13283 - MLK: Webshell leaks (again!)

Status: VERIFIED FIXED

(Core Graveyard :: Embedding: APIs, defect, P3)

Description

[Bruce Mitchener]

The webshell leaks.

This blocks a lot of purify work, any attempt to be sure that gecko doesn't leak
like sieve.  It is probably blocking any resolution or verification of the CSS
leaks (which may not even exist).

Comment 1

[Bruce Mitchener]

Adding CSS leaks as dependencies.

Comment 2

[Mike Shaver (:shaver -- probably not reading bugmail closely)]

kipp: rumour has it that you're going to apply scc's mighty WeakReference hammer
to this particular nail.  If so, can you take this bug?

If not, scc and kipp, could the two of you dump your brains at me or refer me to
previous such dumps so I can take a shot?  We're putting M11 (and subsequent
dependent Ms) at risk by hiding piles of other leaks and naughty behaviour under
the webshell's immense shadow.

Comment 3

[kipp]

False rumour :-(

scc is the king of webshell, and as the new and improved owner he should be all
over it...yes? Feel free to nag him and chip in, but talk with him first!

Comment 4

[Bruce Mitchener]

rumor has it travis might own webshell now?

yes. i'm grabbing for straws. I admit it.

Comment 5

[Mike Shaver (:shaver -- probably not reading bugmail closely)]

Yeah, according to the arch meeting today, scc _doesn't_ own it.  Who does?
This webshell leak is a _major_ problem for us.  Do I have to just start from
first principles?

Comment 6

[Bruce Mitchener]

vidur says that this is in part caused by a circular dependency between the
DocLoader and the LoadGroup, which brings down the webshell.

Comment 7

[Mike Shaver (:shaver -- probably not reading bugmail closely)]

It is with no small glee that I mark this FIXED, thanks to the incredible
efforts of vidur and rpotts.

Comment 8

[Claudius Gayle]

leaving this to te reporter/owner to verify. Goood job guys.

Comment 9

[Claudius Gayle]

code-level fix, marking VERIFIED