Closed (duplicate)
Project:
Atlassian Crowd SSO
Version:
7.x-2.x-dev
Component:
Code
Priority:
Normal
Category:
Bug report
Assigned:
Reporter:
Created:
28 Oct 2010 at 19:42 UTC
Updated:
12 Apr 2013 at 22:50 UTC
Jump to comment: Most recent file
Comments
Comment #1
ben.bunk commentedA Patch for the install file.
Comment #2
ebeyrent commentedI had originally weighted the Crowd module for a reason, which escapes me atm. I'm concerned that setting the module weight in the installer probably wasn't such a great idea, as it clearly doesn't work for all use cases.
I'll poke through my notes to see if I can recall what prompted me to set the weight in the first place.
Comment #3
ben.bunk commentedFor a memory jog: Did this have anything to do with how the ldap modules integrate with crowd? I haven't tested this to verify but it seems possible.
Comment #4
ebeyrent commentedMost likely... :o)
Comment #5
greg2010 commentedI am experiencing a similar issue using Crowd with numerous sub-sites in an Aegir platform. I have set the crowd module weight to -1001 in the system table. It has the lowest weight in the table. When users leave the parent site page to view a sub-site, it does not automatically keep them logged in.
Comment #6
ben.bunk commentedGreg2010 your issue Sounds more like a cookie domain problem then this issue
Comment #7
rjacobs commentedI can confirm that this is also an issue with the 6.x-2.x and 7.x-2.x branches (which are the much more active branches at the moment). I'm going to mark this against those v2 branches and take a look.
Comment #8
rjacobs commentedSo this problem is probably due to the order in which hook_init calls are made. I suppose that Drupal, or some other modules, may force an access denied error before the Crowd module can finish it's SSO business. Because the module weight is so high at the moment (meaning later execution relative to other implementations of hook_init), pretty much all the other init stuff is happening before crowd can do anything. In addition, crowd sso checks never happen for cached page requests, which further limits crowd's ability to do it's SSO magic on all requests.
Anyway, I think that all this will be addressed once a fix for the following issue lands:
#1968994: SSO checks should probably run in hook_boot
Fixing that will mean that the SSO checks will happen in hook_boot, which will always be before hook_init. For this reason, I'm going to mark this as a dup of #1968994.