circular memory leaks: cleanup in all browsers beneficial

[Peter Michaux]

I just ran some circular memory leak tests in IE6, O9, S3, FF2 and it
seems to me they all benefit from doing the same kind of circular
memory leak cleanup that IE requires.

My tests were very similar to Richard Cornford's which are very easy
to set up and watch run.

<URL: http://groups.google.com/group/comp.lang.javascript/msg/e723d3324aaa4127>

On my machine, O, S, and FF all level off at around 38MB of RAM while
the pages load and unload. If I am using the cleanup. If I don't use
cleanup they seem to level off at about 180MB.

I've always been using the onunload cleanup for all browsers and was
considering stopping doing that for all but Internet Explorerer. I
won't be stopping on any browser. It seems all the garbage collectors
can use the help.

Peter

[Peter Michaux]

On Feb 18, 10:33 am, Peter Michaux <[e-mail]:

I just ran some circular memory leak tests in IE6, O9, S3, FF2 and it
seems to me they all benefit from doing the same kind of circular
memory leak cleanup that IE requires.

My tests were very similar to Richard Cornford's which are very easy
to set up and watch run.

<URL:http://groups.google.com/group/comp.lang.javascript/msg/e723d3324aaa4127>

On my machine, O, S, and FF all level off at around 38MB of RAM while
the pages load and unload. If I am using the cleanup. If I don't use
cleanup they seem to level off at about 180MB.

I've always been using the onunload cleanup for all browsers and was
considering stopping doing that for all but Internet Explorerer. I
won't be stopping on any browser. It seems all the garbage collectors
can use the help.

Oddly enough it seems that just the following in the non-IE browsers
is enough to trigger garbage collection onunload

global.addEventListener(
'unload',
function() {},
false);

Peter

[Ed]

On Feb 18, 12:58 pm, Peter Michaux <[e-mail]:

On Feb 18, 10:33 am, Peter Michaux <[e-mail]:

I just ran some circular memory leak tests in IE6, O9, S3, FF2 and it
seems to me they all benefit from doing the same kind of circular
memory leak cleanup that IE requires.

My tests were very similar to Richard Cornford's which are very easy
to set up and watch run.

<URL:http://groups.google.com/group/comp.lang.javascript/msg/e723d3324aaa4127>

On my machine, O, S, and FF all level off at around 38MB of RAM while
the pages load and unload. If I am using the cleanup. If I don't use
cleanup they seem to level off at about 180MB.

I've always been using the onunload cleanup for all browsers and was
considering stopping doing that for all but Internet Explorerer. I
won't be stopping on any browser. It seems all the garbage collectors
can use the help.

Oddly enough it seems that just the following in the non-IE browsers
is enough to trigger garbage collection onunload

global.addEventListener(
'unload',
function() {},
false);

Peter

I think this is explained here:

http://developer.mozilla.org/en/docs...indow.onunload

It says, "using this event handler in your page prevents Firefox 1.5
from caching the page in the in-memory bfcache".

[Peter Michaux]

On Feb 18, 11:42 am, Ed <[e-mail]:

On Feb 18, 12:58 pm, Peter Michaux <[e-mail]:

Oddly enough it seems that just the following in the non-IE browsers
is enough to trigger garbage collection onunload

global.addEventListener(
'unload',
function() {},
false);

Peter

I think this is explained here:

http://developer.mozilla.org/en/docs...indow.onunload

It says, "using this event handler in your page prevents Firefox 1.5
from caching the page in the in-memory bfcache".

Ed, thank you. That seems like a reasonable explanation.

Peter

[David Mark]

On Feb 18, 3:33*pm, Peter Michaux <[e-mail]:

On Feb 18, 11:42 am, Ed <[e-mail]:

On Feb 18, 12:58 pm, Peter Michaux <[e-mail]:

Oddly enough it seems that just the following in the non-IE browsers
is enough to trigger garbage collection onunload

* * * global.addEventListener(
* * * * 'unload',
* * * * function() {},
* * * * false);

Peter

I think this is explained here:

*http://developer.mozilla.org/en/docs...indow.onunload

It says, "using this event handler in your page prevents Firefox 1.5
from caching the page in the in-memory bfcache".

Which also prevents fast history navigation.

Ed, thank you. That seems like a reasonable explanation.

So perhaps those were false positives in the other browsers?

[Peter Michaux]

On Feb 18, 2:32 pm, David Mark <[e-mail]:

On Feb 18, 3:33 pm, Peter Michaux <[e-mail]:

On Feb 18, 11:42 am, Ed <[e-mail]:

On Feb 18, 12:58 pm, Peter Michaux <[e-mail]:
Oddly enough it seems that just the following in the non-IE browsers
is enough to trigger garbage collection onunload

global.addEventListener(
'unload',
function() {},
false);

Peter

I think this is explained here:

http://developer.mozilla.org/en/docs...indow.onunload

It says, "using this event handler in your page prevents Firefox 1.5
from caching the page in the in-memory bfcache".

Which also prevents fast history navigation.

Ed, thank you. That seems like a reasonable explanation.

So perhaps those were false positives in the other browsers?

They do seem to level off at about 180MB of RAM use so I can imagine
the cache is full at that point and just killing the oldest stuff. If
it was a leak it would just keep building, I would think.

Peter

[dhtml]

On Feb 18, 2:32 pm, David Mark <[e-mail]:

On Feb 18, 3:33 pm, Peter Michaux <[e-mail]:

On Feb 18, 11:42 am, Ed <[e-mail]:

On Feb 18, 12:58 pm, Peter Michaux <[e-mail]:
Oddly enough it seems that just the following in the non-IE browsers
is enough to trigger garbage collection onunload

global.addEventListener(
'unload',
function() {},
false);

Peter

I think this is explained here:

http://developer.mozilla.org/en/docs...indow.onunload

It says, "using this event handler in your page prevents Firefox 1.5
from caching the page in the in-memory bfcache".

Which also prevents fast history navigation.

Ed, thank you. That seems like a reasonable explanation.

Not only that, it's a good reason to want to not have an onunload
handler in a lot of cases.

So perhaps those were false positives in the other browsers?

It would be nice there were a good way to detect memory leaks, so you
could conditionally add an onlunload handler.