[log4perl-devel] memory leak

[Kevin G. <ke...@go...>]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Nice work, Mike, as usual.  Now to the question of alleged memory leak.

Here is code and annoted output

=============================================
use Log::Log4perl qw(get_logger);
$Log::Log4perl::CHATTY_DESTROY_METHODS=1;

Log::Log4perl->init(\ q{
log4perl.logger = DEBUG, A1
log4perl.appender.A1 = Log::Log4perl::Appender::Screen
log4perl.appender.A1.layout = SimpleLayout
});

$logger = get_logger();
warn "created logger $logger";
=============================================
created logger Log::Log4perl::Logger=HASH(0x8105018) at test line 13.
	--that's the one I created
Destroying logger Log::Log4perl::Logger=HASH(0x811e474) at
lib/Log/Log4perl/Logger.pm line 56.
	--there goes the root logger
Destroying logger Log::Log4perl::Logger=HASH(0x8105018) at
lib/Log/Log4perl/Logger.pm line 56.
	--there goes mine
Destroying logger Log::Log4perl::Logger=HASH(0x811e474) at
lib/Log/Log4perl/Logger.pm line 56.
	--there goes the root logger again
Destroying appender Log::Log4perl::Appender=HASH(0x82526bc) at
lib/Log/Log4perl/Appender.pm line 263.
	--there goes the appender

	--now we enter the GDP

Destroying appender Log::Log4perl::Appender=HASH(0x82526bc) at
lib/Log/Log4perl/Appender.pm line 263 during global destruction.
	--there goes the appender again

Destroying logger Log::Log4perl::Logger=HASH(0x8105018) at
lib/Log/Log4perl/Logger.pm line 56 during global destruction.
	--there goes the logger again


Now I assume you're worried about those two cleanups during the GDP.
Looking at the logger first, I tried dumping the structure in the
DESTROY method and by the time it's in the last line of output the
logger is totally empty, it's been cleaned out already.  I think the
reference it's cleaning up is the reference in *main:: in the script.  I
added "undef $logger" to the script to delete the reference from *main::
and that line disappeared from the output.  So I don't think that's a
problem.

That leaves the Appender. Again I tried dumping the struct from DESTROY
and it was totally empty.  Means there's a reference to it floating
around.  If you look at Logger::generate_coderef(), you'll see that
inside the coderef is a closure, $appenders is referenced inside the
coderef but is defined outside it.  That's our culprit.  If you change
the eval and return at the end of generate_coderef() to "return sub{}",
then the "Appender during global destruction" line goes away from the
output and that's it for the memory leak.

So I'm pretty sure it's the closure that's the problem.  With the
"return sub{}" change in generate_coderef(), putting a while loop around
the above code and running it forever doesn't use up any memory.

Since the appender is empty by the time it gets to the GDP, I'm pretty
sure that the cause of the memory leak isn't the appenders.  I think
it's the code eval. We're adding to the code each time through.  But
that's as far as I can get.  Devel::Leak might help, if I built a Perl
with debugging, but it's way, way past my bedtime.

On the bright side, it's really not a problem unless somebody is running
init() a thousand times in the life of the process, and we kind of
discourage that anyway, don't we?

- --
Happy Trails. . .

Kevin M. Goess
(and Anne and Frank)
904 Carmel Ave.
Albany, CA  94706
(510)525-5217
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.2.3 (GNU/Linux)
Comment: Using GnuPG with Netscape - http://enigmail.mozdev.org

iD8DBQFAxXI34g4/Tl71vUkRAsxmAKDGSyPd2VO8msvs0kEEmQBz+S0uqQCg1dsK
bzijFBfoGHni9jf22WDk6gk=
=os8v
-----END PGP SIGNATURE-----