x86-64 test/CodeGen/X86/2006-05-02-InstrSched2.ll fails

[edwintorok]

Bugzilla Link	2156
Resolution	WORKSFORME
Resolved on	Apr 19, 2008 02:18
Version	unspecified
OS	Linux
Attachments	fail.bc to reproduce

Extended Description

When running make check on r48371, Linux x86-64:

FAIL: /home/edwin/llvm-svn/llvm/test/CodeGen/X86/2006-05-02-InstrSched2.ll
Failed with exit(1) at line 1
while running: llvm-upgrade < /home/edwin/llvm-svn/llvm/test/CodeGen/X86/2006-05-02-InstrSched2.ll | llvm-as | llc -march=x86 -stats |& grep asm-printer | grep 13
child killed: segmentation violation
child process exited abnormally

If I run it manually I get:

$ llvm-upgrade < /home/edwin/llvm-svn/llvm/test/CodeGen/X86/2006-05-02-InstrSched2.ll | llvm-as | llc -march=x86 -stats | grep asm-printer | grep 13
llc[0x9aed66]
/lib/libc.so.6[0x3932831f70]
llc[0x82db2f]
llc(_ZN4llvm19MachineFunctionPass13runOnFunctionERNS_8FunctionE+0x29)[0x5f00e9]
llc(_ZN4llvm13FPPassManager13runOnFunctionERNS_8FunctionE+0x231)[0x94ca11]
llc(_ZN4llvm23FunctionPassManagerImpl3runERNS_8FunctionE+0x1c9)[0x94d639]
llc(_ZN4llvm19FunctionPassManager3runERNS_8FunctionE+0x31)[0x94d791]
llc(main+0x993)[0x4b42d3]
/lib/libc.so.6(__libc_start_main+0xfa)[0x393281e1ea]
llc(cos+0xf1)[0x4b1cd9]

If I remove either of '-stats' or 'o -' option from llc it doesn't crash anymore.

I tried to get a reduced testcase with bugpoint, but running llc on bugpoint-reduced-simplified.bc didn't reproduce the crash (although it was crashing when called from bugpoint).

Here's a backtrace from a debug build (on x86-64, Linux):
(gdb) r -march=x86 fail.bc -f -o - -stats
Starting program: /home/edwin/llvm-svn/llvm/Debug/bin/llc -march=x86 fail.bc -f -o - -stats
[Thread debugging using libthread_db enabled]
[New Thread 0x7f294d0c76f0 (LWP 14326)]

Program received signal SIGSEGV, Segmentation fault.
[Switching to Thread 0x7f294d0c76f0 (LWP 14326)]
0x000000000089e16a in llvm::ilist_traitsllvm::MachineInstr::getNext (N=0x91)
at /home/edwin/llvm-svn/llvm/include/llvm/CodeGen/MachineBasicBlock.h:37
37 static MachineInstr* getNext(MachineInstr* N) { return N->Next; }
(gdb) bt
#​0 0x000000000089e16a in llvm::ilist_traitsllvm::MachineInstr::getNext (N=0x91)
at /home/edwin/llvm-svn/llvm/include/llvm/CodeGen/MachineBasicBlock.h:37
#​1 0x000000000089f98c in llvm::ilist_iteratorllvm::MachineInstr::operator* (this=0x7fff550ed080)
at /home/edwin/llvm-svn/llvm/include/llvm/ADT/ilist:121
#​2 0x000000000089f9d3 in llvm::ilist_iteratorllvm::MachineInstr::operator-> (this=0x7fff550ed080)
at /home/edwin/llvm-svn/llvm/include/llvm/ADT/ilist:124
#​3 0x0000000000ae0837 in runOnMachineFunction (this=0x1157a40, MF=@0x115f7a0) at TwoAddressInstructionPass.cpp:204
#​4 0x0000000000898ae9 in llvm::MachineFunctionPass::runOnFunction (this=0x1157a40, F=@0x11454c0)
at /home/edwin/llvm-svn/llvm/include/llvm/CodeGen/MachineFunctionPass.h:41
#​5 0x0000000000c17e76 in llvm::FPPassManager::runOnFunction (this=0x1149760, F=@0x11454c0) at PassManager.cpp:1167
#​6 0x0000000000c1809c in llvm::FunctionPassManagerImpl::run (this=0x1144830, F=@0x11454c0) at PassManager.cpp:1125
#​7 0x0000000000c181fa in llvm::FunctionPassManager::run (this=0x7fff550ed320, F=@0x11454c0) at PassManager.cpp:1070
#​8 0x000000000077d1dc in main (argc=7, argv=0x7fff550ed548) at llc.cpp:296
(gdb)

[edwintorok]

This test no longer fails, and it doesn't fail with the attached testcase either.
Closing.