Am I doing something wrong?

["C. Michael Pilato" <cm...@collab.net>]

I'm seeing some weird behavior with Subversion against Google Code.

Using a 1.6.x client, I tried to checkout my Thotkeeper project's tree.  I
copy-n-paste the "svn checkout https://..." line from the "Source" tab on
that project (which specifies "--username cmpilato") and get prompted for
auth.  I provide my Googlecode password, get prompted for "do you wanna save
plaintext passwords", respond "yes", and the checkout works.  I then run
'svn update' on my working copy, and -- even though I can see the cached
creds sitting in my ~/.subversion/auth area -- get prompted for auth again!
 Fill in the auth, answer "yes" to plaintext storage, update completes.
This happens every time I run 'svn update'.  Cached creds totally ignored.

However, if I make one small change -- using my fully qualified "@gmail.com"
username instead of just "cmpilato", I get prompted for auth once, answer
"yes" to plaintext storage, complete the update.  But now my (new) cached
creds finally work for future updates.

Any clues?  Is this a weirdness of Subversion or of Google Code?

Ooh... just noticed that with ra_neon, all works well -- either username is
correctly handled.  So ... this may just be a quirk in ra_serf right now.
Anybody else seeing similar stuff?

Wait... wait... when I try this same routine against the 'svnbook' Google
Code repository, my @gmail.com username doesn't pull from cache correctly.

Oh.  Shoot.  Now it's not working right for Thotkeeper either!  I'm so
confused!!!!

Why would ra_serf a) successful use a set of auth creds, b) write those auth
creds to the disk cache, and then c) ignore the cache?  (I know, I know:
"We call those bugs, Mike.")

</thinkarrhea>

-- 
C. Michael Pilato <cm...@red-bean.com> | http://cmpilato.blogspot.com/

------------------------------------------------------
http://subversion.tigris.org/ds/viewMessage.do?dsForumId=462&dsMessageId=2363244