Shibboleth Developers

[giacomo tenaglia <>]

On Tue, Feb 26, 2008 at 02:24:01PM -0500, Scott Cantor wrote:
> I looked again...this was a bug in the updateRow method, now fixed. However,
> it does demonstrate that updateRow is actually getting called by the SP
> under normal conditions, which I didn't realize.

Every error I've had was a "storing back mapping of NameID for logout",
in which updateText (and consequentely updateRow) is called:

2008-02-26 08:59:54 ERROR XMLTooling.StorageService [363]: error searching 
for (t=texts, c=NameID, 
)
2008-02-26 08:59:54 ERROR XMLTooling.StorageService [363]: ODBC Error: 
HY000:1:1756:[Oracle][ODBC][Ora]ORA-01756: quoted string not properly 
terminated
2008-02-26 08:59:54 ERROR Shibboleth.SessionCache [363]: error storing back 
mapping of NameID for logout: ODBC StorageService search failed.

2008-02-26 16:52:45 ERROR XMLTooling.StorageService [335]: error searching 
for (t=texts, c=NameID, 
)
2008-02-26 16:52:45 ERROR XMLTooling.StorageService [335]: ODBC Error: 
HY000:1:1756:[Oracle][ODBC][Ora]ORA-01756: quoted string not properly 
terminated
2008-02-26 16:52:45 ERROR Shibboleth.SessionCache [335]: error storing back 
mapping of NameID for logout: ODBC StorageService search failed.

So I think you don't have to worry about this.

> Which matters because that's where I had to keep auto-commit off, so it's
> obvious that we'll have to deal with the Oracle error as we discussed and
> retry the statements somehow. Maybe if I try hard I can get a goto statement
> into this code...

Since the enabling of auto-commit, I'm getting the serialization error
we discussed about once per day, and always on createRow.
I'm still studying Oracle concurrency/consistency issues, I'm sure
there's a way to avoid putting a goto in case of failure..

giacomo

-- 
giacomo tenaglia
Technical Student at CERN IT/DES-SIS
CNR Biblioteca d'Area di Bologna - http://biblio.bo.cnr.it
Phone +41 76 5003376 - 
sip: