After an evaluation, GNOME has moved from Bugzilla to GitLab. Learn more about GitLab.
No new issues can be reported in GNOME Bugzilla anymore.
To report an issue in a GNOME project, go to GNOME GitLab.
Do not go to GNOME Gitlab for: Bluefish, Doxygen, GnuCash, GStreamer, java-gnome, LDTP, NetworkManager, Tomboy.
Bug 553193 - free my memory on tab close
free my memory on tab close
Status: RESOLVED INVALID
Product: epiphany
Classification: Core
Component: Backend
git master
Other Linux
: Normal critical
: Future
Assigned To: Xan Lopez
Epiphany Maintainers
Depends on:
Blocks:
 
 
Reported: 2008-09-22 01:14 UTC by Jean-François Fortin Tam
Modified: 2011-12-30 16:21 UTC
See Also:
GNOME target: ---
GNOME version: ---



Description Jean-François Fortin Tam 2008-09-22 01:14:37 UTC
I just killed epiphany (the process was using 120 mb of ram), and started it again with the same tabs, and it was now using only 50 mb.

This tells me that it may be a good idea, even though the webkit migration is supposed to bring us the holy grail and ponies, to make epiphany have each tab in a separate system process. That way, if a tab is closed, ALL of its memory is freed. No leftovers like now. Currently, I frequently have to kill epiphany and restart it to gain back some of my system swiftness.

I think the google chrome guys did that with their webkit-based browser. It also has the advantage that if a tab would crash, it would only crash the tab, not the browser.

Whatddyathink?
Comment 1 Jean-François Fortin Tam 2009-03-29 01:48:54 UTC
I tried epiphany webkit 2.27 from kov's repository, and it suffers from the exact same memory problems as epiphany gecko.

Demonstration of the problem: http://jeff.ecchi.ca/public/epiphany%20webkit%20memory%20usage.ogv

Changing from enhancement to major bug status, because it is a "severe memory leak" and it is just plain crazy that I have to kill epiphany twice a day to go back to a sane level (and stop swapping for nothing).

Just now, Epiphany was using 473.9 Mib of ram (according to gnome-system-monitor's "memory" column), and after killing it, restarting it and waiting for all tabs to load, it is now using 187.6 Mib. That's quite a significant difference that cannot be attributed to caching pages in memory, and it makes epiphany painful to use on machines with less than 16 Exabytes of ram.

For what it's worth, Chromium doesn't do this.
Comment 2 Xan Lopez 2009-03-29 06:59:14 UTC
I'm sure there are a ton of memory leaks, this is very new code, and once all those are gone there's memory fragmentation. Obviously Chrome will free all memory once you close a tab, but each tab will use more memory, so you don't get anything for free.
Comment 3 Xan Lopez 2009-08-05 21:36:17 UTC
I'm closing this bug as it's too generic to be useful. Feel free to open bugs for specific leaks that you could find (and besides, there's already a bug opened to make ephy use one webkit process per tab).
Comment 4 Jean-François Fortin Tam 2009-08-06 00:39:19 UTC
Please make my bug report a duplicate of the one you are referring to, because I could not find it (with keywords such as process, memory, or tab). Reopening until then. Thanks!
Comment 5 Reinout van Schouwen 2009-12-18 18:46:41 UTC
That bug is https://bugs.webkit.org/show_bug.cgi?id=24935 . :)
Comment 6 antistress 2011-12-30 16:21:40 UTC
The current work on webkit2 gtk+ may solve that problem i think ?