Closed GoogleCodeExporter closed 9 years ago
Works over here, so I'd say this must be an App Engine bug for one of your
appids.
Original comment by wkornew...@gmail.com
on 5 Jun 2009 at 1:31
What so that's it? *your* version works on *your* operating system and *your*
computer so therefore there is no bug?
Original comment by idrinkto...@gmail.com
on 5 Jun 2009 at 1:47
OK, could you please try to use remoteapi with some other appid, some other OS,
and
then, please also try to do the same with the beta1 version on our website, so
you
can really nail down that the problem exists on all appids and on which
platforms and
with which AEP versions? That would really save me a ton of work.
What you probably can't know is that the whole remoteapi communication is done
via
Google's SDK while app-engine-patch merely calls a simple function to activate
remoteapi. Since your traceback indicates that it's a simple HTTP communication
error
it's very very unlikely that the problem is anywhere in app-engine-patch. I
think
it's either a problem with urllib on your platform or it's a temporary issue on
Google's servers (we had a few such "temporary" bug reports here, already).
Also, please bear with me that I can't really invest hours into tracking down
everyone's bugs. I'm alone and I earn no money with this and at least 50% of all
issues here are not related to AEP, at all, and some people even post issues
asking
for help with bugs in *their* own code (as if this were a public help-desk
system).
This is quite a lot of work for a single person, so I really have to prioritize
my time.
Original comment by wkornew...@gmail.com
on 5 Jun 2009 at 2:45
Hello.
Actually I am aware that the remote_api was part of google's sdk. The reason I'm
using the repository version is that I understand from the wiki that the
--remote
option is only available in that version and obviously I need all the
appenginepatch/django set up to run for me to be able to use the remote_api in
any
meaningful way, though if it is present in the 1.1 beta from the site then I
will
happily test on that version.
What I have established so far is that the errors appear to be associated in
some way
with the use of google apps account.
So far I have tested as follows
(google = standard google account. google apps = google apps account)
FAIL = update fails
OK = update ok and remote shell ok
App Owned By User Type Result
google apps google apps FAIL
google apps google FAIL
google google OK
google google apps FAIL
I tried the tests for the google apps accounts with two entirely different
domains
and got the same results so it's not just my apps account which is borked.
Next I'm going to test the remote_api against some non appengine-patch code
with the
same accounts. If it falls over the same way then I guess we'll know the error
lies
with google appengine and I will have, as I said before, posted this issue in
the
wrong place!
The platform I am on is OS X by the way. If it turns out that this bug is
appengine-patch related then I have vm's I can use to test it on Linux and
Windows
platforms too if needed.
Original comment by idrinkto...@gmail.com
on 5 Jun 2009 at 3:02
Hello
Further to my last message, I have just tested the remote api using the code
provided
in the google appengine remote api article from
http://code.google.com/appengine/articles/remote_api.html and a super simple non
appingine-patch application and I successfully used the remote api console to
create
a datastore entity despite the fact that I uploaded this application under the
same
appid and google apps account I was using when I reported the issue with
manage.py
update.
It does appear therefore that the issue is in some way related to using a
google apps
account *with* appengine-patch. (At least under OS X)
I'll test the same situation under windows and linux to make sure it isn't
specific
to either os x or my environment but I won't get a chance to do that until
tomorrow
at the earliest.
Original comment by idrinkto...@gmail.com
on 5 Jun 2009 at 3:34
Did you specify a host when using the basic remoteapi console? In
app-engine-patch
you can specify that in settings.py (at the bottom there is remote_host).
Original comment by wkornew...@gmail.com
on 5 Jun 2009 at 3:37
Hello
The basic remote api console defaults to using (my appid).appsopot.com, so
there was
no need. I also double checked using the live dataviewer on
appengine.google.com just
to be sure I hadn't done something dumb like just creating the datastore entity
locally ;)
For the tests I carried out against the appenginepatch-sample app from the
respository the only thing I changed was the appid.
Original comment by idrinkto...@gmail.com
on 5 Jun 2009 at 3:59
My initialization code is almost exactly the same as mentioned in the article.
The
only difference is that I activated the cookie store, so you don't have to
enter the
password on each manage.py run. Anyway, I tested the console with an apps
account and
it doesn't work. It always says "Invalid username or password." Only after three
failed login attempts it exists with your exception. Is that the same behavior
as on
your machine?
On the App Engine discussion group, quite a few people mentioned problems with
using
google apps accounts with App Engine. Google knows about it, but didn't mention
when
they'll fix it.
Original comment by wkornew...@gmail.com
on 5 Jun 2009 at 4:29
Hello
No, I can log in just fine using an apps account. The issue is when I try to do
anything further, such as create a data store entity (and that is only with an
appengine-patch app.)
When you say you tested the console with an apps account what exactly do you
mean?
(the basic console app or manage.py shell --remote)
Original comment by idrinkto...@gmail.com
on 5 Jun 2009 at 4:40
I tested with both by accessing the datastore.
Original comment by wkornew...@gmail.com
on 5 Jun 2009 at 4:42
Hello
No, that isn't the same as on my machine. On my set up there is no issue
logging in
and it fails immediately (i.e. as soon as I do something which acts on the
datastore
such as a put()
Original comment by idrinkto...@gmail.com
on 5 Jun 2009 at 5:04
Could you please try to delete your appcfg_cookies file? I don't know where it's
located on OSX.
Original comment by wkornew...@gmail.com
on 5 Jun 2009 at 6:44
Hello
Yup, tried that. Had to work out how to do that before I could test all those
different users and accounts. (For future reference, on os x it is in your home
directory)
Looking at my schedule for next week and will probably not be able to test on
other
Operating Systems till Tuesday unless the stuff I have to do on Monday (php.
yuk.)
goes quickly.
Original comment by idrinkto...@gmail.com
on 6 Jun 2009 at 6:33
Reopening, for now.
Original comment by wkornew...@gmail.com
on 8 Jun 2009 at 8:43
Any progress on tracking this down?
Original comment by wkornew...@gmail.com
on 11 Jun 2009 at 4:12
Hello
I haven't had a chance to test on other platforms yet. pretty snowed under at
the
moment! Will get back to you as soon as I have though
Original comment by robotlov...@gmail.com
on 12 Jun 2009 at 9:15
Does your traceback contain this:
Please go to
https://www.google.com/accounts/DisplayUnlockCaptcha
and verify you are a human. Then try again.
Could you please try to go to that address and unlock your account?
Original comment by wkornew...@gmail.com
on 16 Jun 2009 at 5:59
Hello,
No, that isn't in my traceback. The account is not locked as far as I know. I
ported my code to appengine_helper
(I needed it to work now and I'm not using the admin interface any more) and I
can use the remote api with the
same accounts and appids with no issues
Original comment by robotlov...@gmail.com
on 16 Jun 2009 at 6:41
Could you please try the attached django-1.1.zip (just copy it into common/zip-
packages, replacing your old django zip file)?
Original comment by wkornew...@gmail.com
on 16 Jun 2009 at 7:05
Attachments:
I suppose you can't help me with this?
Original comment by wkornew...@gmail.com
on 24 Jun 2009 at 5:25
Hello
Yes, I can help but as I already explained I am extremely busy right now. I do
(finally) have some free time on my
hands Friday and Monday so I will look into re-testing this issue and trying
out the attached file.
Thanks
Andy
Original comment by robotlov...@gmail.com
on 25 Jun 2009 at 10:43
OK, thanks, just wanted some feedback about the status of this ticket.
Original comment by wkornew...@gmail.com
on 25 Jun 2009 at 11:47
Still too busy? It's really just a few minutes. Otherwise I have to remove this
due to
inactivity.
Original comment by wkornew...@gmail.com
on 8 Jul 2009 at 1:57
Sorry, this fell off my radar a little since I ported to appengine_helper but I
am happy to test it for you. To be
honest I've rather forgotten *what* I'm supposed to be testing.
You want me to test the appengine_patch 1.1 upload process but using the
supplied .zip file?
Original comment by robotlov...@gmail.com
on 8 Jul 2009 at 2:24
I wanted to know if remoteapi works for you with the django-1.1.zip in comment
19. The
problem was that your google apps accounts couldn't login via remoteapi.
Original comment by wkornew...@gmail.com
on 8 Jul 2009 at 2:32
Hello
I have tested both the attached .zip against the 1.1 branch and the june 18th
release of app-engine-patch-
1.0.2.1 release and both fail at the syncdb step (i.e. when using remote api)
when using a google apps account.
Original comment by robotlov...@gmail.com
on 9 Jul 2009 at 11:46
I really can't explain why the same login (I hope you really used the same
login)
should work with django-helper and fail with aep. The other zip contained 100%
exactly
the same remoteapi login code.
Original comment by wkornew...@gmail.com
on 10 Jul 2009 at 4:37
I don't know either. For the test against the 1.0.2.1 release I simply
downloaded the file, changed the app id and
ran manage.py upload. As before, when using a google apps account the process
fails at the syncdb step where
it uses the remote API.
Are you no longer getting the same issue when using a google apps account for
the app/authentication?
Original comment by robotlov...@gmail.com
on 10 Jul 2009 at 9:39
I do get that issue, but it happens with absolutely all apps (aep, the
console.py
script you mentioned, django-helper, etc.). Also, Google already said that
Google Apps
accounts don't work and recently Alexander Vasiljev reported posted about this
problem,
too. Alexander found out that only one Google Apps account works: the domain
admin. All
others fail. What's strange is that you said you only get the problem with aep
and
after you switched to django-helper it went away. Could it be that you tested
with
different accounts (using the domain admin with django-helper and some other
account
with aep)? That could explain the disparity.
Original comment by wkornew...@gmail.com
on 10 Jul 2009 at 9:51
Hello
I don't get the issue at all with django-helper or with the console app from
the Remote API article. Aside from
when I have used different accounts to test this issue I have used the same
google apps account for all my
GAE development. It is not the domain administrator either.
"Also, Google already said that Google Apps accounts don't work" Could you
point me to where they have said
this. I am planning to deploy a couple of applications into clients google apps
accounts and that is a little
worrying!
With the google apps account you are using to test this, has the GAE
application been added to the google
apps account via the dashboard?
Original comment by robotlov...@gmail.com
on 11 Jul 2009 at 5:52
Hi,
yes, we added the account via the dashboard.
Also, please take a look at the reply from Nick Johnson in this thread (I think
I
also saw other discussions, but this is the only one I could find right now):
http://groups.google.com/group/google-
appengine/browse_thread/thread/b4d0fe4010561585/6881cc2f7bbc7c24
Did you also use the same appid when testing?
Is one of your appids configured to only accept Google Apps accounts?
Original comment by wkornew...@gmail.com
on 13 Jul 2009 at 8:49
Hello
Firstly, thanks for the link to the comment thread.
In answer to your question, the appid I have been using for development of the
project which was previously
using appengine-patch was configured to use only google apps accounts and the
admin user is a member of
that google apps domain.
If you have a look at comment 4 and 5 I have also carried out testing with a
non google apps appid which was
configured to allow authentication from any google account.
I have to admit I'm pretty stumped. It seems strange that we are experiencing
different behavior from the very
similar tests.
My one idea to help track down the issue is for me to create a very simple app
which works under both
appengine-helper and appengine-patch but which will (presumably) display the
issues with remote api under
appengine-patch.
I can then send this to you and you could compare with the results you get in
your environment.
Let me know what you think
Thanks
Andy
Original comment by robotlov...@gmail.com
on 13 Jul 2009 at 1:42
Hey Andy,
independent of the tool I use, I always get the same results as in your comment
4. It
would be great if you could run those tests with exactly the same appid & user
combinations on both aep and the helper. The test app just needs a very simple
model
which can be used via the remote shell - no need to write something fancy.
Thanks!
BTW, in the tests you might have to set the remote_host (see bottom of
settings.py)
if you use a custom domain.
Original comment by wkornew...@gmail.com
on 13 Jul 2009 at 2:52
Hello
I created a test app as discussed and got a rather unexpected result as the
appengine patch version worked!
Just to be clear, the account I tested this under was a google apps account. I
uploaded with a google apps
user from that domain (the appid creator) and (and I'm pretty sure this is
critical) the app had been added to
the google apps domain using the instructions found here -->
http://www.google.com/support/a/bin/answer.py?hl=en&answer=91077
With previous versions of appengine patch this has not been the case, and you
yourself are having problems
with appengine patch and google apps domains so this really wasn't the result I
was expecting at all!
Full Source Code is attached. You will need to set the appid.
The test application is called testapp and I tested the remote api as follows
./manage.py shell --remote
from testapp.models import TestModel
t = TestModel('I am from appengine patch')
t.put()
I then checked the appengine data viewer for my app and the model instance had
been created.
Good news, but slightly confusing...
Original comment by robotlov...@gmail.com
on 14 Jul 2009 at 3:25
Attachments:
OK, I don't have an appid set to a domain like you did, but I think I can now
explain
what the issue is when you try to access a page with required login ("login:
admin"
in the case of remote_api):
When you're not logged in you get redirected to the login page. Now, there are
two
cases:
When you bind your appid to an apps domain you get redirected to the
/a/yourdomain.com version, so login works.
When you don't bind your appid to an apps domain you get redirected to the
normal
Google Accounts login page, but you'd have to get redirected to the
/a/yourdomain.com
version in order to use a Google Apps account.
This means that you always have to bind your appid (i.e., the login system) to
a
domain.
Since you've found a solution that works for you I can close this issue. Now,
it
would be really cool to welcome you back as an app-engine-patch user. ;)
Original comment by wkornew...@gmail.com
on 14 Jul 2009 at 4:21
BTW, your code doesn't work for my appids (which all aren't setup like in your
test).
Original comment by wkornew...@gmail.com
on 14 Jul 2009 at 4:21
Hello. Yes, that makes sense.
Too late to port the current app back to appengine patch but I shall certainly
look at using it for the next one I
do.
Thanks
Andy
Original comment by robotlov...@gmail.com
on 14 Jul 2009 at 5:04
Original issue reported on code.google.com by
idrinkto...@gmail.com
on 5 Jun 2009 at 1:07