Connections / Sessions Problem (unexpected loss/change/create of Session?)

[herzig]

Hello,

is it possible that some "actions" in PLSD 9.0.1.1617 cause a new connection to be opened?

I have several packages which use package session variables. This wasn't a problem until now.

Szenario:
-setting a session variable via sql window anonymous block package proc call
-select some data using this variable in where condition
-edit another package which becomes invalid by recompile
-call one of this package's procedures
-got invalid message (like normal)
-2nd call to same package proc
-package gets validated and works, but:
-got exception which leads to result, that session variable of prior packagage call is now undefined

First time I thought I simply forgot to set the session variable. Then -it happened again- I checked session list and found 3 sessions from me, running on PLSD instance in dual session mode.

Unfortunately I'm not able to reproduce this. Can I do something to catch this next time? Is there a log mode, I would see opening of another(Third) session ?

One thing which is maybe related:
Today was first time I used "multi connection". But again, I'm not sure, if this was same PLSD Instance, where problem occurs. I just know, I closed 2nd connection after some "playing".

regards, jo

 

[Marco Kalter]

is it possible that some "actions" in PLSD 9.0.1.1617 cause a new connection to be opened?

No, this is primarily affected by the "Session Mode" preference.

-got exception which leads to result, that session variable of prior packagage call is now undefined

Isn't this expected? The session state is implicitly reset when compiling/validating packages. Or am I missing something?

 

[herzig]

Well, I'm not sure, how it is supposed to be (officially).

In fact: I can't reproduce the problem, which implies, that full session state reset is not standard behavior. So actually I can invalidate some packages and work within same session with others- not invalidated- and even use their session variables.
And don't forget: I'm talking about two different packages. One hosting a variable, one other which gets recompiled. Certainly, I would acknowledge Your point of view, when looking on a single package.

I suggest, I come around again, if a got more evidence about it.