[BUG] Oracle interface 7.4

potential bug.

Our app connects without issue under 7.3 ,  in 7.4 Cincom added an  
override of #acquireExternal: i.e. into OracleThreadedConnection , which  
in our case fails. Furthermore, there are two issues:

1. we can't connect, so Oracle is for some reason now unhappy
2. the Smalltalk vendor error handling code below relies on a method that  
is absent from the 7.4 Oracle and Database related code base. Namely  
***authenticationMode:*** which I have tagged below is absent. I think  
something like Code Critic may be able to find this out :), nice tool.

Note, that simply removing the new implementation of #acquireExternal: and  
therefore allowing the code to hit the parent's implementation as it did  
under 7.3 will actually work i.e. we are able to connect and the app  
apparently works i.e. we are able to issue the same queries through the  
UIs.

BTW,the new implementation seems to be related to the supposedly now  
supported "connection pooling" capabilities.

Also, I have another question related to Oracle. We currently load the  
oraclient10.dll for both our VW3.0 code base and our 7.x code base. This  
works and it works as well under 7.4 given the minor kludge I point out  
above i.e. remove the new implementation of #acquireExternal:, it  
apparently will also work if we use oci.dll, which is what by default the  
libs specified by VW Oracle related external interfaces.

Question: What's the diff?

oraclient10.dll is certainly a fatter dll.

Can any of you Oracle gurus shed some light?

-Charles
I have Gemstone on my mind -- willie nelson.



OracleThreadedConnection>>>
acquireExternal: aPassword
        "Acquire the external resources involved with a connection to the
        SQL-environment."

        | xif name pw env pool useConnectionPool |
        xif := self class xif.
        name := self username.
        [ name isString ] whileFalse:
                [ name := self class requiredUsernameSignal raiseRequest ].
        pw := aPassword.
        [ pw isString ] whileFalse:
                [ pw := self class requiredPasswordSignal raiseRequest  ].
        pool := self connectionPool.
        useConnectionPool := (pool class name  == #OracleConnectionPool).
        useConnectionPool ifTrue: [
                env := pool environment.
        ] ifFalse: [
                env := self environment.
        ].
        env isNil ifTrue:
                [ env := self class defaultEnvironment ].
        [ env isString ] whileFalse:
                [ env := self class requiredEnvironmentSignal raiseRequest  ].
        env := self class mapLogical: env.

        [ "Begin unwind protection (1)"

                "Allocate necessary handles."
                useConnectionPool ifFalse: [ self unprotectedAllocHenv ].

                ["Begin unwind protection (2)"

                useConnectionPool ifTrue: [
                        self serverAttachConnectionPoolExternal.
                ] ifFalse: [
                        self serverAttachEnvironmentExternal: env.
                ].

                self setAttribute: self hsrvr type: #OCI_ATTR_SERVER
                  descriptor: self hsvcctx type: #OCI_HTYPE_SVCCTX.

                ( (name size == 0) and: [ pw size == 0 ]) ifTrue: [
                        self authenticationMode: xif OCI_CRED_EXT.
                ] ifFalse: [
        ---->> self ***authenticationMode:*** xif OCI_CRED_RDBMS.
                        self setAttribute: name type: #OCI_ATTR_USERNAME
                  descriptor: self hsess type: #OCI_HTYPE_SESSION .
                        self setAttribute: pw type: #OCI_ATTR_PASSWORD
                  descriptor: self hsess type: #OCI_HTYPE_SESSION.
                ].

                self setAttribute: self htxn type: #OCI_ATTR_TRANS
                descriptor: self hsvcctx type: #OCI_HTYPE_SVCCTX.

                [ "Begin unwind protection (3)"

                        self sessionBeginExternal.

                        self setAttribute: self hsess type: #OCI_ATTR_SESSION
              descriptor: self hsvcctx type: #OCI_HTYPE_SVCCTX.

                "End unwind protection (3)" ]
                        ifCurtailed:
                                [self  sessionEndExternal].

                "End unwind protection (2)" ]
                        ifCurtailed: [self  serverDetachExternal].

        "End unwind protection (1)" ]
                ifCurtailed:
                        [ "Something went wrong so we must implicitly free any handles we've  
allocated."
                                useConnectionPool ifFalse: [ self  unprotectedFreeHenv ].
                        ].

        self singleStatementTransactions


Stack:

Unhandled exception: BlockClosure [] in Object>>doesNotUnderstand:
OracleThreadedConnection(Object)>>doesNotUnderstand:
optimized [] in [] in OracleThreadedConnection>>acquireExternal:
BlockClosure>>ifCurtailed:
optimized [] in OracleThreadedConnection>>acquireExternal:
BlockClosure>>ifCurtailed:
OracleThreadedConnection>>acquireExternal:
OracleThreadedConnection(ExternalDatabaseConnection)>>acquire:
OracleThreadedConnection(ExternalDatabaseConnection)>>connect:
OracleThreadedConnection(ExternalDatabaseConnection)>>connect
optimized [] in [] in DsDbInterface class>>oracleConnect
BlockClosure>>ensure:
Cursor>>showWhile:
optimized [] in DsDbInterface class>>oracleConnect
ExternalDatabaseException class(GenericException class)>>handle:do:

--
Charles A. Monteiro

At 02:27 PM 3/16/2006, Charles A. Monteiro wrote:
>potential bug.
>
>Our app connects without issue under 7.3 ,  in 7.4 Cincom added an  
>override of #acquireExternal: i.e. into OracleThreadedConnection , which  
>in our case fails. Furthermore, there are two issues:
>
>1. we can't connect, so Oracle is for some reason now unhappy

It would probably be helpful if you could be a little bit more specific. Like, what platform, what version of Oracle, and any Oracle error messages.

>2. the Smalltalk vendor error handling code below relies on a method that  
>is absent from the 7.4 Oracle and Database related code base. Namely  
>***authenticationMode:*** which I have tagged below is absent. I think  
>something like Code Critic may be able to find this out :), nice tool.

It's in the distribution, implemented on OracleConnection in OracleEXDI. Perhaps you're not loading the right version, which might also be a cause of connectivity problems.

>Also, I have another question related to Oracle. We currently load the  
>oraclient10.dll for both our VW3.0 code base and our 7.x code base. This  
>works and it works as well under 7.4 given the minor kludge I point out  
>above i.e. remove the new implementation of #acquireExternal:, it  
>apparently will also work if we use oci.dll, which is what by default the  
>libs specified by VW Oracle related external interfaces.

There is likely no difference at least as far as our level of usage goes. Oracle follows some unusual naming convention for their libraries. Generally, it's preferable to use a name which isn't going to be different depending on what version of Oracle you're running. Both should provide the OCI functionality, and seem to.  If Oraclient10 is larger, and it isn't just a case of being divided up differently among required libraries, then it's possible there are extra functions they're throwing in there, but presumably not related to basic database connectivity.


--
Alan Knight [|], Cincom Smalltalk Development
[hidden email]
[hidden email]
http://www.cincom.com/smalltalk

"The Static Typing Philosophy: Make it fast. Make it right. Make it run." - Niall Ross

> It would probably be helpful if you could be a little bit more specific.  
> Like, what platform, what version of Oracle, and any Oracle error  
> messages.
>

very true. Windows XP service Pack II, Oracle 10g, no oracle errors per se

> It's in the distribution, implemented on OracleConnection in OracleEXDI.  
> Perhaps you're not loading the right version, which might also be a  
> cause of connectivity problems.

I specifically loaded the Oracle related parcels vial the parcel loader  
 from the 7.4 image.
Now, the distribution that I have is that which is provided to VW  
customers via CD i.e. and which we received in the mail from Cincom. I did  
not download this, so perhaps I did not "check the server" ? when  
installing?

would it be okay to point to the public repository and download whatever  
is there , if indeed available there?

thanks

-Charles



On Thu, 16 Mar 2006 14:46:43 -0500, Alan Knight <[hidden email]> wrote:



--
Charles A. Monteiro

I am not an Oracle guru so I may be wrong, but IIRC oci.dll checks which version of Oracle is installed and calls then the oracleXYZ.dll

-----Ursprüngliche Nachricht-----
Von: Alan Knight [mailto:[hidden email]]
Gesendet: Donnerstag, 16. März 2006 20:47
An: Charles A. Monteiro; [hidden email]
Betreff: Re: [BUG] Oracle interface 7.4


At 02:27 PM 3/16/2006, Charles A. Monteiro wrote:
>potential bug.
>
>Our app connects without issue under 7.3 ,  in 7.4 Cincom added an  
>override of #acquireExternal: i.e. into OracleThreadedConnection , which  
>in our case fails. Furthermore, there are two issues:
>
>1. we can't connect, so Oracle is for some reason now unhappy

It would probably be helpful if you could be a little bit more specific. Like, what platform, what version of Oracle, and any Oracle error messages.

>2. the Smalltalk vendor error handling code below relies on a method that  
>is absent from the 7.4 Oracle and Database related code base. Namely  
>***authenticationMode:*** which I have tagged below is absent. I think  
>something like Code Critic may be able to find this out :), nice tool.

It's in the distribution, implemented on OracleConnection in OracleEXDI. Perhaps you're not loading the right version, which might also be a cause of connectivity problems.

>Also, I have another question related to Oracle. We currently load the  
>oraclient10.dll for both our VW3.0 code base and our 7.x code base. This  
>works and it works as well under 7.4 given the minor kludge I point out  
>above i.e. remove the new implementation of #acquireExternal:, it  
>apparently will also work if we use oci.dll, which is what by default the  
>libs specified by VW Oracle related external interfaces.

There is likely no difference at least as far as our level of usage goes. Oracle follows some unusual naming convention for their libraries. Generally, it's preferable to use a name which isn't going to be different depending on what version of Oracle you're running. Both should provide the OCI functionality, and seem to.  If Oraclient10 is larger, and it isn't just a case of being divided up differently among required libraries, then it's possible there are extra functions they're throwing in there, but presumably not related to basic database connectivity.


--
Alan Knight [|], Cincom Smalltalk Development
[hidden email]
[hidden email]
http://www.cincom.com/smalltalk

"The Static Typing Philosophy: Make it fast. Make it right. Make it run." - Niall Ross

> It's in the distribution, implemented on OracleConnection in OracleEXDI.  
> Perhaps you're not loading the right version, which might also be a  
> cause of connectivity problems.

I have attached the .pst file for OracleEXDI which I obtained from the 7.4  
distribution that came in the CD from Cincom, the customer "SCM" release  
i.e not NC . I cannot find the method in question i.e.  
#authenticationMode: as Alan point out I must not have the current  
distribution. Where can I find a current "officially" supported i.e. not  
dev builds distribution for Oracle ?

I need this so that we can finally put 7.x into production and put our 3.0  
code base to pasture.

FYI, I decided to test this using the Ad Hoc SQL that comes with the  
Database parcel just to totally rule out any interactions with our app.  
Again, if I remove the implementation of #acquireExternal: in  
OracleThreadedConnection which was introduced with 7.4 , I will connect  
and then be able to issue ad hoc sql and successfully get result sets back.

thanks in advance,

-Charles

On Thu, 16 Mar 2006 14:46:43 -0500, Alan Knight <[hidden email]> wrote:


--
Charles A. Monteiro

Dear Charles,

I just tried it: standard installation from the CD, start vanilla image, load OracleEXDI by Parcel Manager, browse implementors of #authenticationMode:
=>
there it is: OracleConnection>>authenticationMode:

It is not a problem of the delivered version on CD.

Cheers
Helge

-----Ursprüngliche Nachricht-----
Von: Charles A. Monteiro [mailto:[hidden email]]
Gesendet: Freitag, 17. März 2006 16:07
An: Alan Knight; [hidden email]
Cc: James Robertson; Nowak, Helge
Betreff: Re: [BUG] Oracle interface 7.4


> It's in the distribution, implemented on OracleConnection in OracleEXDI.  
> Perhaps you're not loading the right version, which might also be a  
> cause of connectivity problems.

I have attached the .pst file for OracleEXDI which I obtained from the 7.4  
distribution that came in the CD from Cincom, the customer "SCM" release  
i.e not NC . I cannot find the method in question i.e.  
#authenticationMode: as Alan point out I must not have the current  
distribution. Where can I find a current "officially" supported i.e. not  
dev builds distribution for Oracle ?

I need this so that we can finally put 7.x into production and put our 3.0  
code base to pasture.

FYI, I decided to test this using the Ad Hoc SQL that comes with the  
Database parcel just to totally rule out any interactions with our app.  
Again, if I remove the implementation of #acquireExternal: in  
OracleThreadedConnection which was introduced with 7.4 , I will connect  
and then be able to issue ad hoc sql and successfully get result sets back.

thanks in advance,

-Charles

On Thu, 16 Mar 2006 14:46:43 -0500, Alan Knight <[hidden email]> wrote:



--
Charles A. Monteiro

Um, the file you attached does have that method in it.

At 10:06 AM 3/17/2006, Charles A. Monteiro wrote:
--
Alan Knight [|], Cincom Smalltalk Development
[hidden email]
[hidden email]
http://www.cincom.com/smalltalk

"The Static Typing Philosophy: Make it fast. Make it right. Make it run." - Niall Ross

You might want to check your $VISUALWORKS setting, which determines where it looks for parcels, and/or see if you have any parcels lying around where they might be found in your parcel path (e.g. in the current directory).

At 10:50 AM 3/17/2006, Alan Knight wrote:
--
Alan Knight [|], Cincom Smalltalk Development
[hidden email]
[hidden email]
http://www.cincom.com/smalltalk

"The Static Typing Philosophy: Make it fast. Make it right. Make it run." - Niall Ross

well, I don't understand, I took the CD delivered to us via US mail ,  
installed it and even provided you with a copy of the file said  
installation generated which shows that the method is missing so either  
I'm crazy , or you are crazy, or perhaps there are two versions of the  
official CD.

Which CD did you use?

Not sure if this is significant but the CD that I have again provided by  
Cincom has been tagged with what seems to be some run number:

E46-CSQE-0512

Secondly, since there is some "official" and more appropriate version, how  
do I get it?

-Charles


On Fri, 17 Mar 2006 10:33:48 -0500, Nowak, Helge <[hidden email]> wrote:



--
Charles A. Monteiro

Dear Charles,

as Alan has already pointed out the file you provided *does* contain the method. I see it in ViusalWorks as well as in a text editor. If you don't see it in VisualWorks there must be something strange with your image.

And BTW: my CD is just an usual CD as you got and has the same tag.

Cheers
Helge

-----Ursprüngliche Nachricht-----
Von: Charles A. Monteiro [mailto:[hidden email]]
Gesendet: Freitag, 17. März 2006 17:30
An: Nowak, Helge; Alan Knight; [hidden email]
Cc: James Robertson
Betreff: Re: AW: [BUG] Oracle interface 7.4


well, I don't understand, I took the CD delivered to us via US mail ,  
installed it and even provided you with a copy of the file said  
installation generated which shows that the method is missing so either  
I'm crazy , or you are crazy, or perhaps there are two versions of the  
official CD.

Which CD did you use?

Not sure if this is significant but the CD that I have again provided by  
Cincom has been tagged with what seems to be some run number:

E46-CSQE-0512

Secondly, since there is some "official" and more appropriate version, how  
do I get it?

-Charles


On Fri, 17 Mar 2006 10:33:48 -0500, Nowak, Helge <[hidden email]> wrote:



--
Charles A. Monteiro

so this is what I did to try to ascertain whether the method was actually  
in the source provided for the parcel in question.

1. I opened the VW 7.4 file browser
2. I found the "database" folder under the 7.4 distribution
3. I found the "OracleEXDI.pst" file.
4. went to the text pane where the source code is displayed
5. Popped mouse and selected "find it"
6. typed in "authenticationMode:"
7. Hit "ok"
8. File browser says: "Not Found"

ah, but now the second time that I'm doing this I noticed this button that  
says
"Show Entire File"

brilliant,

so I press that and then do everything I did above again and
voila it works.

Silly, silly me.

Not that I'm trying to excuse myself but a "hey dude, you are only  
searching through part of the file, it is probably best if you search  
through the entire file i.e. if you care about being accurate that is"  
dialog warn would be nice.

Now, back to main problem why this did not load when I loaded the parcel.

Onto the next email.

-Charles

On Fri, 17 Mar 2006 10:50:39 -0500, Alan Knight <[hidden email]> wrote:



--
Charles A. Monteiro

My $VISUALWORKS setting is set to: C:\vw7.4

I do have several installations of VW namely:

7.2.1
7.3
7.3.1
7.4

all directly under C:\

the image where the Oracle parcels were loaded was a virgin one. I also  
had published these to StORE. It picked up the correct 7.4 version of  
OracleThapiEXDI, so may be a Store load thing i.e. not a bug in StoRE, but  
perhaps a pre-req thing.

If that is the case it would then bring me to wishing that there was some  
sort of component pre-requisite enforcement i.e. OracleThapiEXDI 7.4  
should demand that OracleEXDI 7.4 is loaded regardless of where it is  
loading it from i.e. disk or StORE and any circumvention of that should at  
least be met with some warning.

-Charles


On Fri, 17 Mar 2006 11:11:05 -0500, Alan Knight <[hidden email]> wrote:



--
Charles A. Monteiro