hcitpsmsgkill for 5.6

[Charlie Bursell]

For anyone using hcitpsmsgkill to kill replies in Cloverleaf 5.6, it is possible to get a memory leak since the State 16 message is not killed.

We have rewritten the hcitpsmsgkill proc to handle it.

Topic

[Max Drown (Infor)]

Charlie, we are on 5.6 and we use the hcitpsmsgkill proc quite a lot. I want to make sure I make this replacement properly.

-- Max Drown (Infor)

[Charlie Bursell]

First of all, if you are not using it in an inbound reply context, you do not need the modified version

If you do need it I would not install it in ROOT/tclprocs as a mktclindex in that directory could make other things not work.  Look at the tclIndex in that directory and you will see it uses pointers to other directories which would get lost with a mktclindex.

I would put it in my site or master site directory and maintain the hcitpsmsgkill name.  Since the search path is site, master site, root  the engine would find and load the modified one intead of the one in the root.

I hope this maes sense

[Bob Richardson]

Folks,

We have installed Charlie’s modified hcitpsmsgkill proc in our common site and it works fine. (We are AIX5.2 and have all tclprocs loaded into a shared site that is soft-linked into by other sites).  We were experiencing a leak of the new State 16 handles in our medical instruments interfaces where we disabled using the recovery database for these interfaces.  If I understand the new state 16 handling, if recovery database is enabled then the engine will destroy the state 16 handle at IB Reply context time

anyway (this is the behavior for our interfaces that have the recovery database enabled and use this proc to kill the inbound reply).

Hopefully, this information will prove useful to you.

Good luck!

[Author]

Replies