Blobstore upload redirects to wrong module - google-app-engine

The Problem
When uploading a file using the Go Blobstore API, the success path redirects to the wrong appengine module. Here is a more visual description of the problem:
User lands on the upload page of module A: http://A.my-appengine-app.com/upload
User's browser makes a request to the module for an upload session: http://A.my-appengine-app.com/upload/session
Module A defines a handler for /upload/session which runs the following Go code: url, err := blobstore.UploadURL(c, "/upload/success")
The method returns a URL, similar to: http://A.my-appengine-app.com/_ah/upload/[some long hash]/
This URL is relayed back to the user's browser and inserted into the action of a <form>.
The user submits a multipart POST request to the URL
Whatever handles the URL (some non-user-space appengine handler), attempts to redirect back to /upload/success
This is where things get weird. In development, the server redirects to "/upload/success" in module A. In production, the server redirects to the main module, which we can call B for now. I can tell this is happening because I am getting a 404 in my web console and the logs indicate that the request is being made to module B. I've even gone so far as to explicitly pass the hostname as part of the success path (step #3), but to no effect.
Current Solution (Not Ideal)
It seems my only recourse is to define a handler in module B to handle the request as module A would. Since the goapp architecture globs all the modules together, this isn't the worst tradeoff in the world, but it's semantically wrong given that modules are supposed to be vertically independent. If anyone has any idea how to work around this, I'd be obliged. For now, I'll take the aforementioned approach.

Related

GMAIL API ACCESS ISSUE [duplicate]

On the website https://code.google.com/apis/console I have registered my application, set up generated Client ID: and Client Secret to my app and tried to log in with Google.
Unfortunately, I got the error message:
Error: redirect_uri_mismatch
The redirect URI in the request: http://127.0.0.1:3000/auth/google_oauth2/callback did not match a registered redirect URI
scope=https://www.googleapis.com/auth/userinfo.profile https://www.googleapis.com/auth/userinfo.email
response_type=code
redirect_uri=http://127.0.0.1:3000/auth/google_oauth2/callback
access_type=offline
approval_prompt=force
client_id=generated_id
What does mean this message, and how can I fix it?
I use the gem omniauth-google-oauth2.

The redirect URI (where the response is returned to) has to be registered in the APIs console, and the error is indicating that you haven't done that, or haven't done it correctly.
Go to the console for your project and look under API Access. You should see your client ID & client secret there, along with a list of redirect URIs. If the URI you want isn't listed, click edit settings and add the URI to the list.
EDIT: (From a highly rated comment below) Note that updating the google api console and that change being present can take some time. Generally only a few minutes but sometimes it seems longer.

In my case it was www and non-www URL. Actual site had www URL and the Authorized Redirect URIs in Google Developer Console had non-www URL. Hence, there was mismatch in redirect URI. I solved it by updating Authorized Redirect URIs in Google Developer Console to www URL.
Other common URI mismatch are:
Using http:// in Authorized Redirect URIs and https:// as actual URL, or vice-versa
Using trailing slash (http://example.com/) in Authorized Redirect URIs and not using trailing slash (http://example.com) as actual URL, or vice-versa
Here are the step-by-step screenshots of Google Developer Console so that it would be helpful for those who are getting it difficult to locate the developer console page to update redirect URIs.
Go to https://console.developers.google.com
Select your Project
Click on the menu icon
Click on API Manager menu
Click on Credentials menu. And under OAuth 2.0 Client IDs, you will find your client name. In my case, it is Web Client 1. Click on it and a popup will appear where you can edit Authorized Javascript Origin and Authorized redirect URIs.
Note: The Authorized URI includes all localhost links by default, and any live version needs to include the full path, not just the domain, e.g. https://example.com/path/to/oauth/url
Here is a Google article on creating project and client ID.

If you're using Google+ javascript button, then you have to use postmessage instead of the actual URI. It took me almost the whole day to figure this out since Google's docs do not clearly state it for some reason.

In any flow where you retrieved an authorization code on the client side, such as the GoogleAuth.grantOfflineAccess() API, and now you want to pass the code to your server, redeem it, and store the access and refresh tokens, then you have to use the literal string postmessage instead of the redirect_uri.
For example, building on the snippet in the Ruby doc:
client_secrets = Google::APIClient::ClientSecrets.load('client_secrets.json')
auth_client = client_secrets.to_authorization
auth_client.update!(
:scope => 'profile https://www.googleapis.com/auth/drive.metadata.readonly',
:redirect_uri => 'postmessage' # <---- HERE
)
# Inject user's auth_code here:
auth_client.code = "4/lRCuOXzLMIzqrG4XU9RmWw8k1n3jvUgsI790Hk1s3FI"
tokens = auth_client.fetch_access_token!
# { "access_token"=>..., "expires_in"=>3587, "id_token"=>..., "refresh_token"=>..., "token_type"=>"Bearer"}
The only Google documentation to even mention postmessage is this old Google+ sign-in doc. Here's a screenshot and archive link since G+ is closing and this link will likely go away:
It is absolutely unforgivable that the doc page for Offline Access doesn't mention this. #FacePalm

For my web application i corrected my mistake by writing
instead of : http://localhost:11472/authorize/
type : http://localhost/authorize/

Make sure to check the protocol "http://" or "https://" as google checks protocol as well.
Better to add both URL in the list.

1.you would see an error like this
2.then you should click on request details
after this , you have to copy that url and add this on https://console.cloud.google.com/
go to https://console.cloud.google.com/
click on Menu -> API & Services -> Credentials
you would see a dashboard like this ,click on edit OAuth Client
now in Authorized Javascript Origins and Authorized redirect URLS
add the url that has shown error called redirect_uri_mismatch i.e here it is
http://algorithammer.herokuapp.com , so i have added that in both the places in
Authorized Javascript Origins and Authorized redirect URLS
click on save and wait for 5 min and then try to login again

This seems quite strange and annoying that no "one" solution is there.
for me http://localhost:8000 did not worked out but http://localhost:8000/ worked out.

This answer is same as this Mike's answer, and Jeff's answer, both sets redirect_uri to postmessage on client side. I want to add more about the server side, and also the special circumstance applying to this configuration.
Tech Stack
Backend
Python 3.6
Django 1.11
Django REST Framework 3.9: server as API, not rendering template, not doing much elsewhere.
Django REST Framework JWT 1.11
Django REST Social Auth < 2.1
Frontend
React: 16.8.3, create-react-app version 2.1.5
react-google-login: 5.0.2
The "Code" Flow (Specifically for Google OAuth2)
Summary: React --> request social auth "code" --> request jwt token to acquire "login" status in terms of your own backend server/database.
Frontend (React) uses a "Google sign in button" with responseType="code" to get an authorization code. (it's not token, not access token!)
The google sign in button is from react-google-login mentioned above.
Click on the button will bring up a popup window for user to select account. After user select one and the window closes, you'll get the code from the button's callback function.
Frontend send this to backend server's JWT endpoint.
POST request, with { "provider": "google-oauth2", "code": "your retrieved code here", "redirect_uri": "postmessage" }
For my Django server I use Django REST Framework JWT + Django REST Social Auth. Django receives the code from frontend, verify it with Google's service (done for you). Once verified, it'll send the JWT (the token) back to frontend. Frontend can now harvest the token and store it somewhere.
All of REST_SOCIAL_OAUTH_ABSOLUTE_REDIRECT_URI, REST_SOCIAL_DOMAIN_FROM_ORIGIN and REST_SOCIAL_OAUTH_REDIRECT_URI in Django's settings.py are unnecessary. (They are constants used by Django REST Social Auth) In short, you don't have to setup anything related to redirect url in Django. The "redirect_uri": "postmessage" in React frontend suffice. This makes sense because the social auth work you have to do on your side is all Ajax-style POST request in frontend, not submitting any form whatsoever, so actually no redirection occur by default. That's why the redirect url becomes useless if you're using the code + JWT flow, and the server-side redirect url setting is not taking any effect.
The Django REST Social Auth handles account creation. This means it'll check the google account email/last first name, and see if it match any account in database. If not, it'll create one for you, using the exact email & first last name. But, the username will be something like youremailprefix717e248c5b924d60 if your email is youremailprefix#example.com. It appends some random string to make a unique username. This is the default behavior, I believe you can customize it and feel free to dig into their documentation.
The frontend stores that token and when it has to perform CRUD to the backend server, especially create/delete/update, if you attach the token in your Authorization header and send request to backend, Django backend will now recognize that as a login, i.e. authenticated user. Of course, if your token expire, you have to refresh it by making another request.
Oh my goodness, I've spent more than 6 hours and finally got this right! I believe this is the 1st time I saw this postmessage thing. Anyone working on a Django + DRF + JWT + Social Auth + React combination will definitely crash into this. I can't believe none of the article out there mentions this except answers here. But I really hope this post can save you tons of time if you're using the Django + React stack.

In my case, my credential Application type is "Other". So I can't find Authorized redirect URIs in the credentials page. It seems appears in Application type:"Web application". But you can click the Download JSON button to get the client_secret.json file.
Open the json file, and you can find the parameter like this: "redirect_uris":["urn:ietf:wg:oauth:2.0:oob","http://localhost"]. I choose to use http://localhost and it works fine for me.

When you register your app at https://code.google.com/apis/console and
make a Client ID, you get a chance to specify one or more redirect
URIs. The value of the redirect_uri parameter on your auth URI has to
match one of them exactly.

Checklist:
http or https?
& or &?
trailing slash(/) or open ?
(CMD/CTRL)+F, search for the exact match in the credential page. If
not found then search for the missing one.
Wait until google refreshes it. May happen in each half an hour if you
are changing frequently or it may stay in the pool. For my case it was almost half an hour to take effect.

for me it was because in the 'Authorized redirect URIs' list I've incorrectly put https://developers.google.com/oauthplayground/ instead of https://developers.google.com/oauthplayground (without / at the end).

The redirect url is case sensitive.
In my case I added both:
http://localhost:5023/AuthCallback/IndexAsync
http://localhost:5023/authcallback/indexasync

If you use this tutorial: https://developers.google.com/identity/sign-in/web/server-side-flow then you should use "postmessage".
In GO this fixed the problem:
confg = &oauth2.Config{
RedirectURL: "postmessage",
ClientID: ...,
ClientSecret: ...,
Scopes: ...,
Endpoint: google.Endpoint,
}

beware of the extra / at the end of the url
http://localhost:8000 is different from http://localhost:8000/

It has been answered thoroughly but recently (like, a month ago) Google stopped accepting my URI and it would not worked. I know for a fact it did before because there is a user registered with it.
Anyways, the problem was the regular 400: redirect_uri_mismatch but the only difference was that it was changing from https:// to http://, and Google will not allow you to register http:// redirect URI as they are production publishing status (as opposed to localhost).
The problem was in my callback (I use Passport for auth) and I only did
callbackURL: "/register/google/redirect"
Read docs and they used a full URL, so I changed it to
callbackURL: "https://" + process.env.MY_URL+ "/register/google/redirect"
Added https localhost to my accepted URI so I could test locally, and it started working again.
TL;DR use the full URL so you know where you're redirecting

2015 July 15 - the signin that was working last week with this script on login
<script src="https://apis.google.com/js/platform.js" async defer></script>
stopped working and started causing Error 400 with Error: redirect_uri_mismatch
and in the DETAILS section: redirect_uri=storagerelay://...
i solved it by changing to:
<script src="https://apis.google.com/js/client:platform.js?onload=startApp"></script>

Rails users (from the omniauth-google-oauth2 docs):
Fixing Protocol Mismatch for redirect_uri in Rails
Just set the full_host in OmniAuth based on the Rails.env.
# config/initializers/omniauth.rb
OmniAuth.config.full_host = Rails.env.production? ? 'https://domain.com' : 'http://localhost:3000'
REMEMBER: Do not include the trailing "/"

None of the above solutions worked for me. below did
change authorised Redirect urls to - https://localhost:44377/signin-google
Hope this helps someone.

My problem was that I had http://localhost:3000/ in the address bar and had http://127.0.0.1:3000/ in the console.developers.google.com

Just make sure that you are entering URL and not just a domain.
So instead of:
domain.com
it should be
domain.com/somePathWhereYouHadleYourRedirect

Anyone struggling to find where to set redirect urls in the new console: APIs & Auth -> Credentials -> OAuth 2.0 client IDs -> Click the link to find all your redirect urls

My two cents:
If using the Google_Client library do not forget to update the JSON file on your server after updating the redirect URI's.

I also get This error Error-400: redirect_uri_mismatch
This is not a server or Client side error but you have to only change by checking that you haven't to added / (forward slash) at the end like this
redirecting URL list ❌:
https://developers.google.com/oauthplayground/
Do this only ✅:
https://developers.google.com/oauthplayground

Let me complete #Bazyl's answer: in the message I received, they mentioned the URI
"http://localhost:8080/"
(which of course, seems an internal google configuration). I changed the authorized URI for that one,
"http://localhost:8080/" , and the message didn't appear anymore... And the video got uploaded... The APIS documentation is VERY lame... Every time I have something working with google apis, I simply feel "lucky", but there's a lack of good documentation about it.... :( Yes, I got it working, but I don't yet understand neither why it failed, nor why it worked... There was only ONE place to confirm the URI in the web, and it got copied in the client_secrets.json... I don't get if there's a THIRD place where one should write the same URI... I find nor only the documentation but also the GUI design of Google's api quite lame...

I needed to create a new client ID under APIs & Services -> Credentials -> Create credentials -> OAuth -> Other
Then I downloaded and used the client_secret.json with my command line program that is uploading to my youtube account. I was trying to use a Web App OAuth client ID which was giving me the redirect URI error in browser.

I have frontend app and backend api.
From my backend server I was testing by hitting google api and was facing this error. During my whole time I was wondering of why should I need to give redirect_uri as this is just the backend, for frontend it makes sense.
What I was doing was giving different redirect_uri (though valid) from server (assuming this is just placeholder, it just has only to be registered to google) but my frontend url that created token code was different. So when I was passing this code in my server side testing(for which redirect-uri was different), I was facing this error.
So don't do this mistake. Make sure your frontend redirect_uri is same as your server's as google use it to validate the authenticity.

The main reason for this issue will only come from chrome and chrome handles WWW and non www differently depending on how you entered your URL in the browsers and it searches from google and directly shows the results, so the redirection URL sent is different in a different case
Add all the possible combinations you can find the exact url sent from fiddler , the 400 error pop up will not give you the exact http and www infromation

Try to do these checks:
Bundle ID in console and in your application. I prefer set Bundle ID of application like this "org.peredovik.${PRODUCT_NAME:rfc1034identifier}"
Check if you added URL types at tab Info just type your Bundle ID in Identifier and URL Schemes, role set to Editor
In console at cloud.google.com "APIs & auth" -> "Consent screen" fill form about your application. "Product name" is required field.
Enjoy :)

ng-token-auth, ionic, devise_token_auth; token get lost randomly in xhr request

I have inherited a Ionic app which uses ng-token-auth+devise_token_auth to handle the authentication and the session between front and back.
What happens is quite strange. Sometimes (specially with slow connections) the request (or the response) get lost and after that I get only 401 http errors.
I know that that everytime I send a request the token expires, but when the xhr request is cancelled (by the server I suppose, or by the browser, I don't know) the token is expired without having been replaced by the new one generated by devise_token_auth gem.
I know Rails but I'm not familiar with Angular, neither Ionic and I don't know exactly where to look.
After reading a lot of SO answers where noone seems having my problem (which happens locally and in staging/production), I checked the following
storage is set as localStorage.
config.batch_request_buffer_throttle = 20.seconds
there is no pattern between cancelled requests, sometimes I perform get for the username, sometimes a post or a put to a comment.
Is not a CORS problem because it would happen always or never. (moreover I'm using a proxy as explained in ionic blog)
Maybe it could be related to provisional headers chrome bug. But, how can I be sure?
What puzzles my is that it happens only sometimes and not always. (and there are no errors in the backend)
The only workaround I have found in the devise_token_auth documentation is change config.change_headers_on_each_request to false avoiding in this way the regeneration of the token.
But I don't like this solution because I think it hides the real problem in an insecure way instead of solving the token loss. Any suggestion?

Kindly, please check this thing:
Version: which version of this gem (and ng-token-auth, jToker or Angular2-Token if applicable) are you using?
Request and response headers: these can be found in the "Network" tab of your browser's web inspector.
Rails Stacktrace: this can be found in the log/development.log of your API.
Environmental Info: How is your application different from the reference implementation?
This may include (but is not limited to) the following details:
Routes: are you using some crazy namespace, scope, or constraint?
Gems: are you using MongoDB, Grape, RailsApi, ActiveAdmin, etc.?
Custom Overrides: what have you done in terms of [custom controller overrides]
5?
Custom Frontend: are you using ng-token-auth, jToker, Angular2-
Token, or something else?

is it possible to intercept the response to an HTTP OPTIONS preflight in AngularJS?

I'm trying to implement a simple interceptor that allows me to display a message along the lines of "cannot contact the server" in my Angular app. However as the API is on a different host I'm dealing with CORS pre-flight OPTIONS requests.
I've found that if the API is unavailable Chrome dev tools shows a 503 on the OPTIONS request but Angular's $http interceptor catches a 404 response to the subsequent GET request. I believe this is because the OPTIONS response did not contain the required CORS headers so the GET is actually never performed.
Is is possible to intercept the OPTIONS response? If all I see is a 404 I can't distinguish "server down" from "no such resource".

You can't intercept this request by design - the browser is "checking up" on you, making sure YOU should be allowed to make the request.
We've used three solutions to work around this:
If the problem is that you're using a development environment like NodeJS, and your domain names aren't matching (that is, if you normally wouldn't need to deal with this in Production) you can use a proxy. The https://github.com/substack/bouncyBounceJS NodeJS Module is an easy to use option. Then your Web service request domain will match the domain your page is on, and the check won't be triggered. (You can also use tricks like this in Production, although it can be easily abused!)
Also for temporary use, you can use something like Fiddler or Charles to manipulate the request by faking the required headers, or tell your browser not to check them (--disable-web-security in Chrome).
If you have this problem in Production, you either need to legitimately fix it (adjust the Web service handler to add the required headers - there are only two), or find a way to make the request in a way that doesn't trigger the check. For instance, if you control both the source and target domains, you can put a script on the target that makes the requests to itself. Run this in an IFRAME, invisibly. Then you can use things like postMessage() to communicate back and forth. Large services like Facebook use "XHR bridges" like this for the same reason.

Can I change the welcome page depending on URL?

I have a single GAE app, which serves two different variations of the application depending on the invoked URL. Is it possible to have a per-url welcome page, without having the request map to a servlet which would slow the page load too much.
So, by way of example...
My app lives at myapp.appspot.com
I have two domains mapped to this app, myapp1.com and myapp2.com
If the user accesses myapp1.com, I want to serve myapp1-index.html, and
if he accesses myapp2.com, I want to serve myapp2-index.html

I never played with GAE on Java but I'm guessing you could do something based on the requested URL. I believe you can check the requested URL with the following inside your request:
httpServletRequest.getRequestURI()
Where httpServletRequest is the HttpServletRequest parameter from your doGet() method.

Why does Google App Engine append a path to my 'continue' location during login?

I'm using a very simple GAE instance from a Greasemonkey script. This worked fine for the last months, but now a path is appended to the final 'continue' location, which breaks the login process for me.
The basic workflow, under the assumption that the user is logged into his Google Account, but his token for the GAE instance has timed out:
User opens page A with the GM script installed.
The GM script runs and tries to access the GAE instance with a GM_xmlhttpRequest().
The GAE instance returns "login_needed|<loginurl>". The GM script extracts the loginurl and sets window.location on it.
The user is redirected to the loginurl and eventually back to A. However, this time, actual data is returned by the GM_xmlhttpRequest().
The last step no longer works, as the user is now redirected to the loginurl plus some, which gives a 404 on the target site.
The GAE code is just about half a screen of code. The authentication relevant code is this:
if not users.get_current_user():
self.response.headers['Content-Type'] = 'text/plain'
self.response.out.write('login_needed|'+users.create_login_url(self.request.get('uri')))
The sequence of requests is as follows, all caused by redirects:
GET https://mygaeinstance.appspot.com/?uri=https://targetsite.com/
GET https://www.google.com/accounts/ServiceLogin?service=ah&passive=true&continue=https://appengine.google.com/_ah/conflogin%3Fcontinue%3Dhttps://targetsite.com/&ltmpl=gm&ahname=MyGAEInstance&sig=<some sig>
GET https://appengine.google.com/_ah/conflogin?continue=https%3A%2F%2Ftargetsite.com%2F&pli=1&auth=<some base64 auth token>
GET https://targetsite.com/_ah/conflogin?state=<some base64 state>
targetsite.com doesn't like that path and as you can see, it wasn't in the initial 'continue' argument passed to appengine.google.com, which was just "https://targetsite.com/". What did I do wrong and how can I fix this?

A recent change to our login flow for App Engine has created an issue whereby a login with a continue URL that's outside the app's own domain will result in an erroneous redirect such as the one you're observing.
We're working on fixing this. In the meantime, a workaround is to set up a redirect handler on your own app. Make that the target of the continue parameter, and have it send a final redirect to your actual target.

This redirect is caused by an expired auth token. To make it work again, you need to invalidate the token on the client, as described here: What is the proper URL to get an Auth Cookie from a GAE based Application

Resources