[Pidgin] #9034: Pidgin won't let me block people

[Pidgin]

#9034: Pidgin won't let me block people
----------------------+-----------------------------------------------------
 Reporter:  jewel460  |        Owner:  MarkDoliner
     Type:  defect    |       Status:  new        
Milestone:            |    Component:  AIM        
  Version:  2.5.5     |   Resolution:             
 Keywords:            |  
----------------------+-----------------------------------------------------

Comment(by dustin):

 Some more observations...


 ''Tests done using wireshark''

 1. SSI item list read in on login.

   The SSI item list that is read in by ICQ7.2 or qip contains the same
 entries as libpurple's list.

   (Adding a new entry fails when trying to use certain buddy IDs.
   If this were because of a failure in libpurple when reading in the SSI
 list (missing items), those items would appear in the SSI list of other
 clients.)


 2. Changing AIM user info from within pidgin

   After changing the user info, there is no AIM SSI update from the server
 (which would indicate that there is also a new/changed entry put into the
 SSI list).

   (Changing the user info is how I could reproduce broken/invisible SSI
 entries, see below. If the server changes the SSI list because of this,
 the client would be informed of it... and a failure to receive the update
 would have been a possible cause of the bug.)


 ''Furthermore:''

 3. Reproducing the bug by changing AIM user info in Pidgin.

   When I delete the broken entries on the server and put something into
 the user info field, two broken entries appear again.

   They occupy two low bids (often consecutively), and therefore are likely
 to cause a collison with later entries.

   Unfortunately, somehow the position of earlier entries seems to be
 memorized. So after deletion the new broken entries may have higher bid
 numbers (starting just above the deleted entries or above a former regular
 invisible/blocking entry that was at this position).


 4. Adding new entries.
   I usually found only the two broken entries already mentioned above.

   I added fake entries going up to bid=0x1100, without observing further
 broken entries.
   So it seems the entries stay at the lower end (but may climb to higher
 numbers, as already mentioned).
   This would also mean that collisions may be avoided by only using higer
 bid numbers.

-- 
Ticket URL: <http://developer.pidgin.im/ticket/9034#comment:35>
Pidgin <http://pidgin.im>
Pidgin