Bug 99309 – Web Inspector: convert manual size calculation of different WebKit things into MemoryInstrumentation.

• Component: Web Inspector (Deprecated)
• Status: RESOLVED FIXED

[Ilya Tikhonovsky]

MemoryInstrumentation is our primary method for WebCore internal classes.
But due to historical reasons sizes for some objects  were calculated manually.
I think we could fix this.

[Ilya Tikhonovsky]

Created attachment 168684 [details]
Patch

[Ilya Tikhonovsky]

Created attachment 168685 [details]
Patch

[Build Bot]

Comment on attachment 168685 [details]
Patch

Attachment 168685 [details] did not pass win-ews (win):
Output: http://queues.webkit.org/results/14300666

[Early Warning System Bot]

Comment on attachment 168685 [details]
Patch

Attachment 168685 [details] did not pass qt-ews (qt):
Output: http://queues.webkit.org/results/14289634

[Yury Semikhatsky]

Comment on attachment 168685 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=168685&action=review

> Source/WebCore/inspector/InspectorMemoryAgent.cpp:156
> +static String nodeName(Node* node)

No need to make it static as it is already in anonymous namespace.

[Alexei Filippov]

Comment on attachment 168685 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=168685&action=review

> Source/WebCore/inspector/front-end/NativeMemorySnapshotView.js:-479
> -                    unusedSize -= child.children[j].size;

Why do you need this *.Unused record? What's wrong with calculating it? What's if *.Unused is not present? It seems that unusedSize will be zero, e.g. for JSHeap.

[Build Bot]

Comment on attachment 168685 [details]
Patch

Attachment 168685 [details] did not pass mac-ews (mac):
Output: http://queues.webkit.org/results/14290676

[Early Warning System Bot]

Comment on attachment 168685 [details]
Patch

Attachment 168685 [details] did not pass qt-wk2-ews (qt):
Output: http://queues.webkit.org/results/14288730

[Gyuyoung Kim]

Comment on attachment 168685 [details]
Patch

Attachment 168685 [details] did not pass efl-ews (efl):
Output: http://queues.webkit.org/results/14292740

[Ilya Tikhonovsky]

Committed r131299: <http://trac.webkit.org/changeset/131299>

[Ilya Tikhonovsky]

(In reply to comment #6)
> (From update of attachment 168685 [details])
> View in context: https://bugs.webkit.org/attachment.cgi?id=168685&action=review
> 
> > Source/WebCore/inspector/front-end/NativeMemorySnapshotView.js:-479
> > -                    unusedSize -= child.children[j].size;
> 
> Why do you need this *.Unused record? What's wrong with calculating it? What's if *.Unused is not present? It seems that unusedSize will be zero, e.g. for JSHeap.

This patch flips the code from manual block building to the automatic builder.
The builder builds parent MemoryBlocks on the fly and calculates their sizes.
As a result when I use automatic builder code for JSHeap I need to add two blocks, JSHeap.Used and JSHeap.Unused. The builder makes parent JSHeap block automatically and assigns the right size to it. Latter, on the front end size, we use JSHeap size as the allocated size and JSHeap.Unused as unused size for JSHeap.

[Hans Muller]

Created attachment 168737 [details]
[please ignore]

Added FloatRect::extend() to simplify writing loops that accumulate the bounding box for a sequence of vertices.  For example:

FloatRect box(vertices[0], FloatSize(0,0));
for (unsigned i = 1; i < vertices.length; i++)
    box.extend(vertices[i]);

ExclusionShape and ExclusionPolygon now use FloatRect::extend().

[Hans Muller]

(In reply to comment #12)
> Created an attachment (id=168737) [details]

My apologies for accidentally attaching this patch/comment to the wrong bug.