[LLVMbugs] [Bug 4223] New: Improper handling of "undef" type in simplifycfg

bugzilla-daemon at cs.uiuc.edu bugzilla-daemon at cs.uiuc.edu
Sun May 17 19:00:48 PDT 2009


http://llvm.org/bugs/show_bug.cgi?id=4223

           Summary: Improper handling of "undef" type in simplifycfg
           Product: new-bugs
           Version: unspecified
          Platform: Macintosh
        OS/Version: MacOS X
            Status: NEW
          Severity: normal
          Priority: P2
         Component: new bugs
        AssignedTo: unassignedbugs at nondot.org
        ReportedBy: viridia at gmail.com
                CC: llvmbugs at cs.uiuc.edu


It seems that variables whose value is "undef" aren't being handled correctly
in a number of places. When I pass the following *invalid* code to LLVM, I
would expect LLVM to detect the error in the 4th instruction (the store to
undef):

define %2 @"throwSomething()"() {
entry:
        %0 = call %tart.core.Exception* @tart.core.Exception.type.alloc()      
        ; <%tart.core.Exception*> [#uses=2]
        %construct = call %2
@"tart.core.Exception.construct()"(%tart.core.Exception* %0)               ;
<%2> [#uses=0]
        %1 = bitcast %tart.core.Exception* %0 to %3*            ; <%3*>
[#uses=2]
        store %3* %1, %3** undef
        %2 = getelementptr %3* %1, i32 0, i32 1         ; <%4*> [#uses=1]
        %throw = call i32 @_Unwind_RaiseException(%4* %2)               ; <i32>
[#uses=0]
        unreachable
}

However, that is not what is happening.

The simplifycfg pass appears to fail silently on this line - that is, it
doesn't report an error, but neither does it generate any code from the 4th
line on. The resulting function contains only the instructions before the
"store", and the rest is just chopped off.

The function verifier pass did not report any error either when it ran on this
function either.


-- 
Configure bugmail: http://llvm.org/bugs/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are on the CC list for the bug.



More information about the llvm-bugs mailing list