pike Wrote:...on Openelec I get a toast saying 'Login failed: False', and here's the log: http://paste.ubuntu.com/773576/
Is there some old cache somewhere which interferes with the login and which I can clear?
First to answer your question, No we don't cache anything with regards to login, whenever it fails, it's because the login process has failed.
I'm getting quite tired of this error...
We've seen this almost since day one of the plugin, and created a trac ticket a long time ago:
http://trac.xbmc.org/ticket/11599
We're working on some code that will give a more meaning full response, but it wont fix the problem you're experiencing, since its basically out of our hands.
From your log (and almost 100 other logs we've seen so far):
Code:
T:139740014212864 NOTICE: [YouTube-2.8.0] _fetchPage : 'connecting to server... https://accounts.google.com/ServiceLoginAuth'
T:139740014212864 NOTICE: [YouTube-2.8.0] _fetchPage : 'URLError : <urlopen error _ssl.c:316: Invalid SSL protocol variant specified.>'
Basically whats happening here, is that we're telling python's urllib2 to open the url "https://accounts.google.com/ServiceLoginAuth", and urllib2 is telling us that it doesn't know the "ssl variant" that the page is using. Basically something is wrong with the build of python since its unable to negotiate a proper ssl connecton. Note that the code required for such a connection is only the following 3 lines (very simple):
Code:
request = urllib2.Request("https://accounts.google.com/ServiceLoginAuth")
con = urllib2.urlopen(request)
result = con.read()
We've got running integration test that perform full login procedures that function perfectly well on "standard" unix python distributions (
seen here )
But for some reason this is only a problem for the xbmc distributed version of python. Personally I can recreate this error with very little effort on my OSX and Windows XBMC systems, even with the latest nightly It only takes a couple of login retries before xbmc's python starts throwing these errors, and the only fix i found is restarting xbmc. I'm guessing that the open ELEC guys in their work to create a reduced XBMC installation probably broke something with regards to the python installation.