1465085 - 500 ISE when creating a document that is the child of a redirect with a full URL

Status: RESOLVED FIXED

(developer.mozilla.org Graveyard :: Editing, defect, P5)

Description

[John Whitlock [:jwhitlock]]

What did you do?
================
1. Log into MDN Web Docs as user with page creation permissions
2. Go to https://developer.mozilla.org/en-US/docs/API/Window

What happened?
==============
An Internal Server Error

What should have happened?
==========================
New page creation should start, with https://developer.mozilla.org/en-US/docs/Web/API as parent document.

Is there anything else we should know?
======================================
Part of the bug fix for bug 1052510 included resolving parent redirects as part of new page creation. This code assumes that only relative references are used in valid redirects, such as:

/en-US/docs/API/Window

The redirect for this document (possibly created manually) is a full URL:

https://developer.mozilla.org/en-US/docs/API/Window

The code checks for full URLs and refuses to parse them, returning False:

https://github.com/mozilla/kuma/blob/1bed365b259351e93faab4e914b2ed77f6fa2dae/kuma/wiki/utils.py#L60-L67

The new document code does not handle the case where the following the redirect sets the parent to None, resulting in the exception:

https://github.com/mozilla/kuma/blob/1bed365b259351e93faab4e914b2ed77f6fa2dae/kuma/wiki/views/document.py#L574-L578

Tracked in sentry:

Public: 

https://sentry.prod.mozaws.net/share/issue/37332e34333531393130/

Staff:

https://sentry.prod.mozaws.net/operations/mdn-prod/issues/4351910/

Comment 1

[John Whitlock [:jwhitlock]]

Fixed for this case by manually updating https://developer.mozilla.org/en-US/docs/API/Window. Looking to see how many additional pages have this issue.

Comment 2

[John Whitlock [:jwhitlock]]

254 redirect documents include full URLs. Some possible fixes:

1) Adjust the parent document redirect following to allow full URLs on developer.mozilla.org.
2) Update Document.get_redirect_url() to return the relative reference.
3) Update the 254 documents manually to use relative URLs.
4) Eliminate the issues by re-implementing redirects (bug 1364709).

Comment 3

[Safwan Rahman (:safwan)]

I strongly believe the 2 number point is most acceptable solution because the document should have relative url instead of full url.
But it may need some time to write a script or do it manually.

In the meantime, I am going to submit a patch which will eliminate it.

Comment 4

[[github robot]]

Commits pushed to master at https://github.com/mozilla/kuma

https://github.com/mozilla/kuma/commit/616dda0a1bed25fd5107826ef918957242347dc8
bug 1465085: Tests for get_doc_components_from_url

Refactor the tests for get_doc_components_from_url to pytest style, and
expand the test cases.

https://github.com/mozilla/kuma/commit/1c49677006f52a2a7c6f1d6773d8ac9d50fcfda2
bug 1465085: Refactor some Document model tests

Refactor DocumentTests to pytest style, and expand some tests cases.

https://github.com/mozilla/kuma/commit/f7596e4cd3572b49a79d8950a78d83407e275a0e
bug 1465085: Parse full URL for local domains

In get_doc_components_from_url with check_host=True, continue parsing
full URLs on the local domain.

https://github.com/mozilla/kuma/commit/29cda03dbbe83a14efe336411739798e33f25dc0
bug 1465085: Return the path from a full URL

Redirect URLs are allowed to be the full URL to a local document.
Process these and return just the path.

https://github.com/mozilla/kuma/commit/f264fe85288cb6fb019423a2c27c44977309ec51
Merge pull request #4863 from safwanrahman/redirect_fix

bug 1465085: Handle a child of a redirect with a full URL without raising an error

Comment 5

[John Whitlock [:jwhitlock]]

Deployed to production, https://developer.mozilla.org/en-US/docs/API/Window now starts the Create a New Article workflow.

Comment 6

[John Whitlock [:jwhitlock]]

I was a little quick to close this one. I'd fixed /API/Window by fixing the redirect, so it wasn't a good test.

This document has a standard redirect:

https://developer.mozilla.org/en-US/docs/Mozilla/Projects/SpiderMonkey/JSAPI_reference/JS_GetPropertyByIdDefault

to

https://developer.mozilla.org/en-US/docs/Mozilla/Projects/SpiderMonkey/JSAPI_reference/JS_GetPropertyDefault

and creating a new subpage works:

https://developer.mozilla.org/en-US/docs/Mozilla/Projects/SpiderMonkey/JSAPI_reference/JS_GetPropertyByIdDefault/SubPage


There are several non-standard redirects. For example, this URL:

https://developer.mozilla.org/en-US/Extensions

is a redirect to:

https://developer.mozilla.org/en-US/Add-ons

and raises a 500 error when you view:

https://developer.mozilla.org/en-US/Extensions/SubPage

This is also a redirect to a zoned URL, which is scheduled for removal in bug 1462475.

I'm going to see how many "non-standard" redirects there are that remain problems, to see if it is useful to write more code or manually fix them.

Comment 7

[John Whitlock [:jwhitlock]]

Attachment: bad_redirects.csv

There are 74 bad redirects remaining, or 30% of the original 254. This seems like it could be fixed quickly.

Comment 8

[John Whitlock [:jwhitlock]]

I manually fixed 66 of the redirects. There are 8 remaining that are redirects to non-documents. For example, https://developer.mozilla.org/en-US/docs/Main_page redirects to the homepage. Going to a subpage of one of these redirects (https://developer.mozilla.org/en-US/docs/Main_page/SubPage) still results in a 500 ISE when logged in. Instead, we should return a 404.

Comment 9

[John Whitlock [:jwhitlock]]

Opened https://github.com/mozilla/kuma/pull/4887 to fix issue. Tracked in sentry:

https://sentry.prod.mozaws.net/share/issue/37332e34333838303638/

Comment 10

[[github robot]]

Commits pushed to master at https://github.com/mozilla/kuma

https://github.com/mozilla/kuma/commit/73edca756852b4e176981d1c569dfa5e1cabb4cc
bug 1465085: Return 404 for child of non-doc 301

If /en-US/docs/Main_page is a redirect to a non-Document MDN URL, then
return a 404 to /en-US/docs/Main_page/SubPage when logged in, rather
than raise an Internal Server Error (500 ISE).

https://github.com/mozilla/kuma/commit/4b5577b4cb52518b9f290337f741fbb6e5936178
Merge pull request #4887 from jwhitlock/subpage-404-1465085

bug 1465085: Return 404 for the child of a 301 to a non-document URL

Comment 11

[John Whitlock [:jwhitlock]]

Fix deployed to production, https://developer.mozilla.org/en-US/docs/Main_page/SubPage is now 404.