[dev] [commits] Horde branch master updated. 6473f788bd06787aecea9fbce3cde8ea6f3491ea

[Jan Schneider]

Zitat von Thomas Jarosch <thomas.jarosch at intra2net.com>:

> Hi,
>
> On Tuesday, 10. September 2013 12:51:08 you wrote:
>> commit 6473f788bd06787aecea9fbce3cde8ea6f3491ea
>> Author: Jan Schneider <jan at horde.org>
>> Date:   Tue Sep 10 14:50:52 2013 +0200
>>
>>     [jan] Don't ping server before each query in PDO drivers (Bug #12656).
>>
>>  framework/Db/lib/Horde/Db/Adapter/Pdo/Base.php |    2 --
>>  framework/Db/package.xml                       |    2 ++
>>  2 files changed, 2 insertions(+), 2 deletions(-)
>>
>> http://git.horde.org/horde-git/-/commit/6473f788bd06787aecea9fbce3cde8ea6f
>> 3491ea
>
> while this change is a very good idea, I think it should be mentioned
> in the ChangeLog that the reconnect() call was removed. Some users might
> be unhappy about the silent removal of a safety net.
>
>
> Chuck added the code in 86a3abf06cb476637e25b8612c7424fa600f44de
> --------------
> "Don't consider the connection active if the connection object
> is missing/false, and try reconnecting if the connection
> isn't active when execute() is called"
> --------------
>
>
> Probably the side-effect of doing an additional SELECT
> for each execute() call went unnoticed back then.
>
>
> Long running scripts (ActiveSync?) might bail out badly
> if f.e. the SQL server is quickly reconfigured and
> therefore the DB connection socket was closed.
> Or think about a TCP idle timeout.
>
> BR,
> Thomas

Is this the real world issue, i.e. how often is a database server  
really going away during a request? The original patch to remove the  
ping included an additional re-try loop if a statement failed, but  
only incompletely implemented, thus I left it out.
If people feel like we should really have this check (it only ever  
existed in the PDO backends anyway), the we should properly implement  
such a check-and-retry.
-- 
Jan Schneider
The Horde Project
http://www.horde.org/