[XCOFF] change the decoding of External symbol's function auxiliary entry in XCOFF32 for llvm-readobj

[DiggerLin]

Summary

1. llvm-readobj decide whether to decode the external symbol's function auxiliary entry based on whether symbol is function or not currently. But the XCOFFSymbolRef::isFunction() do not work properly when -ffunction-sections is enabled. we will not decode the function auxiliary entry based on the XCOFFSymbolRef::isFunction()

2. we will decode the function auxiliary entry based on following:

According to the https://www.ibm.com/docs/en/aix/7.2?topic=formats-xcoff-object-file-format#XCOFF__c0f91ad419jbau

In XCOFF32, there are only "one csect Auxiliary Entry" and "a function auxiliary symbol table entry" for the C_EXT, C_WEAKEXT, and C_HIDEXT Symbols. and By convention, the csect auxiliary entry in an XCOFF32 file must be the last auxiliary entry for any external symbol that has more than one auxiliary entry( that means for the C_EXT, C_WEAKEXT, and C_HIDEXT Symbols. if there more than one auxiliary Entries. we look the last one as csect auxiliary entry. and others auxiliary entries as function entries).

[hubert.reinterpretcast]

llvm/test/tools/llvm-readobj/XCOFF/symbols-invalid.test

12–13

Minor: May want to re-number the cases.

llvm/tools/llvm-readobj/XCOFFDumper.cpp

625

I think we should have functional tests that demonstrate the new behaviour.

[jhenderson]

In the patch description, you've stated "In XCOFF32 ..." twice. Should one of them be "In XCOFF64"?

[DiggerLin]

In D136950#3898550, @jhenderson wrote:

In the patch description, you've stated "In XCOFF32 ..." twice. Should one of them be "In XCOFF64"?

There is flag name "x_auxtype" in XCOFF64 to indicate the symbol auxiliary entry type(but XCOFF32 do not have the field.)

[DiggerLin]

llvm/tools/llvm-readobj/XCOFFDumper.cpp

625

I added symbol which has two function auxiliary and a Csect auxiliary entries in the tools/llvm-readobj/XCOFF/symbols.test

[DiggerLin]

address comment

[hubert.reinterpretcast]

llvm/test/tools/llvm-readobj/XCOFF/symbols-invalid.test

12–13

The temp file names are also numbered (should re-sync to match the case numbers).

[DiggerLin]

address comment.

llvm/test/tools/llvm-readobj/XCOFF/symbols-invalid.test

12–13

thanks

[hubert.reinterpretcast]

LGTM with comment.

llvm/test/tools/llvm-readobj/XCOFF/symbols.test

141

One of the changes made in this patch is that the DT_FCN bit is not checked.

339

llvm/test/tools/llvm-readobj/XCOFF/symbols-invalid.test

 ## Test that we report warnings or dump raw data when symbols are invalid.
 
-# RUN: yaml2obj %s --docnum=1 -o %t1
-# RUN: llvm-readobj --syms %t1 2>&1 | FileCheck %s -DFILE=%t1 --check-prefix=CASE1
-
-# CASE1: warning: '[[FILE]]': the non-function C_EXT symbol at index 1 should have only 1 auxiliary entry, i.e. the CSECT auxiliary entry
-
 --- !XCOFF
 FileHeader:
   MagicNumber: 0x1DF
 Symbols:
   - Name: .sym
   - Name:               .fun
     StorageClass:       [[STORAGECLASS='C_EXT']]
     NumberOfAuxEntries: 2
 
-# RUN: yaml2obj %s --docnum=1 -DSTORAGECLASS='C_WEAKEXT' -o %t2
-# RUN: llvm-readobj --syms %t2 2>&1 | FileCheck %s -DFILE=%t2 --check-prefix=CASE2
-
-# CASE2: warning: '[[FILE]]': the non-function C_WEAKEXT symbol at index 1 should have only 1 auxiliary entry, i.e. the CSECT auxiliary entry
-
-# RUN: yaml2obj %s --docnum=1 -DSTORAGECLASS='C_HIDEXT' -o %t3
-# RUN: llvm-readobj --syms %t3 2>&1 | FileCheck %s -DFILE=%t3 --check-prefix=CASE3
-
-# CASE3: warning: '[[FILE]]': the non-function C_HIDEXT symbol at index 1 should have only 1 auxiliary entry, i.e. the CSECT auxiliary entry
-
 # RUN: yaml2obj %s --docnum=1 -DSTORAGECLASS='C_STAT' -o %t4
-# RUN: llvm-readobj --syms %t4 2>&1 | FileCheck %s -DFILE=%t4 --check-prefix=CASE4
+# RUN: llvm-readobj --syms %t4 2>&1 | FileCheck %s -DFILE=%t4 --check-prefix=CASE1
 
-# CASE4: warning: '[[FILE]]': the C_STAT symbol at index 1 should not have more than 1 auxiliary entry
+# CASE1: warning: '[[FILE]]': the C_STAT symbol at index 1 should not have more than 1 auxiliary entry
 
 # RUN: yaml2obj %s --docnum=1 -DSTORAGECLASS='C_DWARF' -o %t5
-# RUN: llvm-readobj --syms %t5 2>&1 | FileCheck %s -DFILE=%t5 --check-prefix=CASE5
+# RUN: llvm-readobj --syms %t5 2>&1 | FileCheck %s -DFILE=%t5 --check-prefix=CASE2
 
-# CASE5: warning: '[[FILE]]': the C_DWARF symbol at index 1 should not have more than 1 auxiliary entry
+# CASE2: warning: '[[FILE]]': the C_DWARF symbol at index 1 should not have more than 1 auxiliary entry
 
 # RUN: yaml2obj %s --docnum=1 -DSTORAGECLASS='C_BLOCK' -o %t6
-# RUN: llvm-readobj --syms %t6 2>&1 | FileCheck %s -DFILE=%t6 --check-prefix=CASE6
+# RUN: llvm-readobj --syms %t6 2>&1 | FileCheck %s -DFILE=%t6 --check-prefix=CASE3
 
-# CASE6: warning: '[[FILE]]': the C_BLOCK symbol at index 1 should not have more than 1 auxiliary entry
+# CASE3: warning: '[[FILE]]': the C_BLOCK symbol at index 1 should not have more than 1 auxiliary entry
 
 # RUN: yaml2obj %s --docnum=1 -DSTORAGECLASS='C_FCN' -o %t7
-# RUN: llvm-readobj --syms %t7 2>&1 | FileCheck %s -DFILE=%t7 --check-prefix=CASE7
+# RUN: llvm-readobj --syms %t7 2>&1 | FileCheck %s -DFILE=%t7 --check-prefix=CASE4
 
-# CASE7: warning: '[[FILE]]': the C_FCN symbol at index 1 should not have more than 1 auxiliary entry
+# CASE4: warning: '[[FILE]]': the C_FCN symbol at index 1 should not have more than 1 auxiliary entry
 
 ## This case tests the raw data output ability when a file auxiliary entry does
 ## not have the matching auxiliary type.
 # RUN: yaml2obj %s --docnum=2 -o %t8
-# RUN: llvm-readobj --syms %t8 | FileCheck %s --strict-whitespace --match-full-lines --check-prefix=CASE8
+# RUN: llvm-readobj --syms %t8 | FileCheck %s --strict-whitespace --match-full-lines --check-prefix=CASE5
 
-#      CASE8:Symbols [
-# CASE8-NEXT:  Symbol {
-# CASE8-NEXT:    Index: 0
-# CASE8-NEXT:    Name: .fun
-# CASE8-NEXT:    Value (SymbolTableIndex): 0x0
-# CASE8-NEXT:    Section: N_UNDEF
-# CASE8-NEXT:    Source Language ID: TB_C (0x0)
-# CASE8-NEXT:    CPU Version ID: 0x0
-# CASE8-NEXT:    StorageClass: C_FILE (0x67)
-# CASE8-NEXT:    NumberOfAuxEntries: 1
-# CASE8-NEXT:    !Unexpected raw auxiliary entry data:
-# CASE8-NEXT:    00000000 00000001 00020300 00000000 00fb
-# CASE8-NEXT:  }
-# CASE8-NEXT:]
+#      CASE5:Symbols [
+# CASE5-NEXT:  Symbol {
+# CASE5-NEXT:    Index: 0
+# CASE5-NEXT:    Name: .fun
+# CASE5-NEXT:    Value (SymbolTableIndex): 0x0
+# CASE5-NEXT:    Section: N_UNDEF
+# CASE5-NEXT:    Source Language ID: TB_C (0x0)
+# CASE5-NEXT:    CPU Version ID: 0x0
+# CASE5-NEXT:    StorageClass: C_FILE (0x67)
+# CASE5-NEXT:    NumberOfAuxEntries: 1
+# CASE5-NEXT:    !Unexpected raw auxiliary entry data:
+# CASE5-NEXT:    00000000 00000001 00020300 00000000 00fb
+# CASE5-NEXT:  }
+# CASE5-NEXT:]
 
 --- !XCOFF
 FileHeader:
   MagicNumber: 0x1F7
 Symbols:
   - Name:         .fun
     StorageClass: C_FILE
     AuxEntries:
       - Type:                   AUX_CSECT
         ParameterHashIndex:     1
         TypeChkSectNum:         2
         SymbolAlignmentAndType: 3
         StorageMappingClass:    XMC_PR

llvm/test/tools/llvm-readobj/XCOFF/symbols.test

## The C_FCN symbol with a Block auxiliary entry.
   - Name:               .fun6
     StorageClass:       C_FCN
     NumberOfAuxEntries: 1
     AuxEntries:
       - Type:      AUX_SYM
         LineNumHi: 2
         LineNumLo: 3
 
+##The C_WEAKEXT symbol with two Function auxiliary entries and a CSECT auxiliary entry.
+  - Name:               .fun7
+    Value:              0x0
+    Section:            .text 
+    Type:               0x20
+    StorageClass:       C_WEAKEXT
+    NumberOfAuxEntries: 3
+    AuxEntries:
+      - Type:                 AUX_FCN
+        OffsetToExceptionTbl: 2
+        SizeOfFunction:       3
+        SymIdxOfNextBeyond:   4
+        PtrToLineNum:         5
+      - Type:                 AUX_FCN
+        OffsetToExceptionTbl: 8
+        SizeOfFunction:       5
+        SymIdxOfNextBeyond:   8
+        PtrToLineNum:         5
+      - Type:                   AUX_CSECT
+        ParameterHashIndex:     11
+        TypeChkSectNum:         22
+        SymbolAlignmentAndType: 33
+        StorageMappingClass:    XMC_PR
+        SectionOrLength:        256
+        StabInfoIndex:          44
+        StabSectNum:            55
+
 # SYMBOL32:      Symbols [
 # SYMBOL32-NEXT:   Symbol {
 # SYMBOL32-NEXT:     Index: 0
 # SYMBOL32-NEXT:     Name: .file
 # SYMBOL32-NEXT:     Value (SymbolTableIndex): 0x0
 # SYMBOL32-NEXT:     Section: N_DEBUG
 # SYMBOL32-NEXT:     Source Language ID: TB_C (0x0)
 # SYMBOL32-NEXT:     CPU Version ID: 0x0

 # SYMBOL32-NEXT:     StorageClass: C_FCN (0x65)
 # SYMBOL32-NEXT:     NumberOfAuxEntries: 1
 # SYMBOL32-NEXT:     Block Auxiliary Entry {
 # SYMBOL32-NEXT:       Index: 20
 # SYMBOL32-NEXT:       LineNumber (High 2 Bytes): 0x2
 # SYMBOL32-NEXT:       LineNumber (Low 2 Bytes): 0x3
 # SYMBOL32-NEXT:     }
 # SYMBOL32-NEXT:   }
+# SYMBOL32-NEXT:   Symbol {
+# SYMBOL32-NEXT:     Index: 21
+# SYMBOL32-NEXT:     Name: .fun7
+# SYMBOL32-NEXT:     Value (RelocatableAddress): 0x0
+# SYMBOL32-NEXT:     Section: .text
+# SYMBOL32-NEXT:     Type: 0x20
+# SYMBOL32-NEXT:     StorageClass: C_WEAKEXT (0x6F)
+# SYMBOL32-NEXT:     NumberOfAuxEntries: 3
+# SYMBOL32-NEXT:     Function Auxiliary Entry {
+# SYMBOL32-NEXT:       Index: 22
+# SYMBOL32-NEXT:       OffsetToExceptionTable: 0x2
+# SYMBOL32-NEXT:       SizeOfFunction: 0x3
+# SYMBOL32-NEXT:       PointerToLineNum: 0x5
+# SYMBOL32-NEXT:       SymbolIndexOfNextBeyond: 4
+# SYMBOL32-NEXT:     }
+# SYMBOL32-NEXT:     Function Auxiliary Entry {
+# SYMBOL32-NEXT:       Index: 23
+# SYMBOL32-NEXT:       OffsetToExceptionTable: 0x8
+# SYMBOL32-NEXT:       SizeOfFunction: 0x5
+# SYMBOL32-NEXT:       PointerToLineNum: 0x5
+# SYMBOL32-NEXT:       SymbolIndexOfNextBeyond: 8
+# SYMBOL32-NEXT:     }
+# SYMBOL32-NEXT:     CSECT Auxiliary Entry {
+# SYMBOL32-NEXT:       Index: 24
+# SYMBOL32-NEXT:       SectionLen: 256
+# SYMBOL32-NEXT:       ParameterHashIndex: 0xB
+# SYMBOL32-NEXT:       TypeChkSectNum: 0x16
+# SYMBOL32-NEXT:       SymbolAlignmentLog2: 4
+# SYMBOL32-NEXT:       SymbolType: XTY_SD (0x1)
+# SYMBOL32-NEXT:       StorageMappingClass: XMC_PR (0x0)
+# SYMBOL32-NEXT:       StabInfoIndex: 0x2C
+# SYMBOL32-NEXT:       StabSectNum: 0x37
+# SYMBOL32-NEXT:     }
+# SYMBOL32-NEXT:   }
 # SYMBOL32-NEXT: ]

llvm/tools/llvm-readobj/XCOFFDumper.cpp

for (int I = 1; I <= NumberOfAuxEntries; I++) {
       const XCOFFFileAuxEnt *FileAuxEntPtr =
           getAuxEntPtr<XCOFFFileAuxEnt>(AuxAddress);
       printFileAuxEnt(FileAuxEntPtr);
     }
     break;
   case XCOFF::C_EXT:
   case XCOFF::C_WEAKEXT:
   case XCOFF::C_HIDEXT: {
-    if (!SymbolEntRef.isFunction() && NumberOfAuxEntries > 1)
-      reportUniqueWarning("the non-function " +
-                          enumToString(static_cast<uint8_t>(SymbolClass),
-                                       makeArrayRef(SymStorageClass)) +
-                          " symbol at index " + Twine(SymbolIdx) +
-                          " should have only 1 auxiliary entry, i.e. the CSECT "
-                          "auxiliary entry");
-
     // For 32-bit objects, print the function auxiliary symbol table entry. The
     // last one must be a CSECT auxiliary entry.
     // For 64-bit objects, both a function auxiliary entry and an exception
     // auxiliary entry may appear, print them in the loop and skip printing the
     // CSECT auxiliary entry, which will be printed outside the loop.
     for (int I = 1; I <= NumberOfAuxEntries; I++) {
-      if ((I == NumberOfAuxEntries && !Obj.is64Bit()) ||
-          !SymbolEntRef.isFunction())
+      if (I == NumberOfAuxEntries && !Obj.is64Bit())
         break;
 
       uintptr_t AuxAddress = XCOFFObjectFile::getAdvancedSymbolEntryAddress(
           SymbolEntRef.getEntryAddress(), I);
 
       if (Obj.is64Bit()) {
         XCOFF::SymbolAuxType Type = *Obj.getSymbolAuxType(AuxAddress);
         if (Type == XCOFF::SymbolAuxType::AUX_CSECT)