Mark Banner (:standard8) Reporter
	Description • 17 years ago

The following tests all fail on Thunderbird and Sunbird trunk builds with the same assertion on exit:

test_bug299716.js
test_bug378216.js
test_bug394300.js

The tests themselves seem to pass, but on exit, the assertion below is raised:

###!!! ASSERTION: ### mem cache leaking entries?
: 'mEntryCount == 0', file /Users/moztest/mozilla/hg/mozilla/netwerk/cache/src/nsMemoryCacheDevice.cpp, line 130

This happens on Mac & Linux.

Both Thunderbird and Sunbird have

NECKO_DISK_CACHE=

in confvars.sh. If I remove this (at least on the Sunbird build) then the tests pass completely.

To complete the loop, if I remove the "NECKO_DISK_CACHE=" line AND specify "browser.cache.disk.enable" as false in the relevant prefs file, then the assertion occurs again.

So I don't think Thunderbird and Sunbird can get around this without a fix in core.

	Mark Banner (:standard8) Reporter
	Comment 1 • 17 years ago

I have just done some further testing, this assertion only happens in debug mode. 

However, I believe it is implying a memory leak, and seeing as users can disable the disk cache (e.g. for mobile devices?) I'm leaving it open, but removing it from blocking the setting up of unit tests for Thunderbird/Sunbird.

Expanding on comment 0:

test_bug299716.js
test_bug378216.js
test_bug394300.js

these tests are all in toolkit/mozapps/extensions

test_download_manager.js

in toolkit/components/downloads also causes this assertion.

	Gary Kwong [:gkw] [:nth10sd] (NOT official MoCo now)
	Comment 3 • 17 years ago

I hit this occasionally on shutdown as well, in a debug 3.0a2pre TB Mac build.

	Gary Kwong [:gkw] [:nth10sd] (NOT official MoCo now)
	Comment 4 • 17 years ago

Hope this helps.

	Serge Gautherie (:sgautherie)
	Comment 5 • 16 years ago

[Mozilla/5.0 (Windows; U; Windows NT 5.0; en-US; rv:1.9.1a2pre)
Gecko/2008080202 SeaMonkey/2.0a1pre] (home, debug default) (W2Ksp4)

Usual |make check| leaks, but without assertion:
{{
test_bug299716.js.log
 => mFreeCount:           26959  --  LEAKED 436 !!!

test_bug378216.js.log
 => mFreeCount:           33597  --  LEAKED 434 !!!

test_bug394300.js.log
 => mFreeCount:           12169  --  LEAKED 434 !!!
}}

	Serge Gautherie (:sgautherie)
	Comment 6 • 16 years ago

(In reply to comment #5)
> [Mozilla/5.0 (Windows; U; Windows NT 5.0; en-US; rv:1.9.1a2pre)
> Gecko/2008080202 SeaMonkey/2.0a1pre] (home, debug default) (W2Ksp4)

<test_download_manager.js> is not "built / run",
as SM does not use Toolkit D.M. (yet).

	Mark Banner (:standard8) Reporter
	Comment 7 • 16 years ago

(In reply to comment #5)
> [Mozilla/5.0 (Windows; U; Windows NT 5.0; en-US; rv:1.9.1a2pre)
> Gecko/2008080202 SeaMonkey/2.0a1pre] (home, debug default) (W2Ksp4)

Serge, please read comment 0 FULLY. This bug only happens in TB & SB because NECKO_DISK_CACHE is not set (i.e. not 1).

It does NOT happen in SM or FF because they don't have that set by default.

	Mark Banner (:standard8) Reporter
	Comment 8 • 16 years ago

FWIW test_bug394717.js now exhibits this assertion/failure as well (in TB/SB).

	Serge Gautherie (:sgautherie)
	Comment 9 • 16 years ago

(In reply to comment #5)
>  => mFreeCount:           26959  --  LEAKED 436 !!!
>  => mFreeCount:           33597  --  LEAKED 434 !!!
>  => mFreeCount:           12169  --  LEAKED 434 !!!

These leak(s) looks like to be bug 448980 (++).


(In reply to comment #7)
> Serge, please read comment 0 FULLY. This bug only happens in TB & SB because
> NECKO_DISK_CACHE is not set (i.e. not 1).

I did.
The "missing" word in comment 0 was *only*:
I thought it wouldn't hurt to add comment 5.

	Patrick McManus [:mcmanus]
	Comment 10 • 9 years ago

new cache code