Shell I get valgrid data?

Yes.

I've tried few times but it crashed each time. On Valgrid mailing list I didn't get any help.

I too use epiphany and I can confirm this happens.
Eventually I have to close all the browser windows and restart epiphany. When you close the last forked window, you will see in the system monitor a huge square wave, as all of a sudden a bunch of memory becomes free.
Let me know if you need more information. This is particularly annoying and bad. thanks! :(
James, purpleidea@gmail.com

Present in 2.20.3. However not always reproducable (sometimes after few houres sometimes days).

@James: Although I will try with valgrid once more since the bug is marked with NEEDINFO any information will be appriciated.

(In reply to comment #5)
> Present in 2.20.3. However not always reproducable (sometimes after few houres
> sometimes days).
> 
> @James: Although I will try with valgrid once more since the bug is marked with
> NEEDINFO any information will be appriciated.
> 

Hey,
I'm not very experienced with these matters, I've used valgrind a bit, but not sure what I can do to help, but I am willing if you could suggest something / guid e me. Let me know,
cheers,

_J
ps: I know that epiphany will be soon supporting/changing to the webkit backend ( http://live.gnome.org/Epiphany/WebKit ) do you think this change will solve the problem? I just mention it in case it would save you a lot of work fixing a bug in a backend that will get replaced.

Well - it depends where the problem is. If the problem is with Gecko or integration with Gecko it will be solved (of course if it is not copied into new backend). However the bug should be fixed since both engines will be supported.

If the problem is somewhere else it will not.

Created attachment 104171 [details]
G_SLICE=always-malloc valgrind --leak-check=full --log-file=$HOME/Desktop/Bugs/epiphany-valgrid epiphany

It is the start-up. Unfortunatly my computer is too weak to provide a full debug (memcheck consumed 1.5 of RAM etc.).

I see a lot of references to Python in the valgrind log...
Can you disable all Python extensions and see if you can narrow down the problem that way?

1. As far as I remember the python is a little bit broken with valgrind
2. Which are those plugins?

A little bit of research doesn't hurt...

[reinout@reinout-rcs extensions]$ find . -name '*.py'
./cc-license-viewer/cc-license-viewer.py
./favicon/favicon.py
./epilicious/epilicious.py
./epilicious/config.py
./epilicious/magnolia.py
./epilicious/backend.py
./epilicious/progress.py
./epilicious/EpiphanyStore.py
./epilicious/BaseStore.py
./epilicious/DeliciousStore.py
./epilicious/__init__.py
./python-console/console.py
./sample-python/sample-python-statusbar.py
./sample-python/sample-python-location.py
./sample-python/sample-python-menu.py
./sample-python/sample-python.py

Plus any python extns you may have saved in ~/.gnome2/epiphany/extensions/.

I had only favicon enabled. If the problem will be stopped I will notify.

Created attachment 114456 [details]
G_SLICE=always-malloc valgrind --leak-check=full --log-file=$HOME/Desktop/Bugs/epiphany-valgrid epiphany (without python plugins)

Valgrid log without python plugins

PS. The log is from 2.22.3.

Which mem leak from the valgrind log is this bug about, then? Note that not all 'definitive' leaks are really leaks, they may also just be global data that's alive during the whole programme run and just (intentionally) not cleaned up on exit.

(In reply to comment #15)
> Which mem leak from the valgrind log is this bug about, then? Note that not all
> 'definitive' leaks are really leaks, they may also just be global data that's
> alive during the whole programme run and just (intentionally) not cleaned up on
> exit.
> 

I'm not sure. However I can trace the whole live of epiphany by valgrid if somebody will sponsor a new procesor (any since you cannot upgrade it in my case without upgrading motherboard... ;) ). The fact is (was? after upgrade of RAM to 2 GiB it not seems to consume more then 66% in a worst case) that epiphany consumed 1.5 GiB of virtual space (as I had 512 MiB it cannot consume 1.5 GiB of memory).

It was pointed out that many warrings etc. is in python and it was suggested that it might be responsible for this effect.

Also there is hypothesis that the grow is in virtual address space only (I need to check it) or it was solved during development.

Seems to happen no more. I'll reopen if I reproduce.

Problem still seems to be valid although not so dramatic.

Just a second ago it consumed 1 GiB of memory and it rised sharply.

This is just way to vague to do anything about it IMHO. If you manage to figure out something specific please re-open...