Bug 25782 – Allow custom memory allocation control in JavaScriptCore

• Status: RESOLVED
• Resolution: FIXED
• Component: JavaScriptCore

[Zoltan Horvath]

FastAllocBase class landed in the trunk (bug #20422) at r42344. 
Now we have to continue the work with inherite some classes from it. 
TODOs:
- We have to inherit every classes from FastAllocBase which has been instantiated by 'new'. 
- We have to put FastAllocBase into inheritance tree as far up as possible.

First I file a patch for one class called HashTable. After it'll be in the trunk I'll file a list which classes instantiated by 'new'.

[Zoltan Horvath]

Created attachment 30319 [details]
proposed patch for HashTable

[Maciej Stachowiak]

Comment on attachment 30319 [details]
proposed patch for HashTable

r=me

[Brent Fulgham]

Landed in @r44161.

[Zoltan Horvath]

I've analysed the source and noticed that I inherited the wrong HashTable class. Sorry, I've filed a new patch which contains the correction of the mistake and the allocation support for the 'real' HashTable. Details in the patch.

[Zoltan Horvath]

Created attachment 30868 [details]
proposed patch for HashTable struct

[Zoltan Horvath]

Simplier method is to change 'new' call to fastNew, and 'delete' to fastDelete. (In this case I need to modify only where HashTable instantiated by 'new' and destructed with delete. JSGlobalData.cpp:106-112) 
If you wish I could attach a patch with this solution.

[Zoltan Horvath]

Created attachment 30871 [details]
proposed patch for HashTable struct

Based on Darin's https://bugs.webkit.org/show_bug.cgi?id=20422#c36 comment, I filed the fastNew/fastDelete solution.

[Darin Adler]

Comment on attachment 30871 [details]
proposed patch for HashTable struct

r=me

[Brent Fulgham]

Assign for landing.

[Brent Fulgham]

Landed in @r44417.