[Date Prev][Date Next] [Chronological] [Thread] [Top]

RE : RE : RE : RE : Crash on entry modification in back-sql/unixODBC/oracle (ITS#2641)

I have tested the correction in entry-id.c : OK
Then I have tested the correction in delete.c : the crash does not occur any
Still another error occurs : when performing the SQLExecute(), the driver
replies : invalid character.
Then I modified the metadata to change the create_proc and delete_proc from
{blabla} to blabla, assuming the "invalid character" was the bracket
Then I re-tested : it worked ! I succeeded in deleting one person. And it
correctly removed the deleted author from the book entries: great.
So the oracle samples needs to be modified to suppress the brackets
surrounding the stored procedure calls.

-----Message d'origine-----
De : Pierangelo Masarati [mailto:ando@sys-net.it] 
Envoyé : mardi 16 mars 2004 11:04
À : emmanuel.duru@atosorigin.com
Cc : openldap-its@OpenLDAP.org
Objet : Re: RE : RE : RE : Crash on entry modification in
back-sql/unixODBC/oracle (ITS#2641)

> In my case delete fails before crashing because backsql_count_children
> returns an error because the SQL request returns "0.0000000000" (and not
> 0) and the test after strtol at the end of the function fails.

I'll have a look at this.  Is this a RDBMS related issue?
I mean: the query looks like

SELECT COUNT(distinct subordinates.id) from ...

I do not expect COUNT to return a fractional number; is it reasonable to
have fractional counts of discrete pieces of information?  I've committed
a hack in HEAD, can you go on testing?  Thanks, p.

Pierangelo Masarati