912462 - Memory usage increase with the fix for bug 898558

Status: NEW

(Firefox :: General, defect)

Description

[Ioana (away)]

A memory usage increase has been detected by the Mozmill endurance testrun:

Last good: http://hg.mozilla.org/mozilla-central/rev/a81349b61b4e
First bad: http://hg.mozilla.org/mozilla-central/rev/d4f1b39a8575

testFlashIframe used 17mb more resident memory, and testFlashObject 8mb more, with the first bad build. The differences for resident memory for the other tests go around 1-3mb. All tests used more explicit memory than with the last good build:
http://mozmill-crowd.blargon7.com/#/endurance/report/3c3cd991de01b704f3f65783a1d848b5
http://mozmill-crowd.blargon7.com/#/endurance/report/3c3cd991de01b704f3f65783a1e92b6d

Comment 1

[Dave Hunt [:davehunt] [he/him] ⌚BST]

Is this an expected side-effect of bug 898558, Nicholas?

Comment 2

[Nicholas Nethercote [inactive]]

Bug 898558 will increase the 'explicit' measurement significantly -- it's now including more stuff in its measurement.  (You can see a big spike on the 'explicit' graph on areweslimyet.com, and it has an annotation explaining it.)

'resident' should be unchanged, though.

Comment 3

[Dave Hunt [:davehunt] [he/him] ⌚BST]

Does this need further investigation then, as resident memory has clearly increased? Does AWSY test many flash components? It's these tests that show the largest increase.

Comment 4

[Nicholas Nethercote [inactive]]

AFAIK, AWSY doesn't test any flash components.

Comment 5

[Dave Hunt [:davehunt] [he/him] ⌚BST]

Mike: Could you recommend someone to look into this? It appears that bug 898558 unexpectedly regressed resident memory usage.

Comment 6

[Mike Hommey [:glandium]]

How are those resident memory usage numbers taken?

Comment 7

[Dave Hunt [:davehunt] [he/him] ⌚BST]

We use the nsIMemoryReporterManager, the relevant code is here: http://hg.mozilla.org/qa/mozmill-tests/file/c5db4e89e35d/lib/performance.js#l5