842003 - Leak-check ipc/chromium stuff

Status: NEW

(Core :: IPC, defect, P3)

Description

[Chris Jones [:cjones] inactive; ni?/f?/r? if you need me]

See bug 841993 for the disaster the lack of this caused; without DMD, we would have been in a really bad spot.

The reason this stuff isn't leak checked is
 - in original and current desktop usage, we only hold onto one class there explicitly, the child process host.  Everything else is managed by it.
 - it's not moz code, sorta

But now with |opens| and |bridge| channels, we drag IPC:Channel out of ipc/chromium.  We were leaking those for ages and didn't know it.

Easiest thing to do is just let the moz code start trickling in there, starting with COUNT_CTOR/COUNT_DTOR.

Comment 1

[Ben Turner (not reading bugmail, use the needinfo flag!)]

We currently have [1]:

  typedef IPC::Channel Transport;

We could just inherit into our own class that adds logging instead.

Comment 2

[Ben Turner (not reading bugmail, use the needinfo flag!)]

Oops, that is in http://mxr.mozilla.org/mozilla-central/source/ipc/glue/Transport.h#24

Comment 3

[Jet Villegas (inactive)]

bent: Will a leak reporter here help real leaks (ie. non-leak on shutdown?)

Comment 4

[Ben Turner (not reading bugmail, use the needinfo flag!)]

Hm, not sure I understand the question. We definitely had real leaks that were not reclaimed at shutdown. We never saw them because these classes aren't instrumented.

Comment 5

[Justin Lebar (not reading bugmail)]

> We definitely had real leaks that were not reclaimed at shutdown.

Okay, that was the question.  Thanks.