[LLVMbugs] [Bug 15056] New: LLVM LIT test fails on Win8 (ExecutionEngine/MCJIT/2003-05-06-LivenessClobber.ll)

[bugzilla-daemon at llvm.org]

http://llvm.org/bugs/show_bug.cgi?id=15056

             Bug #: 15056
           Summary: LLVM LIT test fails on Win8
                    (ExecutionEngine/MCJIT/2003-05-06-LivenessClobber.ll)
           Product: libraries
           Version: trunk
          Platform: PC
        OS/Version: Windows NT
            Status: NEW
          Severity: normal
          Priority: P
         Component: Support Libraries
        AssignedTo: unassignedbugs at nondot.org
        ReportedBy: marina.yatsina at intel.com
                CC: llvmbugs at cs.uiuc.edu
    Classification: Unclassified


The following LLVM LIT test fails when running on Win8: 
ExecutionEngine/MCJIT/2003-05-06-LivenessClobber.ll 
This failures happens in the current llvm trunk (rev. 173279). 
This test failure might be related to transition to MCJIT (This test was
previously ran with the regular execution enginen and passend on LLVM 3.0
release) or the DynamicLibrary which is responsible for loading modules. 

Test's command line: 
not.exe lli.exe -mtriple=x86_64-pc-win32-elf -use-mcjit
c:\llvm\test\ExecutionEngine\MCJIT\2003-05-06-LivenessClobber.ll 

The lli.exe is expected to fail (return val != 0). 
It seems that lli.exe runs the function from the LIT tests, which returns 1,
and then runs the "exit(int)" windows system call with the given value. 

on Win7 this system call terminates the process with return code "1", while on
Win8 this system call terminates the process with return code "0". 

After investigation I saw that on Win7 the exit() system call was loaded from
C:\Windows\System32\msvcr100d.dll, while  on Win8 it was loaded from
C:\Windows\System32\KernelBase.dll and they were 2 completley different system
calls and implemented differently.

-- 
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.