158805 - Image reload eats lots of memory

Status: VERIFIED INVALID

(Core :: Networking: Cache, defect, P3)

Description

[Peter 'vt100' Schwindt]

From Bugzilla Helper:
User-Agent: Mozilla/5.0 (Windows; U; Windows NT 5.0; en-US; rv:1.1b) Gecko/20020721
BuildID:    2002071208

Watching upper page eats lots of memory - every image reloaded seems to be put
into ram - and none of the old ones being removed

Reproducible: Always
Steps to Reproduce:
1. open http://cam0.ph-freiburg.de
2. start taskmanager
3. see the memory use rising

Actual Results:  Memory being eaten up

Expected Results:  don't use more ram every image reload

Comment 1

[Boris Zbarsky [:bzbarsky]]

and the relationship to the JS engine is.... what?

Comment 2

[Peter 'vt100' Schwindt]

The reload is being done by javascript - I thought there could be a problem...

But well, if you don't think so...

Comment 3

[WD]

I think this is a cache issue?

See also bug 157187
(changing webcam images refreshed via javascript causes memory usage to increase
without end)

Comment 4

[WD]

*** Bug 173928 has been marked as a duplicate of this bug. ***

Comment 5

[Daniel de Wildt]

Is this bug still valid?
The URL above is not available, so I can not compare this bug with bug 229839.
Maybe someone could dup that bug, if this bug has an image reload based on a
redirect.

Otherwise:
The URL in bug 173928 reloads it's image with a dynamic generated url, which is
never be the same. It seems that each image got an other cache key. If the
reload for the URL above also generates a dynamic URL, then the observed memory
increase is the cache filling process. If this is so, then this bug should be
closed with WONTFIX. 
(This could be wrong because bug 173928 said that "Finally Mozilla *crashes* or
eats up *all* available memory".)

Comment 6

[Daniel de Wildt]

Report, maybe what you observe is the filling of the cache.
How big is the memory cache? (Enter about:cache in the location bar)
Is Mozilla/Firebird still leaking if the cache is full?

How much memory is used (task manager) after you cleared the (memory) cache?

Comment 7

[Peter 'vt100' Schwindt]

cache size (memory) is 21504 k - the disk one is 50000 k

I did not see it "leek" memory after having filled up - but it seems that
mozilla is somehow fixed as it doesn't allocate tons of ram just for that one
image being reloaded...

it only goes up to about 17M, then stops using more of it

Comment 8

[Daniel de Wildt]

In that case it seems that the memory consumption is the normal process, when
the cache is filled. I think that the reload generates an dynamical (not
indentical) url for each new image. The cache might 'think' in such cases that
it's not the same image and will cache it. It will release not needed images
from the cache when it needs some new memory.

Is it somehow possible for me to get access to that webcam or could you create
an minimal attachment with the html source?

BTW, maybe the server could send a header "Pragma: no-cache" to prevend caching
that images.

Comment 9

[Peter 'vt100' Schwindt]

resolving as fixed after getting mail from Stephan and Daniel

Comment 10

[Jason Bassford]

Was something actually broken and fixed in Mozilla - or was the behaviour normal
behaviour?  It's not clear from the comments.  Was there some patch that was
checked in, here or elsewhere that corrected things?  (Nothing's been attached
to this bug.)

If nothing was actually checked in, anywhere, then "Fixed" is NOT the correct
resolution.  It should be reopened and marked as either Invalid or Worksforme.

Comment 11

[Jason Bassford]

In fact, I'm reopening it myself.  If something DID fix it please comment
appropriately to indicate what it was before marking it Fixed again.  Otherwise
resolve it correctly.

Comment 12

[Daniel de Wildt]

I mailed the webcam administrator (because the webcam is down) and he sent me a
link to the webcam manufacturer and a link to a working webcam. This webcam has
a dynamical generated url (see comment #5). So I suggested that the bug
description is the normal process when the cache is filled (comment #8). The
webcam administrator agreed that the memory consumption doesn't increase further
after the cache is filled.
I couldn't close this bug because I was in holidays, so I recommended the
reporter to close the bug by himself. On basis of this results I think this bug
should be closed as INVALID.

Comment 13

[benc]

V: Daniel, thanks for looking into that problem