Bugs item #2983860, was opened at 2010-04-08 17:00
Message generated for change (Comment added) made by shcheklein
You can respond by visiting:
https://sourceforge.net/tracker/?func=detail&atid=713730&aid=2983860&group_id=129076
Please note that this message will contain a full copy of the comment thread,
including the initial issue submission, for this request,
not just the latest update.
Category: Query Execution
>Group: Sedna (current)
Status: Open
Resolution: Accepted
>Priority: 7
Private: No
Submitted By: Yanko Yankulov (core-dumped)
>Assigned to: Alexander Kalinin (allexk)
Summary: seg fault on module & trigger
Initial Comment:
import module namespace ps = "www.packetscale.com/test1";
CREATE TRIGGER "t1" BEFORE INSERT ON doc("someDoc")/* FOR EACH NODE DO {
ps:error(); } &
This statement is accepted by the engine, but the static context holding
module functions is lost after its end, leaving behind PPFunCall object holding
fn_id to destroyed static context. When executing an UPDATE triggering the
----------------------------------------------------------------------
>Comment By: Ivan Shcheklein (shcheklein)
Date: 2010-05-05 16:47
Message:
Yanko, sorry for the delay in answer. Unfortunately, the team is working
now on other features. I believe, that it will be posssible for us to fix
this issue in month or two.
Probably, alternative solution you described is possible, however I think
it's also not that simple to implement.
----------------------------------------------------------------------
Comment By: Yanko Yankulov (core-dumped)
Date: 2010-04-08 18:48
Message:
Thanks for the fast reply Ivan.
I suspected that a full refactoring of the contexts will be necessary to
properly handle this issue. Do you have any idea about the expected
timeline for this to happen ?
As an alternative/partial/temporary solution I was thinking to implement a
default prolog witch will be added to all queries for a given database.
What do you/the community think about this. We at packetScale are
definitely interested in using user functions in triggers, and we might be
able to dedicate some man hours in doing this. Any
comments/objections/comments about this idea?
----------------------------------------------------------------------
Comment By: Ivan Shcheklein (shcheklein)
Date: 2010-04-08 18:14
Message:
Yanko, thank you for bug report! This is a known issue with Sedna:
http://thread.gmane.org/gmane.text.xml.sedna/1454 . It requires full
refactoring of dynamic and static contexts.
----------------------------------------------------------------------
Comment By: Yanko Yankulov (core-dumped)
Date: 2010-04-08 17:06
Message:
pressed some buttons too early :) ...
When executing an UPDATE causing the trigger to fire, fn_id is looked up
in current context which hodls undefined function at this fn_id.
Same problem must exist for other static context members -> vars,
namespaces, etc.
Workaround: There's an ugly workaround for this -> use always the same
prolog for every query.
----------------------------------------------------------------------
You can respond by visiting:
https://sourceforge.net/tracker/?func=detail&atid=713730&aid=2983860&group_id=129076
------------------------------------------------------------------------------
_______________________________________________
Sedna-discussion mailing list
Sedna-discussion@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/sedna-discussion
