98721 - TestGtkEmbed fails to luanch with seg fault (core dump)

Status: VERIFIED FIXED

(Core Graveyard :: Embedding: APIs, defect)

Description

[Tracy Walker [:tracy]]

seen on TestGtkEmbed 2001-09-07-06-trunk

-Install the build
-launch it

it fails to start up with just a seg fault message and (core dump) on the 
console

Comment 1

[Tracy Walker [:tracy]]

this is blocking the memory usage test.  perhaps this is also the reason for the 
increased leakage? go to 
http://ftp.mozilla.org/pub/data/PerformanceAndFootprints.html
to see graphical results of these tests.  

Note: The improved tag on the memory test is a obviously incorrect.  Don't we 
all wish the app could use zero memory.  I've resubmitted the data to this page 
it should only take a few minutes before it is updated to the world.

Comment 2

[Adam Lock]

mfcEmbed and PPembed have problems too.

I suspect it is something to do with recent component manager changes because
the component.reg file is not being created correctly.

Investigating further.

Comment 3

[Adam Lock]

CC'ing Doug.

NS_InitEmbedding is calling NS_InitXPCOM2 which is returning a nsnull
nsIServiceManager in the first argument. This makes NS_InitEmbedding unhappy so
it is returns an error code.

I will try and hack a patch around this, but I need to know if this new
behaviour is deliberate or not. From inspecting our embed api code I think we
could (and should) switch over to the do_QueryService method.

Comment 4

[Adam Lock]

Attachment: Patch stops NS_InitEmbedding from caring whether NS_InitXPCOM2 returns a good service manager pointer or not

Comment 5

[Adam Lock]

r=[e-mail]

Comment 6

[Christopher Blizzard (:blizzard)]

sr=blizzard

The larger bug needs to be fixed, though.  Please don't let it fall through the
cracks.

Comment 7

[Adam Lock]

Fix is checked in. Bug 98735 opened to deal with the issue of whether
NS_InitXPCOM2 even needs that argument anymore.

Comment 8

[Tracy Walker [:tracy]]

verified fixed on linux TestGtkEmbed 2001-09-07-11-trunk