[Webkit-unassigned] [Bug 145062] New: LICM should be more careful about settings node origin's forExit

[bugzilla-daemon at webkit.org]

https://bugs.webkit.org/show_bug.cgi?id=145062

            Bug ID: 145062
           Summary: LICM should be more careful about settings node
                    origin's forExit
    Classification: Unclassified
           Product: WebKit
           Version: 528+ (Nightly build)
          Hardware: Unspecified
                OS: Unspecified
            Status: NEW
          Severity: Normal
          Priority: P2
         Component: JavaScriptCore
          Assignee: webkit-unassigned at lists.webkit.org
          Reporter: basile_clement at apple.com

This can be hit by running e.g. sunspider/access-nsieve with

DYLD_FRAMEWORK_PATH=WebKitBuild/Debug WebKitBuild/Debug/jsc --forceEagerCompilation=true --useFTLJIT=true --dumpGraphAtEachPhase=true

In this case, a Phi node is converted into a JSConstant by the DFGConstantFoldingPhase, and doesn't have a NodeOrigin.
Then it gets LICM'd, which unconditionally sets the NodeOrigin's forExit, and now we have a NodeOrigin with a set forExit and unset semantic, which we assert against in various places.

-- 
You are receiving this mail because:
You are the assignee for the bug.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.webkit.org/pipermail/webkit-unassigned/attachments/20150515/6c520790/attachment-0001.html>