Bug 26750 – Allow custom memory allocation control for JavaScriptCore/runtime directory's classes

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

[Zoltan Horvath]

Inherits runtime directory's classes from FastAllocBase (bug #20422), which has been instantiated by 'new':

class IdentifierTable
-> instantiated in runtime/Identifier.cpp: line 70

stuct RegExpObjectData
-> runtime/RegExpObject.cpp: line 62

class HashEntry
-> runtime/Lookup.h: line 47

struct DateInstance::Cache
-> runtime/DateInstance.cpp: line 33

class UString
-> runtime/UString.cpp: line 212

struct RegExpConstructorPrivate
-> runtime/RegExpConstructor.cpp: line 108

class ScopeChain
-> untime/JSFunction.h: line 109

[Zoltan Horvath]

Created attachment 31916 [details]
proposed patch

[Zoltan Horvath]

Created attachment 31917 [details]
updated proposed patch

[Eric Seidel (no email)]

Comment on attachment 31917 [details]
updated proposed patch

Looks fine.  I still think we could just land these all as one big patch.

[Eric Seidel (no email)]

Committing to http://svn.webkit.org/repository/webkit/trunk ...
	M	JavaScriptCore/ChangeLog
	M	JavaScriptCore/runtime/DateInstance.cpp
	M	JavaScriptCore/runtime/Identifier.cpp
	M	JavaScriptCore/runtime/Lookup.h
	M	JavaScriptCore/runtime/RegExpConstructor.cpp
	M	JavaScriptCore/runtime/RegExpObject.h
	M	JavaScriptCore/runtime/ScopeChain.h
	M	JavaScriptCore/runtime/UString.h
Committed r45259
http://trac.webkit.org/changeset/45259

[Eric Seidel (no email)]

Rolling out http://trac.webkit.org/changeset/45259.  Caused:
http://build.webkit.org/builders/Leopard%20Intel%20Debug%20%28Tests%29/builds/2322
http://build.webkit.org/results/Leopard%20Intel%20Debug%20(Tests)/r45259%20(2322)/results.html

plugins/undefined-property-crash.html
ASSERTION FAILED: s <= HeapConstants<heapType>::cellSize
(/Volumes/Big/WebKit-BuildSlave/leopard-intel-debug-tests/build/JavaScriptCore/runtime/Collector.cpp:278 void* JSC::Heap::heapAllocate(size_t) [with JSC::HeapType heapType = PrimaryHeap])

[Eric Seidel (no email)]

Comment on attachment 31917 [details]
updated proposed patch

Caused crash on the bots.

[Eric Seidel (no email)]

Committing to http://svn.webkit.org/repository/webkit/trunk ...
	M	JavaScriptCore/ChangeLog
	M	JavaScriptCore/runtime/DateInstance.cpp
	M	JavaScriptCore/runtime/Identifier.cpp
	M	JavaScriptCore/runtime/Lookup.h
	M	JavaScriptCore/runtime/RegExpConstructor.cpp
	M	JavaScriptCore/runtime/RegExpObject.h
	M	JavaScriptCore/runtime/ScopeChain.h
	M	JavaScriptCore/runtime/UString.h
Committed r45269
http://trac.webkit.org/changeset/45269

[Eric Seidel (no email)]

I need to add a --no-close option to my landing script.

[Zoltan Horvath]

Created attachment 33606 [details]
updated proposed patch

I need to send one patch for one class, so we will see which causes the assertion.

[Darin Adler]

I'm concerned, because this seems to indicate that adding FastAllocBase as a base class is making some objects bigger!

[Zoltan Horvath]

I think I made a mistake in the r-d patch and inherit a class which has been instantiated by JS's GC also and that caused the problem.

[Zoltan Horvath]

I've checked the problem: I've inherited DateInstance::Cache form FastAllocBase but DataInstance is a JSObject so this caused the problem. DataInstance::Cache doesn't need to be inherited.

[Eric Seidel (no email)]

Comment on attachment 33606 [details]
updated proposed patch

Assuming everything still builds.  Does this need to be public FastAllocBase?

[Adam Barth]

Comment on attachment 33606 [details]
updated proposed patch

Clearing review flag on attachment: 33606

Committing to http://svn.webkit.org/repository/webkit/trunk ...
	M	JavaScriptCore/ChangeLog
	M	JavaScriptCore/runtime/RegExpObject.h
Committed r46907
	M	JavaScriptCore/runtime/RegExpObject.h
	M	JavaScriptCore/ChangeLog
r46907 = 44404206ca651e03bd169e7e01bca730b59bcd41 (trunk)
No changes between current HEAD and refs/remotes/trunk
Resetting to the latest refs/remotes/trunk
http://trac.webkit.org/changeset/46907

[Adam Barth]

All reviewed patches have been landed.  Closing bug.