fix: use userinfo() with okta and don't try decoding a hash as json#2292
fix: use userinfo() with okta and don't try decoding a hash as json#2292Lewiscowles1986 wants to merge 3 commits intodpgaspar:masterfrom
Conversation
|
Am not a contributor to this repo but by coincidence I've also been trying to trace this issue today. I believe this PR would fix a non-obvious "gotcha" when configuring a provider with the name In fact, it already works just fine if you name your provider something other than To continue down this train of thought - I think The mappings of the retrieved data to flask_appbuilder's user info values needn't be hardcoded either - some organizations might wish to customize them. It might be better to provide them as overridable defaults for each provider. Allowing the values to be extracted from the ID token claims (if present) could also prevent the need for a round trip to the userinfo endpoint altogether. |
|
@jtv8 while this feels related, the problem I am solving actually relates to the Apache Superset usage of this library, where additional conventions have been setup around this. We've actually gone old-school and manually pulled and patched the If you look at the auth0 just below okta definition, we are essentially borrowing from another of Okta's services (Auth0) conventions, to get this to work. Server metadata url is not required by the way, you could also define the user info endpoint url, but I like that I can set the base uri and the metadata url and be more or less done. |
|
@dpgaspar any chance of 👀 on this? |
Description
This plainly doesn't work with Okta, and it is not so hard to get it to work with Okta... There is no custom security class, no hacks to rely on identity token, just using the metadata url, to do what it should.
Fixes #2291
ADDITIONAL INFORMATION
I Would love it if someone can cut a release once this merges, so that I can build a superset image based upon this.