[Dbi-interbase-devel] Leak, but no leak (was: Re: DBD::InterBase Leaking?)

[Edwin P. <ed....@co...>]

Edwin Pratomo wrote:
> 
> Michael Samanov wrote:
> 
> >
> > I'm  not  thinking it's a deal of SS or classic server: you are client
> > and  it  is server, and you are both different processes communicating
> > via  TCP/IP.  If IBPerl code shows the same problem then it means that
> > some  leakage  is  in  the  Interbase client library - gds.a (or, more
> > likely, gds.so).

Silly question: where is the libgds.a?

> 
> I've just tried this test using DSQL C API, and it seems that _both_
> drivers, mine and IBPerl has leaks.
^^^^^^^^
whoops, IBPerl is not a driver. sorry.

> For now I haven't found where exactly it leaks. Need investigate this
> further.
> 

After spending hours observing this behavior, finally I've located the
problem.
I've attached the test sources and the result files as well for you to
see.
My conclusion is the problem is not in the driver (I dont know about
IBPerl, haven't taken a closer look at that), because the _same_
behaviour is also observed when directly using InterBase's DSQL_API,
with the same logic flow as in the driver.
My leak.c works like the driver in AutoCommit on, and noauto.c works
like in AutoCommit off.
From the log file created by leak.pl, we can see how the ibserver
process growing. comparing the top_autocommit_1.txt and
top_autocommit_2.txt, seems a bit confusing. the second log shows a
fixed size increment of 4 KB at every 16304 loop (what is this
mysterious number?), after the first increment at the 12349-th loop. 
After repeating the test many times, I see that ibserver starts growing
after the 10000-th loop. 
The size is restored after disconnection. So we may encountered serious
problem if we are running DBD::InterBase or more exactly, any client, in
a persistent database connection such as Apache::DBI if we make such
loops (more than 10000 loops) many times, before the connection is
disconected (and reconnected) due to timeout.

Rgds,
Edwin.

> 
> > I just have tested - FreeBSD has the same problem.
> >