Fix for issue #883 #889
Fix for issue #883 #889
Conversation
|
The behaviour after this fix seems normal. 48 PwWorkflows have been launched (not finished yet) that each of them has sub-workflows and I don't see the previous problems. |
| @@ -181,3 +180,49 @@ def test_session_update_and_expiration_4(self): | |||
| code.store() | |||
|
|
|||
| self.drop_connection() | |||
|
|
|||
There was a problem hiding this comment.
I would also add a test without any SQL specific command (in this case expunge_all etc), as this is what the user would do (this could be a generic test) that would have triggered the error before.
There was a problem hiding this comment.
This is a good point.
The problem is that I have not managed to find until today (and it has been a while that I investigate this) a way to isolate the problem at that level (which more or less implies to follow the DbNode object generation in @nmounet workflows - that run by a Daemon that I don't control by my debugger).
Maybe an idea is if I can easily advance the steps of the daemon manually to emulate the daemon steps and see if I have the same behaviour. And if it is true to check what happens at every step with the session and the dbnodes. But this will not be a 10-20 mins job.
There was a problem hiding this comment.
Ok. For now, then, at least add the same identical test (but without any sqla command) as a general test. Then for me we can merge. Thanks!
There was a problem hiding this comment.
OK, I will.
Just keep in mind that it will be a test that doesn't test anything. I.e. that test would not fail even with the problematic code of set_value() that was corrected.
There was a problem hiding this comment.
I created the test but I did not push it yet. Let's have a short chat before I add it to this PR.
There was a problem hiding this comment.
ok, I see. Let's merge this, then.
(Fix for issue #883)
It seems to run with one WF. It would be good to launch many before merging it.
But in any case, it is an improvement comparing to the previous case