Created attachment 783 [details]
Details of the failures

This is the "testrun.log" output stripped of all the PASS lines. It shows the
failures associated with this PR.

This patch:

http://lists.cs.uiuc.edu/pipermail/llvm-commits/Week-of-Mon-20070409/047707.html

fixes these three GEP upgrade issues:
test/Assembler/2002-08-19-BytecodeReader.llx
test/Assembler/2003-08-21-ConstantExprCast-Fold.llx
test/Assembler/2004-01-11-getelementptrfolding.llx

This patch:

http://lists.cs.uiuc.edu/pipermail/llvm-commits/Week-of-Mon-20070409/047710.html

fixes:

test/CFrontend/2006-01-16-BitCountIntrinsicsUnsigned.c

This patch:

http://lists.cs.uiuc.edu/pipermail/llvm-commits/Week-of-Mon-20070409/047711.html

fixes:

test/Transforms/IndVarsSimplify/exit_value_tests.llx

This patch:

http://lists.cs.uiuc.edu/pipermail/llvm-commits/Week-of-Mon-20070409/047713.html

fixes:

test/Transforms/IndVarsSimplify/exit_value_tests.llx

er .. fixes:
test/Transforms/GlobalOpt/memset.ll
(itself).

This patch:

http://lists.cs.uiuc.edu/pipermail/llvm-commits/Week-of-Mon-20070409/047716.html

fixes:

test/Transforms/GlobalDCE/2002-08-17-FunctionDGE.ll

This patch:
http://lists.cs.uiuc.edu/pipermail/llvm-commits/Week-of-Mon-20070409/047718.html
fixes:
test/CodeGen/Generic/vector-constantexpr.ll

test/CodeGen/CBackend/2004-02-15-PreexistingExternals.llx
just needed to have an intrinsic manually upgraded.

I have resolved 16 of the regressions for this bug. For the remainder, I need a
little help. Either I don't understand what the test is trying to do or it looks
 like there's an actual bug in some part of LLVM I'm not familiar with.

The remaining tests are:

FAIL: /proj/llvm/llvm-2/test/C++Frontend/2004-09-27-CompilerCrash.cpp: exit(1) 
FAIL: /proj/llvm/llvm-2/test/C++Frontend/2007-04-11-InlineStorageClassC++.cpp:
FAIL: /proj/llvm/llvm-2/test/CFrontend/2002-11-07-Redefinition.c: exit(1) 
FAIL: /proj/llvm/llvm-2/test/CFrontend/2003-07-22-ArrayAccessTypeSafety.c.tr:
FAIL: /proj/llvm/llvm-2/test/CodeGen/ARM/align.ll: exit(1) at line 2
FAIL: /proj/llvm/llvm-2/test/CodeGen/ARM/hello.ll: exit(1) at line 3
FAIL: /proj/llvm/llvm-2/test/CodeGen/Alpha/cmov.ll: exit(1) at line 1
FAIL: /proj/llvm/llvm-2/test/CodeGen/Alpha/ctlz.ll: exit(1) at line 2
FAIL: /proj/llvm/llvm-2/test/CodeGen/CBackend/2005-07-14-NegationToMinusMinus.ll
FAIL: /proj/llvm/llvm-2/test/CodeGen/X86/2004-02-14-InefficientStackPointer.llx:
FAIL: /proj/llvm/llvm-2/test/CodeGen/X86/fp-stack-ret.ll: exit(1) at line 4
FAIL: /proj/llvm/llvm-2/test/CodeGen/X86/fp_constant_op.llx: exit(1) at line 1
FAIL: /proj/llvm/llvm-2/test/CodeGen/X86/fp_load_fold.llx: exit(1) at line 1
FAIL: /proj/llvm/llvm-2/test/CodeGen/X86/sse-fcopysign.ll: exit(1) at line 1
FAIL: /proj/llvm/llvm-2/test/CodeGen/X86/sse-load-ret.ll: exit(1) at line 1

If the owners of these areas could take a look, I would appreciate it.

We're done to these now:

FAIL: /proj/llvm/llvm-2/test/CodeGen/ARM/align.ll: exit(1) at line 2
FAIL: /proj/llvm/llvm-2/test/CodeGen/ARM/hello.ll: exit(1) at line 3
FAIL: /proj/llvm/llvm-2/test/CodeGen/Alpha/cmov.ll: exit(1) at line 1
FAIL: /proj/llvm/llvm-2/test/CodeGen/Alpha/ctlz.ll: exit(1) at line 2
FAIL: /proj/llvm/llvm-2/test/CodeGen/X86/2004-02-14-InefficientStackPointer.llx:
FAIL: /proj/llvm/llvm-2/test/CodeGen/X86/fp-stack-ret.ll: exit(1) at line 4
FAIL: /proj/llvm/llvm-2/test/CodeGen/X86/fp_constant_op.llx: exit(1) at line 1
FAIL: /proj/llvm/llvm-2/test/CodeGen/X86/fp_load_fold.llx: exit(1) at line 1
FAIL: /proj/llvm/llvm-2/test/CodeGen/X86/sse-fcopysign.ll: exit(1) at line 1
FAIL: /proj/llvm/llvm-2/test/CodeGen/X86/sse-load-ret.ll: exit(1) at line 1

-Chris

Fixing the llvm-upgrade problem for test/CodeGen/Generic/vector-constantexpr.ll
was only half the story. llc also failed in getIntrinsicID because the function
has no name. This patch fixes it:

http://lists.cs.uiuc.edu/pipermail/llvm-commits/Week-of-Mon-20070416/047738.html

I fixed an Alpha feature option bug. The list is down to:

FAIL: /proj/llvm/llvm-2/test/CodeGen/ARM/align.ll: exit(1) at line 2
FAIL: /proj/llvm/llvm-2/test/CodeGen/ARM/hello.ll: exit(1) at line 3
FAIL: /proj/llvm/llvm-2/test/CodeGen/Alpha/cmov.ll: exit(1) at line 1
FAIL: /proj/llvm/llvm-2/test/CodeGen/X86/2004-02-14-InefficientStackPointer.llx:
FAIL: /proj/llvm/llvm-2/test/CodeGen/X86/fp-stack-ret.ll: exit(1) at line 4
FAIL: /proj/llvm/llvm-2/test/CodeGen/X86/fp_constant_op.llx: exit(1) at line 1
FAIL: /proj/llvm/llvm-2/test/CodeGen/X86/fp_load_fold.llx: exit(1) at line 1
FAIL: /proj/llvm/llvm-2/test/CodeGen/X86/sse-fcopysign.ll: exit(1) at line 1
FAIL: /proj/llvm/llvm-2/test/CodeGen/X86/sse-load-ret.ll: exit(1) at line 1

I fixed all the X86 test cases. The list is now down to:

FAIL: /proj/llvm/llvm-2/test/CodeGen/ARM/align.ll: exit(1) at line 2
FAIL: /proj/llvm/llvm-2/test/CodeGen/ARM/hello.ll: exit(1) at line 3
FAIL: /proj/llvm/llvm-2/test/CodeGen/Alpha/cmov.ll: exit(1) at line 1

Fixed CodeGen/ARM/hello.ll and align.ll.

Chris asked me to attach the results of llvm/test on powerpc-apple-darwin8, checkout as of 9:25PDT:

Running /Volumes/MacOS9/gcc/llvm-gcc-apple/llvm/test/C++Frontend/dg.exp ...
FAIL: /Volumes/MacOS9/gcc/llvm-gcc-apple/llvm/test/C++Frontend/2003-08-28-ForwardType.cpp: 
exit(1) at line 1
while running: /Volumes/MacOS9/gcc/llvm-gcc-apple.0402.install/bin/llvm-g++ -emit-llvm -S /
Volumes/MacOS9/gcc/llvm-gcc-apple/llvm/test/C++Frontend/2003-08-28-For\
wardType.cpp -o - | llvm-as -f -o /dev/null
/Volumes/MacOS9/gcc/llvm-gcc-apple/llvm/test/C++Frontend/2003-08-28-ForwardType.cpp:4: 
error: 'operator new' takes type 'size_t' ('long unsigned int') as first p\
arameter
/Volumes/MacOS9/gcc/llvm-gcc-apple/llvm/test/C++Frontend/2003-08-28-ForwardType.cpp:4: 
warning: inline function 'void* operator new(long unsigned int, void*)' us\
ed but never defined
Running /Volumes/MacOS9/gcc/llvm-gcc-apple/llvm/test/CFrontend/dg.exp ...
FAIL: /Volumes/MacOS9/gcc/llvm-gcc-apple/llvm/test/CFrontend/2006-09-25-DebugFilename.c.tr: 
exit(1) at line 1
while running: /Volumes/MacOS9/gcc/llvm-gcc-apple.0402.install/bin/llvm-gcc -emit-llvm -xc /
Volumes/MacOS9/gcc/llvm-gcc-apple/llvm/test/CFrontend/2006-09-25-Debu\
gFilename.c.tr -S -o /dev/null |&  grep fluffy | grep 2006-09-25-DebugFilename.c.tr
child process exited abnormally
XPASS: /Volumes/MacOS9/gcc/llvm-gcc-apple/llvm/test/CFrontend/bit-accurate-int.c
FAIL: /Volumes/MacOS9/gcc/llvm-gcc-apple/llvm/test/CFrontend/cast-to-bool.c: exit(1) at line 1
while running: /Volumes/MacOS9/gcc/llvm-gcc-apple.0402.install/bin/llvm-gcc -emit-llvm -S /
Volumes/MacOS9/gcc/llvm-gcc-apple/llvm/test/CFrontend/cast-to-bool.c -\
o - | grep {trunc.*to i8}
child process exited abnormally
Running /Volumes/MacOS9/gcc/llvm-gcc-apple/llvm/test/CodeGen/ARM/dg.exp ...
FAIL: /Volumes/MacOS9/gcc/llvm-gcc-apple/llvm/test/CodeGen/ARM/long.ll: exit(1) at line 5
while running: llvm-upgrade < /Volumes/MacOS9/gcc/llvm-gcc-apple/llvm/test/CodeGen/ARM/
long.ll | llvm-as | llc -march=arm | grep subs | wc -l | grep 1
child process exited abnormally
Running /Volumes/MacOS9/gcc/llvm-gcc-apple/llvm/test/CodeGen/Alpha/dg.exp ...
FAIL: /Volumes/MacOS9/gcc/llvm-gcc-apple/llvm/test/CodeGen/Alpha/ctlz.ll: exit(1) at line 2
while running: llvm-as < /Volumes/MacOS9/gcc/llvm-gcc-apple/llvm/test/CodeGen/Alpha/ctlz.ll | llc 
-march=alpha -mattr=+CIX | grep -i ctlz
'+cix' is not a recognized feature for this target (ignoring feature)

Running /Volumes/MacOS9/gcc/llvm-gcc-apple/llvm/test/CodeGen/Generic/dg.exp ...
FAIL: /Volumes/MacOS9/gcc/llvm-gcc-apple/llvm/test/CodeGen/Generic/vector-constantexpr.ll: 
signal(SIGBUS) at line 1
while running: llvm-upgrade < /Volumes/MacOS9/gcc/llvm-gcc-apple/llvm/test/CodeGen/Generic/
vector-constantexpr.ll | llvm-as | llc
        .machine ppc970
.section __TEXT,__textcoal_nt,coalesced,pure_instructions
.section __TEXT,__symbol_stub1,symbol_stubs,pure_instructions,16
.text

        .subsections_via_symbols
child killed: bus error
Running /Volumes/MacOS9/gcc/llvm-gcc-apple/llvm/test/CodeGen/X86/dg.exp ...
XPASS: /Volumes/MacOS9/gcc/llvm-gcc-apple/llvm/test/CodeGen/X86/vec_call.ll
Running /Volumes/MacOS9/gcc/llvm-gcc-apple/llvm/test/Transforms/GlobalDCE/dg.exp ...
FAIL: /Volumes/MacOS9/gcc/llvm-gcc-apple/llvm/test/Transforms/GlobalDCE/2002-08-17-
FunctionDGE.ll: exit(1) at line 1
while running: llvm-upgrade < /Volumes/MacOS9/gcc/llvm-gcc-apple/llvm/test/Transforms/
GlobalDCE/2002-08-17-FunctionDGE.ll | llvm-as | opt -globaldce | llvm-dis |\
not grep foo
llvm-upgrade: <stdin>:13: error: Reference to an invalid definition: 'foo' of type 'i32 *'
llvm-as: <stdin>:1,0: <stdin>:1: error: parse error, expecting `GLOBAL' or `CONSTANT' while reading 
token: 'llvm-upgrade'
opt: Standard Input is empty!
llvm-dis: Standard Input is empty!

Running /Volumes/MacOS9/gcc/llvm-gcc-apple/llvm/test/Transforms/LICM/dg.exp ...
FAIL: /Volumes/MacOS9/gcc/llvm-gcc-apple/llvm/test/Transforms/LICM/scalar_promote.ll: signal
(SIGPIPE) at line 1
while running: llvm-upgrade < /Volumes/MacOS9/gcc/llvm-gcc-apple/llvm/test/Transforms/LICM/
scalar_promote.ll | llvm-as | opt  -licm -stats |&  grep {memory locat\
ions promoted to register}
Binary file (standard input) matches

Devang,

Many of those problems I've already fixed. Please update, rerun and post again.
The ones that remain are the CFE ones.  

Reid.

After my last round of changes (to get rid of &&), add two more:

test/CFrontend/2006-09-25-DebugFilename.c.tr
test/Transforms/PredicateSimplifier/2006-10-22-IntOr.ll

Removed CodeGen/X86/sse-load-ret.ll xfail.

test/CFrontend/2006-09-25-DebugFilename.c.tr
anton fixed this

test/Transforms/PredicateSimplifier/2006-10-22-IntOr.ll
I fixed this.

I also fixed several others.  Dejagnu is now finally clean for me.  Should we close this?

-Chris

Let me just verify it here and if its clean I'll close.

Andrew,

There is only one regression remaining in this PR. It is:
 test/CodeGen/Alpha/cmov.ll

The RUN: lines are:

; RUN: llvm-upgrade < %s | llvm-as | llc -march=alpha | not grep cmovlt
; RUN: llvm-upgrade < %s | llvm-as | llc -march=alpha | grep cmoveq

so no cmovlt instructions should be found. However, they are. The assembly
output is below. Please correct this test case/code gen at your earliest
convenience. When you do, please un-XFAIL the test case.

Thanks, Reid.

Alpha Assembly output:
       .arch ev6
        .set noat

.text
        .align  4
        .globl cmov_lt
        .ent cmov_lt
cmov_lt:
        ldah $29,0($27)         !gpdisp!1
        lda $29,0($29)          !gpdisp!1
        $cmov_lt..ng:

        lda $0,10($31)
        cmovlt $17,$16,$0
        ret $31,($26),1
        .end cmov_lt

        .align  4
        .globl cmov_const
        .ent cmov_const
cmov_const:
        ldah $29,0($27)         !gpdisp!2
        lda $29,0($29)          !gpdisp!2
        $cmov_const..ng:

        cmplt $16,$17,$0
        cmoveq $0,10,$18
        bis $18,$18,$0
        ret $31,($26),1
        .end cmov_const

        .align  4
        .globl cmov_lt2
        .ent cmov_lt2
cmov_lt2:
        ldah $29,0($27)         !gpdisp!3
        lda $29,0($29)          !gpdisp!3
        $cmov_lt2..ng:

        cmovgt $17,10,$16
        bis $16,$16,$0
        ret $31,($26),1
        .end cmov_lt2

this is done