expected behaviour in step 6: automatically login using the previously saved password.
Created attachment 146461 [details] proposed patch This patch solves the issue in Ganymede and Galileo. (I haven't tested in other release.)
I could reproduce the problem. This problem will occur if the password auth method is disabled and the keyboard-interactive is enabled on the remote sshd. It seems to me that this problem will not appear if those methods are enabled simultaneously. I'll try to fix the problem. The proposed patch is not good, because it may save the incorrect password into the datastore.
I ran into the same problem, but failed to find this Bugzilla entry. So I went the longer way and reported it as a bug in JSch: http://sourceforge.net/tracker/?func=detail&aid=2964328&group_id=64920&atid=509122 Note that I came up with a different patch which fixes the problem for me at the JSch level, the patch can be seen in the mentioned bug report. Adding,
Adding Martin Oberhuber to CC per his request in orbit-dev mailing list where he directed me to the proper place to report the bug to.
On a superficial review, the JSch patch looks good to me, though there may be potential for improving (e.g. also allow case-insensitive password prompt, or a prompt like "Enter Password :"). Unfortunately, due to the Eclipse requirement for Legal Review (which may take some time), we can not guarantee getting an updated version of JSch into Eclipse Platform for the 3.6 (Helios) release. At any rate, please ping here on the bug when a new JSch release with the fix is available such that we could try getting it in. If possible, fixing this on the Eclipse side may be preferrable though.
(In reply to comment #5) >Unfortunately, due to the Eclipse requirement for Legal Review (which may take >some time), we can not guarantee getting an updated version of JSch into >Eclipse Platform for the 3.6 (Helios) release. > >At any rate, please ping here on the bug when a new JSch release with the fix >is available such that we could try getting it in. > > If possible, fixing this on the Eclipse side may be preferrable though. I agree with you. I don't want to become the release blocker. At first, this problem should be fixed on the Eclipse side, and then fixed on JSch lib side. As for the proposed fix, I confirmed that it never save the incorrect password into the datastore. It was my misunderstanding. I'll accept it. May I apply that patch to CVS?
(In reply to comment #6) > As for the proposed fix, I confirmed that it never save the incorrect password > into the datastore. It was my misunderstanding. I'll accept it. > > May I apply that patch to CVS? Sure. This is a good time to do it since we've just got M6 declared.
I have committed the fix to CVS HEAD.
As I understand there will be no further work on the problem. I'm marking the bug as FIXED since the patch has been in CVS for two weeks. The change will be available in today's I build. Michal, could you please get the build (I20100330-0800) and verify the problem is gone?