109166 - Minimized components don't release memory

Status: RESOLVED WORKSFORME

(SeaMonkey :: General, defect)

Description

[selmer (gone)]

Hong's team ran a test scenario that shows we hold onto all of our memory until
*every* open component (nav, mail, AIM) gets minimized and only *then* do we let
things go and get into a quiescent state.

I doubt this is high in the performance hierarchy, but I wanted it in the list
for tracking purposes.

Comment 1

[timeless]

can you please provide steps to reproduce, system information (os/platform), and
expected results?  I seem to recall that on w2k windows would merily swap out
memory associated w/ minimized windows which seemed to match the behavior of ie.

if mozilla releases memory, what functionality is lost?

Comment 2

[selmer (gone)]

Hong, could you post your test results here?

Comment 3

[grylchan]

Attachment: attaching memory test case for win2k and NT 4.0

Adding my test case for Hong.

Comment 4

[Ere Maijala (slow)]

Even the habit of Windows itself minimizing the working set of a minimized
window often causes unnecessary heavy swapping. I believe it's generally a bad
idea for a program to try to minimize the working set unless it's very well
known that the memory isn't needed for a long time or so (as stated in MSDN). It
was tried upon closing windows and it caused horrible performance problems.

Comment 5

[Vedran Miletic]

retargeting

Comment 6

[OstGote!]

(In reply to comment #4)
> Even the habit of Windows itself minimizing the working set of a minimized
> window often causes unnecessary heavy swapping. 

see related Bug 76831, now we can configure the behavior of the working set

Comment 7

[Wayne Mery (:wsmwk)]

WFM per comment 6.
if you disagree, please reopen but change product to core.