1323344 - multiple (nested?) previous session restorations draining ram

Status: RESOLVED INCOMPLETE

(Firefox :: Session Restore, defect)

Description

[harcesz]

Attachment: just 16 windows full of tabs to read over the weekend.

User Agent: Mozilla/5.0 (X11; Ubuntu; Linux x86_64; rv:50.0) Gecko/20100101 Firefox/50.0
Build ID: 20161130094234

Steps to reproduce:

I kill firefox a lot, to be able to come back to what I was looking at a particular point and read up, while atm I need a 'clean new' browser, since the last session had to much.

basically, open up 20-100 tabs, kill, reopen firefox, don't retrieve the last session, leave the restoration waiting, open another 30-50 tabs, kill firefox, repeat 20 times. (that's my actual numbers, just checked).


Actual results:

even with just a few pages open in the current session (and 20 retrieval's one in another) the entire system is mulled by firefox

oddly after retrieving all these into about 16 windows and few hundred (unloaded) tabs firefox seemed to manage better, and is much better off with all of that packed back into a single session retrieval (yes, again)


Expected results:

I would expect multiple session retrieving, nested in each other, not to take any more ram than displaying a text file. definitely not the case.

Comment 1

[Wayne Mery (:wsmwk)]

>  basically, open up 20-100 tabs, kill, reopen firefox, don't retrieve the last session, leave the restoration waiting, open another 30-50 tabs, kill firefox, repeat 20 times.

It doesn't necessarily work reliably that way.  Been there.

1. what is the size of your sessionstore.js?
2. what is size reported after using about:sessionstore (not about:sessionrestore) and cleaning out closed tabs and windows at bottom of the page?

Comment 2

[Firefox Bug Husbandry Bot]

2019-03-06

This bug is part of a group of bugs which have had an open needinfo for at least 12 weeks.

The request for information has not been answered, and we can't move forward on the bug so we are closing it.

If the defect is still present, please reopen this bug with an updated report.