Bug 56673 – Leak of DocumentLoader seen on Leaks bot

• Component: Page Loading
• Status: NEW

[Adam Roben (:aroben)]

To see the leaks:

1. Go to http://build.webkit.org/LeaksViewer/?url=http%3A%2F%2Fbuild.webkit.org%2F%2Fresults%2FSnowLeopard%20Intel%20Leaks%2Fr81496%20(15670)%2F
2. Dig into malloc_zone_malloc > malloc > WTF::fastMalloc > RefCounted<DocumentLoader>::operator new

We're leaking a DocumentLoader! You can find a few other callstacks that stem from the DocumentLoader constructor (e.g., ApplicationCacheHost), too.

[Jessie Berlin]

<rdar://problem/9156333>

[Brady Eidson]

It'd awesome to know the first revision where this happened.

[Adam Roben (:aroben)]

I wonder if this could be the cause of some of the CFNetwork-related leaks we see on the bots, too?

[Adam Roben (:aroben)]

Looks like the leak happens in this range of tests (starting in fast/dom):

........................................................................................................................................................................................................................................................................
fast/dom/Attr ....
fast/dom/CSSStyleDeclaration .....
fast/dom/DOMException ....
fast/dom/DOMImplementation .....
fast/dom/Document ...................
fast/dom/Document/CaretRangeFromPoint ....
fast/dom/Element .................................
fast/dom/EntityReference .
fast/dom/Geolocation ...................................
fast/dom/HTMLAnchorElement .........
fast/dom/HTMLAreaElement .
fast/dom/HTMLBaseElement .
fast/dom/HTMLButtonElement/value .
fast/dom/HTMLDivElement/align .
fast/dom/HTMLDocument ..................
fast/dom/HTMLElement ..............................
fast/dom/HTMLFontElement .
fast/dom/HTMLFormElement ......
fast/dom/HTMLHeadElement .......
fast/dom/HTMLHtmlElement ..
fast/dom/HTMLImageElement ..........
fast/dom/HTMLInputElement ............
fast/dom/HTMLKeygenElement .
fast/dom/HTMLLabelElement ...
fast/dom/HTMLLabelElement/form .
fast/dom/HTMLLinkElement ...
fast/dom/HTMLMetaElement .
fast/dom/HTMLMeterElement ............
fast/dom/HTMLObjectElement ....
fast/dom/HTMLObjectElement/form ..
fast/dom/HTMLOptionElement ....
fast/dom/HTMLOutputElement ......
fast/dom/HTMLProgressElement .....
fast/dom/HTMLScriptElement .....................
fast/dom/HTMLSelectElement .........
fast/dom/HTMLStyleElement .
fast/dom/HTMLTableColElement .
fast/dom/HTMLTableElement ........
fast/dom/HTMLTableRowElement ..
fast/dom/HTMLTableSectionElement .
fast/dom/HTMLTextAreaElement .
fast/dom/Node ........
fast/dom/NodeList ...........
fast/dom/Orientation .
fast/dom/Range ..............................
fast/dom/Selection ..
fast/dom/SelectorAPI ..................
fast/dom/StyleSheet ................
fast/dom/Text .
fast/dom/TreeWalker ........
fast/dom/Window ...............................................................................
fast/dom/Window/Location ..................
fast/dom/beforeload ...................
fast/dom/getElementsByClassName ................
fast/dom/getElementsByClassName/array ....
fast/dynamic .......................................................................
fast/encoding ..........................................................................................
fast/encoding/GBK ............
fast/encoding/dumpAsText .
fast/encoding/hebrew .......
fast/events .............................
 ? checking for leaks in DumpRenderTree
 + 29 leaks (7840 bytes) were found, details in /Volumes/Data/WebKit-BuildSlave/snowleopard-intel-leaks/build/layout-test-results/DumpRenderTree3-leaks.txt

[David Kilzer (:ddkilzer)]

(In reply to comment #4)
> Looks like the leak happens in this range of tests (starting in fast/dom):

We need a way to bisect layout tests!  :)  Hmm...I guess that's Bug 16903.  See also Bug 16873.