Bug 150042

Summary: CodeBlock write barriers should be precise
Product: WebKit Reporter: Geoffrey Garen <ggaren>
Component: New BugsAssignee: Geoffrey Garen <ggaren>
Status: REOPENED    
Severity: Normal CC: fpizlo, mark.lam, mhahnenb
Priority: P2    
Version: WebKit Nightly Build   
Hardware: Unspecified   
OS: Unspecified   
Attachments:
Description Flags
Patch none

Geoffrey Garen
Reported 2015-10-12 11:30:30 PDT
CodeBlock write barriers should be precise
Attachments
Patch (4.97 KB, patch)
2015-10-12 11:36 PDT, Geoffrey Garen 		no flags
DetailsFormatted DiffDiff
View All Add attachment proposed patch, testcase, etc.
Geoffrey Garen
Comment 1 2015-10-12 11:36:58 PDT
Created attachment 262909 [details] Patch
Mark Lam
Comment 2 2015-10-12 14:21:21 PDT
Comment on attachment 262909 [details] Patch View in context: https://bugs.webkit.org/attachment.cgi?id=262909&action=review > Source/JavaScriptCore/dfg/DFGOSRExitCompilerCommon.cpp:278 > + codeBlocksToWriteBarrier.add(jit.codeBlock()->baselineAlternative()); Is this necessary? I may be wrong but wouldn't this baseline codeBlock be covered when you iterate the inlineCallFrame's directCaller back to the top level (which you are doing below)?
Geoffrey Garen
Comment 3 2015-10-12 14:49:06 PDT
> Is this necessary? I may be wrong but wouldn't this baseline codeBlock be > covered when you iterate the inlineCallFrame's directCaller back to the top > level (which you are doing below)? The inlineCallFrame stack only includes inlined functions, and not the baseline CodeBlock.
Saam Barati
Comment 4 2015-10-13 12:19:11 PDT
Comment on attachment 262909 [details] Patch View in context: https://bugs.webkit.org/attachment.cgi?id=262909&action=review r=me >> Source/JavaScriptCore/dfg/DFGOSRExitCompilerCommon.cpp:278 >> + codeBlocksToWriteBarrier.add(jit.codeBlock()->baselineAlternative()); > > Is this necessary? I may be wrong but wouldn't this baseline codeBlock be covered when you iterate the inlineCallFrame's directCaller back to the top level (which you are doing below)? I don't think the loop below covers it.
Geoffrey Garen
Comment 5 2015-10-13 13:09:26 PDT
Committed r191003: <http://trac.webkit.org/changeset/191003>
Geoffrey Garen
Comment 6 2015-10-15 14:53:16 PDT
Rolled out in <http://trac.webkit.org/changeset/191139>.
Note You need to log in before you can comment on or make changes to this bug.