URL fragment doesn't persist through server-side redirect on iOS

[boonebgorges]

See #79. That ticket fixed the issue authenticated POSTs, but not for those that need to go through wp-login.php. I'll need to build some sort of JS-based replacement for wp-login's redirect_to behavior.

[moui72]

let us know when/if this needs testing

[boonebgorges]

Ran out of time for this one. It's extremely difficult to debug.

[boonebgorges]

I've determined that this is a more general Safari problem, affecting both iOS and OSX. WP is never receiving a POST request to begin with, which, at least on the dev site, is a regression from #79.

I have a suspicion that it's a problem only for our dev site, because of the way that Safari handles form submissions that contain HTTP authentication. @drdrew42 - questions for you:

1. Have you heard of anyone having redirect issues when using Safari in real use of OL-WW?
2. Can you share credentials to one of the courses on the production site? I see there's an "OpeningGateways" course. Send them to me privately boone@gorg.es. If it looks like this is specific to openlabdev.org, there may be nothing to do here.

[boonebgorges]

@drdrew42 - thanks for sending the credentials! I can now confirm that the POST/GET problem is specific to the dev server (so no problem with #79 on production) but the URL fragment is still being lost. This gives me enough info to debug.

[boonebgorges]

@drdrew42 - thanks for sending the credentials! I can now confirm that the POST/GET problem is specific to the dev server (so no problem with #79 on production) but the URL fragment is still being lost.

I just put in place a kinda insane but, apparently, working fix. Here's the high-level rundown:

• On wp-login.php, if it's detected that the problemId info is included in the redirect_to parameter, the problemId is stored in a cookie
• When loading a logged-in user page, if the problemId cookie is detected, then (a) that cookie is deleted (so the redirect happens just once), and (b) you're redirected to the correct problemId URL.

It's not currently possible to test this on openlabdev.org because of the weird behavior of HTTP auth + redirects/form submits + Safari.

However, I've monkey-patched the production server for the purposes of testing, and since it works, I'm going to leave it in place. @bree-z and @moui72 - the testing process here is, on Safari (mobile or otherwise):

1. to go to the production testing site (ask @drdrew42 for creds if you don't have them)
2. Make sure you're not logged into the OL (production)
3. Ask for Help
4. You should see the OL wp-login.php screen.
5. Look at the URL. It should contain a long ?redirect_to parameter that includes the string problemId as well as a lot of other stuff. If you see a shorter redirect_to, something else has gone wrong, and you may want to try a different source problem on WW (Safari is extremely finicky)
6. If you want, check your browser at this point for a webwork-problem-id cookie
7. After login, you should be redirected to the proper problemId URL

[bree-z]

@drdrew42 could you please send creds (and URL, just to be sure I'm testing the right thing) to work@bree.bz?

thanks!

[drdrew42]

I believe the course you're looking for is http://mathww.citytech.cuny.edu/webwork2/MAT1275/

I'll email the credentials.

[bree-z]

Thanks @drdrew42!

[bree-z]

I think this is working for me. I went to the site above in Safari, and after clicking 'Ask for Help' the login page opened with following URL:

https://openlab.citytech.cuny.edu/ol-webwork/wp-login.php?redirect_to=https%3A%2F%2Fopenlab.citytech.cuny.edu%2Fol-webwork%3Fpost_data_key%3Dwebwork_post_data_bfc4d4bde72ce1abf49141c912e246f0%23%3AproblemId%3Dlocal%2FReducingRationalExpressions%2FmonoDenom-NS.pg

After logging in, I was on the correct problem (ReducingRationalExpressions: Problem 1), with this URL:

https://openlab.citytech.cuny.edu/ol-webwork/?post_data_key=webwork_post_data_bfc4d4bde72ce1abf49141c912e246f0#:problemId=local/ReducingRationalExpressions/monoDenom-NS.pg

I also appear to have a webwork-problem-id cookie.

If this sounds good I think we can close this. Thanks!

[boonebgorges]

Yes, this sounds good. Thanks for verifying, @bree-z !