Re: still memory leaks with libpgtcl

ljb wrote:

> Sure people are using it. As I understand it: The Tcl interface included
> with PostgreSQL-7.3.1 is the preferred, production release. The new
> libpgtcl on gborg.postgresql.org is beta-test; when it is done, the Tcl
> interface will be unbundled from the PostgreSQL releases (like the perl5
> interface already is). The latest release of the unbundled libpgtcl-1.4b3
> seems pretty good to me (I was unable to break it), so I suggest you try it
> if you can.
>

OK, good to know. In the meanwhile I upgraded to 7.3.1, so 1) and 2) are OK.

> As to your specific issues:
> 
> 1) The PostgreSQL-7.3.1 released libpgtcl does have the connection loss
> handling function.
> 
> 2) I'm not sure what you are referring to by "fix for finding a free
> connection slot (PSetResultID)". If you mean finding a free result
> structure slot (PGSetResultID), then I see code changes there and I think
> that is fixed in both 7.3.1 and 1.4b3.  It seems to work as it should: I
> can allocate 128, but not 129 result structures.
> 
> 3) Memory leak on connect/disconnect unfortunately is not fixed in either
> the production or beta versions.  I'm seeing about 4KB loss per
> connect/disconnect. I'm sure this is serious for some applications, but
> others would probably not have trouble here. I can't remember: was there a
> patch - has this leak been tracked down?

Definitely serious, took me a long time to find out, why two machines hat to be 
hard rebooted because they ran out of memory after some weeks running with no 
problems. Changed the "connect/disconnect when idle" strategy to "connect at 
startup and stay so".
If I'd track down the problem and sent a patch, has the wheel to be reinvented 
for gborg ? Two concurrent developments is not a good thing.

Oops, forgot to check that thread. Thanks for your response.

-- 
Gerhard Hintermayer
http://www.inode.at/g.hintermayer