Tickets :: [#2749] bugfix (and whitespace fixes) for lib/Horde/SessionHandler/pgsql.php

6.0.0-git

2021-01-18

Summary	bugfix (and whitespace fixes) for lib/Horde/SessionHandler/pgsql.php
Queue	Horde Framework Packages
Queue Version	FRAMEWORK_3
Type	Bug
State	Resolved
Priority	2. Medium
Owners
Requester	cbs (at) cts (dot) ucla (dot) edu
Created	2005-10-09 (5580 days ago)
Due
Updated	2005-10-15 (5574 days ago)
Assigned	2005-10-11 (5578 days ago)
Resolved	2005-10-12 (5577 days ago)
Milestone
Patch	No

2005-10-15 03:30:47	cbs (at) cts (dot) ucla (dot) edu	Comment #6	Reply to this comment
Committed, thanks. An extra BEGIN crept into my last diff. Line 106 currently reads $query = sprintf('BEGIN; SELECT session_data FROM %s WHERE ' . but should read $query = sprintf('SELECT session_data FROM %s WHERE ' .

2005-10-12 17:53:30	Chuck Hagenbuch	Comment #5 State ⇒ Resolved	Reply to this comment
Committed, thanks.

2005-10-12 15:23:17	Chuck Hagenbuch	Deleted Original Message

2005-10-12 07:50:18	cbs (at) cts (dot) ucla (dot) edu	Comment #4 New Attachment: pgsql.php.diff	Reply to this comment
The attached diff is against the latest framework HEAD snapshot.

2005-10-11 22:00:21	Chuck Hagenbuch	Taken from Horde Developers Taken from Chuck Hagenbuch

2005-10-11 22:00:05	Chuck Hagenbuch	Comment #3 State ⇒ Feedback	Reply to this comment
This patch does not apply cleanly to either HEAD or FRAMEWORK_3 CVS. Please re-do the patch against HEAD and we'll get it backported once it's tested.

2005-10-10 19:06:51	Chuck Hagenbuch	Deleted Original Message

2005-10-09 22:21:36	cbs (at) cts (dot) ucla (dot) edu	Comment #2 New Attachment: pgsql.php[1].diff	Reply to this comment
I have a better patch that more closely approximates what mysql.php is doing. This patch modified read() to only do a SELECT FOR UPDATE on the session id. If the row exists, this will lock it for the duration of the transaction. It changes write() to do a SELECT on the session id to determine if there are any rows with that id. If there aren't any, it does an INSERT. If there is a row it does an UPDATE. This is more closely akin to the MySQL REPLACE INTO from mysql.php. I am now able to authenticate and get a session. Turba, Gollem, Kronolith, and Nag are all functioning normally as far as I can tell. I am having problems with Imp -- trying to select a message or change folders (anything that involves imp/message.php, imp/mailbox.php, imp/search.php) returns me to imp/mailbox.php?mailbox=INBOX&actionID=login

2005-10-09 22:10:26	Chuck Hagenbuch	Assigned to Chuck Hagenbuch Assigned to Horde Developers State ⇒ Assigned

2005-10-09 10:36:31	cbs (at) cts (dot) ucla (dot) edu	Comment #1 Type ⇒ Bug State ⇒ Unconfirmed Priority ⇒ 2. Medium Summary ⇒ bugfix (and whitespace fixes) for lib/Horde/SessionHandler/pgsql.php Queue ⇒ Horde Framework Packages New Attachment: pgsql.php.diff	Reply to this comment
The write() handler in pgsql.php is checking the result of the COMMIT, not the result of the UPDATE. The current check tests pg_rows_affected(), which is 0, and always results in true. The test should (I think) be checking result of the UPDATE to ensure that 1 row was updated. See the thread on horde@lists with the subject "horde 3.0.5, postgresql session handler fails" from Oct. 8 for more details.