385549 - various cycle collector debugging patches, round 2

Status: RESOLVED FIXED

(Core :: XPCOM, defect)

Description

[David Baron :dbaron: (⌚️UTC-4, no longer working on Mozilla)]

Here's another round of cycle collector debugging patches.  My previous round was bug 383234.  Except for the first one, which I've had around for a while, these are the patches I used to debug the leak regression from bug 381199, which is now filed as bug 385548.

Comment 1

[David Baron :dbaron: (⌚️UTC-4, no longer working on Mozilla)]

Attachment: log the native object (mIdentity) in XPCWrappedNative::DebugDump

This makes it easier to correlate pointers between different types of debugging output.

Comment 2

[David Baron :dbaron: (⌚️UTC-4, no longer working on Mozilla)]

Attachment: shouldBeFreed call for outer window

This adds an nsCycleCollector_DEBUG_shouldBeFreed for the outer window, useful for ExplainLiveExpectedGarbage.  (These don't need to be perfect, but they're true in most cases.)

Comment 3

[David Baron :dbaron: (⌚️UTC-4, no longer working on Mozilla)]

Attachment: shouldBeFreed call for elements

This adds an nsCycleCollector_DEBUG_shouldBeFreed for elements, useful for ExplainLiveExpectedGarbage.  (These don't need to be perfect, but they're true in most cases.)

Comment 4

[David Baron :dbaron: (⌚️UTC-4, no longer working on Mozilla)]

Attachment: more accurate warnings about failure to suspect or unlink

This is the fun one.

The bugs mentioned in comment 0 are the first time I hit the warning about failure to unlink or suspect, and they weren't too useful because they print warnings for the entire graph of objects that would be freed.  This determines the strongly connected components of the graph and prints them only for the objects in the root components, which are the ones we care about.  And it even prints component identifiers so you can see which are related to each other.

Comment 5

[David Baron :dbaron: (⌚️UTC-4, no longer working on Mozilla)]

All patches now checked in.