[llvm-readelf/obj][test] - Document what we print in various places for unnamed section symbols.

[grimar]

Summary

We have an issue with ELFDumper<ELFT>::getSymbolSectionName:

1. It is used deeply for both LLVM/GNU styles and might return LLVM-style only values to describe symbols: "Undefined", "Processor Specific", "Absolute", etc.

2. getSymbolSectionName is used by getFullSymbolName and these special values might appear in instead of symbol names in many places. This occurs for unnamed section symbols.

It was not noticed because for most cases I've found it is unexpected to have an
unnamed section symbol. This patch documents the existent behavior, adds tests and FIXMEs.

These FIXMEs are fixed in D87764.

[jhenderson]

llvm/test/tools/llvm-readobj/ELF/dyn-symbols.test

326

I think it would be slightly more obvious to use Name: [[NAME=bar]] and then just set -DNAME='', if that's possible?

348

If you add the check for the first line of GNU output that says how many symbols there are, I think you can probably get away with simplifying the above patterns to single line checks for the names (i.e. without the need for -SAME).

llvm/test/tools/llvm-readobj/ELF/hash-symbols.test

87

Same comment as above.

llvm/test/tools/llvm-readobj/ELF/mips-got.test

707–708

Is there any particular reason you've chosen to have both absolute and common here, but none of the other varieties?

llvm/test/tools/llvm-readobj/ELF/symbol-shndx.test

63

Same comment as above re. naming.

[grimar]

• Addressed review comments.

llvm/test/tools/llvm-readobj/ELF/dyn-symbols.test

326

Interesting idea. Done.

llvm/test/tools/llvm-readobj/ELF/mips-got.test

707–708

No. I've just showed 2 arbitrary symbols revealing the issue and an arbitrary normal symbol in between of them.
I do not feel it worth to check all possible values here? They are tested in symbol-shndx.test, we just have the same issue in all other places because of using the same API for getting a symbol name.

[jhenderson]

LGTM.

llvm/test/tools/llvm-readobj/ELF/dyn-symbols.test

Entries:
       - Version:    1
         Flags:      0
         VersionNdx: 3
         Hash:       0
         Names:
           - NonDefault
 DynamicSymbols:
   - Name: foo
-  - Name: bar
-  - Name: zed
+  - Name:  [[NAME=bar]]
+    Type:  [[TYPE=STT_NOTYPE]]
+    Index: [[INDEX=<none>]]
+  - Name: [[NAME=zed]]
+    Type: [[TYPE=STT_NOTYPE]]
+
+## Check the behavior for unnamed versioned section symbols.
+## TODO: we should print proper symbol names instead of descriptions.
+# RUN: yaml2obj %s -DTYPE=STT_SECTION -DNAME="''" -DINDEX=SHN_ABS --docnum=6 -o %t6.sec.sym
+# RUN: llvm-readobj -V --dyn-symbols %t6.sec.sym | FileCheck %s --check-prefix=VERSIONED-SEC-SYM-LLVM
+# RUN: llvm-readelf -V --dyn-symbols %t6.sec.sym | FileCheck %s --check-prefix=VERSIONED-SEC-SYM-GNU
+
+# VERSIONED-SEC-SYM-LLVM: DynamicSymbols [
+# VERSIONED-SEC-SYM-LLVM:  Name: foo (12)
+# VERSIONED-SEC-SYM-LLVM:  Name: Absolute (0)
+# VERSIONED-SEC-SYM-LLVM:  Name: Undefined (0)
+# VERSIONED-SEC-SYM-LLVM: VersionSymbols [
+# VERSIONED-SEC-SYM-LLVM:  Name: foo
+# VERSIONED-SEC-SYM-LLVM:  Name: Absolute
+# VERSIONED-SEC-SYM-LLVM:  Name: Undefined
+
+# VERSIONED-SEC-SYM-GNU: Symbol table '.dynsym' contains 4 entries:
+# VERSIONED-SEC-SYM-GNU:    Num: {{.*}} Ndx Name
+# VERSIONED-SEC-SYM-GNU:      1: {{.*}} UND foo
+# VERSIONED-SEC-SYM-GNU-NEXT: 2: {{.*}} ABS Absolute
+# VERSIONED-SEC-SYM-GNU-NEXT: 3: {{.*}} UND Undefined
 
 ## Case 8: Check what we print when:
 ## a) The dynamic symbol table does not exist.
 # RUN: yaml2obj %s --docnum=7 -o %t7
 # RUN: llvm-readobj --dyn-symbols %t7 | FileCheck %s --check-prefix=NO-DYNSYM-LLVM
 # RUN: llvm-readelf --dyn-symbols %t7 | count 0
 ## b) The dynamic symbol table has a size of 0.
 # RUN: yaml2obj %s --docnum=8 -o %t8

llvm/test/tools/llvm-readobj/ELF/hash-symbols.test

## This matches the file offset of the .hash section.
       - Tag:   DT_GNU_HASH
 ## PT_LOAD's p_vaddr is 0x0. DT_GNU_HASH value is 0x2c (size of .hash = 0x2c).
 ## llvm-readelf will read .gnu.hash content from PT_LOAD's p_offset + (DT_GNU_HASH value - p_vaddr).
 ## This matches the file offset of the .gnu.hash section.
         Value: 0x000000000000002C
       - Tag:   DT_NULL
         Value: 0x0000000000000000
 DynamicSymbols:
-  - Name:    ccc
+  - Name:    [[NAME=ccc]]
     Binding: STB_GLOBAL
-  - Name:    aaa
+    Type:    [[TYPE=STT_NOTYPE]]
+  - Name:    [[NAME=aaa]]
     Section: .hash
     Binding: STB_GLOBAL
     Value:   0x0000000000001000
-  - Name:    ddd
+    Type:    [[TYPE=STT_NOTYPE]]
+  - Name:    [[NAME=ddd]]
     Index:   SHN_ABS
     Binding: STB_GLOBAL
     Value:   0x0000000000000001
-  - Name:    eee
+    Type:    [[TYPE=STT_NOTYPE]]
+  - Name:    [[NAME=eee]]
     Section: .gnu.hash
     Binding: STB_GLOBAL
-  - Name:    bbb
+    Type:    [[TYPE=STT_NOTYPE]]
+  - Name:    [[NAME=bbb]]
     Section: .hash
     Binding: STB_WEAK
     Value:   0x0000000000001001
+    Type:    [[TYPE=STT_NOTYPE]]
 ProgramHeaders:
   - Type:  PT_LOAD
     Flags: [ PF_R, PF_X ]
     Sections:
       - Section: .hash
       - Section: .gnu.hash
       - Section: .dynamic
 
+## Check what we print for unnamed section symbols.
+## TODO: we should print proper symbol names instead of descriptions.
+# RUN: yaml2obj --docnum=1 -DBITS=64 -DTYPE=STT_SECTION -DNAME="''" %s -o %t1-sec-syms.so
+# RUN: llvm-readelf --hash-symbols %t1-sec-syms.so | FileCheck %s --check-prefix=UNNAMED-SEC-SYMS
+
+# UNNAMED-SEC-SYMS:      Symbol table of .hash for image:
+# UNNAMED-SEC-SYMS-NEXT:  Num   {{.*}} Ndx Name
+# UNNAMED-SEC-SYMS-NEXT:    1   {{.*}} UND Undefined
+# UNNAMED-SEC-SYMS-NEXT:    5   {{.*}}   1 .hash
+# UNNAMED-SEC-SYMS-NEXT:    3   {{.*}} ABS Absolute
+# UNNAMED-SEC-SYMS-NEXT:    2   {{.*}}   1 .hash
+# UNNAMED-SEC-SYMS-NEXT:    4   {{.*}}   2 .gnu.hash
+# UNNAMED-SEC-SYMS-EMPTY:
+# UNNAMED-SEC-SYMS:      Symbol table of .gnu.hash for image:
+# UNNAMED-SEC-SYMS-NEXT:  Num {{.*}} Ndx Name
+# UNNAMED-SEC-SYMS-NEXT:    2 {{.*}}   1 .hash
+# UNNAMED-SEC-SYMS-NEXT:    3 {{.*}} ABS Absolute
+# UNNAMED-SEC-SYMS-NEXT:    4 {{.*}}   2 .gnu.hash
+# UNNAMED-SEC-SYMS-NEXT:    5 {{.*}}   1 .hash
+
 ## Check the output when only .hash section is present.
 
 # RUN: yaml2obj --docnum=2 %s -o %t2-32.so
 # RUN: llvm-readelf --hash-symbols %t2-32.so \
 # RUN:   | FileCheck %s --strict-whitespace --match-full-lines --check-prefix ONLY-HASH-32
 
 # ONLY-HASH-32: Symbol table of .hash for image:
 # ONLY-HASH-32-NEXT:  Num Buc:    Value  Size   Type   Bind Vis      Ndx Name

llvm/test/tools/llvm-readobj/ELF/mips-got.test

Entries:
       - Tag:   DT_MIPS_LOCAL_GOTNO
         Value: 1
       - Tag:   DT_MIPS_GOTSYM
         Value: 1
       - Tag:   DT_PLTGOT
         Value: 0x1122
 DynamicSymbols:
   - Name: foo
+
+## Check how we print global GOT entries when they are unnamed section symbols.
+## TODO: we should print proper symbol names instead of descriptions.
+# RUN: yaml2obj --docnum=5 %s -o %t.err8.o
+# RUN: llvm-readobj -A %t.err8.o 2>&1 | FileCheck %s -DFILE=%t.err8.o --check-prefix=SEC-SYMS-LLVM
+# RUN: llvm-readelf -A %t.err8.o 2>&1 | FileCheck %s -DFILE=%t.err8.o --check-prefix=SEC-SYMS-GNU
+
+# SEC-SYMS-LLVM:      Global entries [
+# SEC-SYMS-LLVM-NEXT:   Entry {
+# SEC-SYMS-LLVM:          Section: Absolute (0xFFF1)
+# SEC-SYMS-LLVM-NEXT:     Name: Absolute (0)
+# SEC-SYMS-LLVM-NEXT:   }
+# SEC-SYMS-LLVM-NEXT:   Entry {
+# SEC-SYMS-LLVM:          Section: .got (0x1)
+# SEC-SYMS-LLVM-NEXT:     Name: .got (0)
+# SEC-SYMS-LLVM-NEXT:   }
+# SEC-SYMS-LLVM-NEXT:   Entry {
+# SEC-SYMS-LLVM:          Section: Common (0xFFF2)
+# SEC-SYMS-LLVM-NEXT:     Name: Common (0)
+# SEC-SYMS-LLVM-NEXT:   }
+# SEC-SYMS-LLVM-NEXT: ]
+
+# SEC-SYMS-GNU:      Global entries:
+# SEC-SYMS-GNU-NEXT:  {{.*}} Ndx Name
+# SEC-SYMS-GNU-NEXT:  {{.*}} ABS Absolute
+# SEC-SYMS-GNU-NEXT:  {{.*}}   1 .got
+# SEC-SYMS-GNU-NEXT:  {{.*}} COM Common
+
+--- !ELF
+FileHeader:
+  Class:   ELFCLASS64
+  Data:    ELFDATA2LSB
+  Type:    ET_REL
+  Machine: EM_MIPS
+Sections:
+  - Name:    .got
+    Type:    SHT_PROGBITS
+    Address: 0x1122
+    Size:    48
+  - Name: .dynamic
+    Type: SHT_DYNAMIC
+    Entries:
+      - Tag:   DT_MIPS_LOCAL_GOTNO
+        Value: 1
+      - Tag:   DT_MIPS_GOTSYM
+        Value: 1
+      - Tag:   DT_PLTGOT
+        Value: 0x1122
+DynamicSymbols:
+  - Type:    STT_SECTION
+    Index:   SHN_ABS
+  - Type:    STT_SECTION
+    Section: .got
+  - Type:    STT_SECTION
+    Index:   SHN_COMMON

llvm/test/tools/llvm-readobj/ELF/mips-plt.test

 # RUN: not llvm-readobj -A %t.err6.o 2>&1 | FileCheck %s -DFILE=%t.err6.o -check-prefix ERR6
 
 # ERR6: error: '[[FILE]]': unable to get a symbol table linked to the SHT_PROGBITS section with index 2: invalid section index: 2863311530
 
 # RUN: yaml2obj --docnum=2 -DVAL2=0x100 %s -o %t.err7.o
 # RUN: not llvm-readobj -A %t.err7.o 2>&1 | FileCheck %s -DFILE=%t.err7.o -check-prefix ERR7
 
 # ERR7: error: '[[FILE]]': unable to get a string table for the SHT_DYNAMIC section with index 1: invalid sh_type for symbol table, expected SHT_SYMTAB or SHT_DYNSYM
+
+## Check how we print PLT entries when they are unnamed section symbols.
+## TODO: we should print proper symbol names instead of descriptions.
+# RUN: yaml2obj --docnum=3 %s -o %t.3
+# RUN: llvm-readobj -A %t.3 2>&1 | FileCheck %s -DFILE=%t.err8.o --check-prefix=SEC-SYMS-LLVM
+# RUN: llvm-readelf -A %t.3 2>&1 | FileCheck %s -DFILE=%t.err8.o --check-prefix=SEC-SYMS-GNU
+
+# SEC-SYMS-LLVM:      PLT GOT {
+# SEC-SYMS-LLVM:        Entries [
+# SEC-SYMS-LLVM:          Entry {
+# SEC-SYMS-LLVM:            Section: Absolute (0xFFF1)
+# SEC-SYMS-LLVM-NEXT:       Name: Absolute (0)
+# SEC-SYMS-LLVM-NEXT:     }
+# SEC-SYMS-LLVM-NEXT:     Entry {
+# SEC-SYMS-LLVM:            Section: .got.plt (0x2)
+# SEC-SYMS-LLVM-NEXT:       Name: .got.plt (0)
+# SEC-SYMS-LLVM-NEXT:     }
+# SEC-SYMS-LLVM-NEXT:     Entry {
+# SEC-SYMS-LLVM:            Section: Common (0xFFF2)
+# SEC-SYMS-LLVM-NEXT:       Name: Common (0)
+# SEC-SYMS-LLVM-NEXT:     }
+# SEC-SYMS-LLVM-NEXT:   ]
+# SEC-SYMS-LLVM-NEXT: }
+
+# SEC-SYMS-GNU:      PLT GOT:
+# SEC-SYMS-GNU:       Entries:
+# SEC-SYMS-GNU-NEXT:   Address          {{.*}} Ndx Name
+# SEC-SYMS-GNU-NEXT:   0000000000002010 {{.*}} ABS Absolute
+# SEC-SYMS-GNU-NEXT:   0000000000002018 {{.*}}   2 .got.plt
+# SEC-SYMS-GNU-NEXT:   0000000000002020 {{.*}} COM Common
+
+--- !ELF
+FileHeader:
+  Class:   ELFCLASS64
+  Data:    ELFDATA2LSB
+  Type:    ET_REL
+  Machine: EM_MIPS
+Sections:
+  - Name:    .rel.plt
+    Type:    SHT_REL
+    Flags:   [ SHF_ALLOC ]
+    Address: 0x1000
+    Link:    .dynsym
+    Relocations:
+      - Offset: 0x1
+        Symbol: 1
+        Type:   R_MIPS_JUMP_SLOT
+      - Offset: 0x2
+        Symbol: 2
+        Type:   R_MIPS_JUMP_SLOT
+      - Offset: 0x2
+        Symbol: 3
+        Type:   R_MIPS_JUMP_SLOT
+  - Name:    .got.plt
+    Type:    SHT_PROGBITS
+    Flags:   [ SHF_WRITE, SHF_ALLOC ]
+    Address: 0x2000
+    Size:    40 ## (dynamic symbols number + 2) * 8
+  - Name: .dynamic
+    Type: SHT_DYNAMIC
+    Entries:
+      - Tag:   DT_JMPREL
+        Value: 0x1000
+      - Tag:   DT_MIPS_PLTGOT
+        Value: 0x2000
+DynamicSymbols:
+  - Type:    STT_SECTION
+    Index:   SHN_ABS
+  - Type:    STT_SECTION
+    Section: .got.plt
+  - Type:    STT_SECTION
+    Index:   SHN_COMMON

llvm/test/tools/llvm-readobj/ELF/symbol-shndx.test

 Sections:
   - Name: .text
     Type: SHT_PROGBITS
   - Name: .symtab_shndx
     Type: SHT_SYMTAB_SHNDX
     Link: .symtab
     Entries: [ 0, 0, 0, 0, 0, 0, 0, 0, 1 ]
 Symbols:
-  - Name:    undef
+  - Name:    [[NAME=undef]]
     Binding: STB_GLOBAL
-  - Name:    normal
+    Type:    [[TYPE=STT_NOTYPE]]
+  - Name:    [[NAME=normal]]
     Section: .text
     Binding: STB_GLOBAL
-  - Name:    common
+    Type:    [[TYPE=STT_NOTYPE]]
+  - Name:    [[NAME=common]]
     Index:   SHN_COMMON
     Binding: STB_GLOBAL
-  - Name:    absolute
+    Type:    [[TYPE=STT_NOTYPE]]
+  - Name:    [[NAME=absolute]]
     Index:   SHN_ABS
     Binding: STB_GLOBAL
-  - Name:    proc
+    Type:    [[TYPE=STT_NOTYPE]]
+  - Name:    [[NAME=proc]]
     Index:   0xff01
     Binding: STB_GLOBAL
-  - Name:    os
+    Type:    [[TYPE=STT_NOTYPE]]
+  - Name:    [[NAME=os]]
     Index:   0xff21
     Binding: STB_GLOBAL
-  - Name:    reserved
+    Type:    [[TYPE=STT_NOTYPE]]
+  - Name:    [[NAME=reserved]]
     Index:   0xfffe
     Binding: STB_GLOBAL
-  - Name:    xindex
+    Type:    [[TYPE=STT_NOTYPE]]
+  - Name:    [[NAME=xindex]]
     Index:   SHN_XINDEX
     Binding: STB_GLOBAL
+    Type:    [[TYPE=STT_NOTYPE]]
+
+## Check the behavior for section symbols.
+# RUN: yaml2obj --docnum=1 -DTYPE=STT_SECTION %s -o %t1-sec
+# RUN: llvm-readobj --symbols %t1-sec | FileCheck %s --check-prefix=LLVM1
+# RUN: llvm-readelf --symbols %t1-sec | FileCheck %s --check-prefix=GNU1
+
+## Check the behavior for unnamed section symbols.
+## TODO: we should print proper symbol names instead of descriptions.
+# RUN: yaml2obj --docnum=1 -DTYPE=STT_SECTION -DNAME="''" %s -o %t1-sec-unnamed
+# RUN: llvm-readobj --symbols %t1-sec-unnamed | FileCheck %s --check-prefix=LLVM1-SEC-SYMS
+# RUN: llvm-readelf --symbols %t1-sec-unnamed | FileCheck %s --check-prefix=GNU1-SEC-SYMS
+
+# LLVM1-SEC-SYMS:      Symbols [
+# LLVM1-SEC-SYMS-NEXT:   Symbol {
+# LLVM1-SEC-SYMS-NEXT:     Name:  (0)
+# LLVM1-SEC-SYMS:          Section: Undefined (0x0)
+# LLVM1-SEC-SYMS:        Symbol {
+# LLVM1-SEC-SYMS-NEXT:     Name: Undefined (0)
+# LLVM1-SEC-SYMS:          Section: Undefined (0x0)
+# LLVM1-SEC-SYMS:        Symbol {
+# LLVM1-SEC-SYMS-NEXT:     Name: .text (0)
+# LLVM1-SEC-SYMS:          Section: .text (0x1)
+# LLVM1-SEC-SYMS:        Symbol {
+# LLVM1-SEC-SYMS-NEXT:     Name: Common (0)
+# LLVM1-SEC-SYMS:          Section: Common (0xFFF2)
+# LLVM1-SEC-SYMS:        Symbol {
+# LLVM1-SEC-SYMS-NEXT:     Name: Absolute (0)
+# LLVM1-SEC-SYMS:          Section: Absolute (0xFFF1)
+# LLVM1-SEC-SYMS:        Symbol {
+# LLVM1-SEC-SYMS-NEXT:     Name: Processor Specific (0)
+# LLVM1-SEC-SYMS:          Section: Processor Specific (0xFF01)
+# LLVM1-SEC-SYMS:        Symbol {
+# LLVM1-SEC-SYMS-NEXT:     Name: Operating System Specific (0)
+# LLVM1-SEC-SYMS:          Section: Operating System Specific (0xFF21)
+# LLVM1-SEC-SYMS:        Symbol {
+# LLVM1-SEC-SYMS-NEXT:     Name: Reserved (0)
+# LLVM1-SEC-SYMS:          Section: Reserved (0xFFFE)
+# LLVM1-SEC-SYMS:        Symbol {
+# LLVM1-SEC-SYMS-NEXT:     Name: .text (0)
+# LLVM1-SEC-SYMS:          Section: .text (0x1)
+
+# GNU1-SEC-SYMS:      Num: {{.*}} Ndx Name
+# GNU1-SEC-SYMS-NEXT:   0: {{.*}} UND
+# GNU1-SEC-SYMS-NEXT:   1: {{.*}} UND Undefined
+# GNU1-SEC-SYMS-NEXT:   2: {{.*}}   1 .text
+# GNU1-SEC-SYMS-NEXT:   3: {{.*}} COM Common
+# GNU1-SEC-SYMS-NEXT:   4: {{.*}} ABS Absolute
+# GNU1-SEC-SYMS-NEXT:   5: {{.*}} PRC[0xff01] Processor Specific
+# GNU1-SEC-SYMS-NEXT:   6: {{.*}} OS[0xff21] Operating System Specific
+# GNU1-SEC-SYMS-NEXT:   7: {{.*}} RSV[0xfffe] Reserved
+# GNU1-SEC-SYMS-NEXT:   8: {{.*}}   1 .text
 
 ## In this case, the index does not correspond to a real section. Check that GNU
 ## style just prints the section index as normal and LLVM style prints a warning
 ## (but only once for each warning).
 # RUN: yaml2obj --docnum=2 %s -o %t2
 # RUN: llvm-readobj --symbols %t2 2> %t2.llvm.err | FileCheck %s --check-prefix=LLVM2
 # RUN: FileCheck %s --input-file=%t2.llvm.err --check-prefix=BAD-SHNDX --implicit-check-not=warning
 # RUN: llvm-readelf --symbols %t2 2> %t2.gnu.err | FileCheck %s --check-prefix=GNU2