[llvm-objcopy] Add additional testing for various cases

[jhenderson]

Summary

This patch adds a number of tests to test various cases not covered by existing tests. All of them work correctly, with no need to change llvm-objcopy itself, although some do indicate possible areas for improvement.

[jhenderson]

Add one more test case for overlapping sections inside segments.

[MaskRay]

test/tools/llvm-objcopy/ELF/common-suffix-string-tables.test

21

Indent the output as other tests do?

test/tools/llvm-objcopy/ELF/invalid-e_phoff.test

15

'r+b' if you want to make it work on Python 3 as well though.

[jhenderson]

Address review comments:

• Use extra indentation
• Use r+b for python binary editing

[MaskRay]

test/tools/llvm-objcopy/ELF/common-suffix-string-tables.test

3

It is tail merge, not substring merge.

69

Add another section to make tail merging clear? e.g. .a.b.c .b.c .c

test/tools/llvm-objcopy/ELF/strip-sections.test

39

Nit: # -> # .

[jhenderson]

Address review comments:

• Add extra spacing.
• Rename tail merging test/update comments.
• Add extra symbol and section to tail merging test.
• Remove unnecessary switches from tail merging test.

[thakis]

llvm/trunk/test/tools/llvm-objcopy/ELF/shared-strtab-shstrtab.s

63 ↗

(On Diff #199006)

This makes the test fail if "foo" is a substring of the path of your build dir. Could you make this more robust against accidentally matching the directory name?

[jhenderson]

llvm/trunk/test/tools/llvm-objcopy/ELF/shared-strtab-shstrtab.s

63 ↗

(On Diff #199006)

See D77546.

test/tools/llvm-objcopy/ELF/common-suffix-string-tables.test

+## This test shows that when symbol/section names are optimized in the string
+## table to share the same entry, because one is a substring of the other,
+## llvm-objcopy operations function correctly without corrupting the names of
+## unchanged sections/symbols.
+
+# RUN: yaml2obj %s -o %t.o
+## Sanity check that the strings have been pooled.
+# RUN: llvm-readobj %t.o --string-dump .strtab --string-dump .shstrtab \
+# RUN:   | FileCheck %s --check-prefix=VALIDATE --implicit-check-not=bar --implicit-check-not=.blag
+
+# VALIDATE: String dump of section '.strtab':
+# VALIDATE: foobar
+# VALIDATE: String dump of section '.shstrtab':
+# VALIDATE: .blah.blag
+
+## Case 1: Basic copy.
+# RUN: llvm-objcopy %t.o %t.copy
+# RUN: llvm-readobj --section-headers --symbols %t.copy | FileCheck %s --check-prefix=COPY
+
+# COPY: Sections [
+# COPY: Name: .blah.blag (
+# COPY: Name: .blag (
+# COPY: Symbols [
+# COPY: Name: bar (
+# COPY: Name: foobar (
+
+## Case 2: Rename section.
+# RUN: llvm-objcopy %t.o %t.rename-section --rename-section .blag=.blab
+# RUN: llvm-readobj --section-headers --symbols %t.rename-section \
+# RUN:   | FileCheck %s --check-prefix=RENAME-SECTION
+
+# RENAME-SECTION: Sections [
+# RENAME-SECTION: Name: .blah.blag (
+# RENAME-SECTION: Name: .blab (
+
+## Case 3: Rename symbol.
+# RUN: llvm-objcopy %t.o %t.rename-symbol --redefine-sym bar=rab
+# RUN: llvm-readobj --section-headers --symbols %t.rename-symbol \
+# RUN:   | FileCheck %s --check-prefix=RENAME-SYMBOL
+
+# RENAME-SYMBOL: Symbols [
+# RENAME-SYMBOL: Name: rab (
+# RENAME-SYMBOL: Name: foobar (
+
+## Case 4: Remove section.
+# RUN: llvm-objcopy %t.o %t.remove-section -R .blag
+# RUN: llvm-readobj --section-headers --symbols %t.remove-section \
+# RUN:   | FileCheck %s --check-prefix=REMOVE-SECTION --implicit-check-not .blag
+
+# REMOVE-SECTION: Sections [
+# REMOVE-SECTION: Name: .blah.blag (
+
+## Case 5: Remove symbol.
+# RUN: llvm-objcopy %t.o %t.remove-symbol -N bar
+# RUN: llvm-readobj --section-headers --symbols %t.remove-symbol \
+# RUN:   | FileCheck %s --check-prefix=REMOVE-SYMBOL --implicit-check-not bar
+
+# REMOVE-SYMBOL: Symbols [
+# REMOVE-SYMBOL: Name: foobar (
+
+--- !ELF
+FileHeader:
+  Class:   ELFCLASS64
+  Data:    ELFDATA2LSB
+  Type:    ET_EXEC
+  Machine: EM_X86_64
+Sections:
+  - Name: .blah.blag
+    Type: SHT_PROGBITS
+  - Name: .blag
+    Type: SHT_PROGBITS
+Symbols:
+  - Name: bar
+  - Name: foobar

test/tools/llvm-objcopy/ELF/invalid-e_phoff.test

+## If the e_phoff field is set to a value either past the end of the file, or
+## such that e_phoff + e_phnum * sizeof(Elf_Phdr) is past the end of the file,
+## we should emit an error. This test checks that the emitted error is sensible.
+
+# RUN: yaml2obj %s -o %t.o
+## Remove the default section headers so that the section header table is not invalid.
+# RUN: llvm-objcopy --strip-sections %t.o
+# RUN: cp %t.o %t2.o
+
+## Sanity check that the phdr table is at offset 64:
+# RUN: llvm-readobj --file-headers %t.o | FileCheck %s --check-prefix=VALIDATE
+# VALIDATE: ProgramHeaderOffset: 0x40{{$}}
+
+## Truncate the file to end before the program header table ends.
+# RUN: %python -c "with open('%/t.o', 'r+') as input: input.truncate(65)"
+# RUN: not llvm-objcopy %t.o 2>&1 | FileCheck %s
+
+## Set the e_phoff field to a value much larger than the object file size.
+# RUN: %python -c "with open('%/t2.o', 'r+') as input: import struct; bytes = struct.pack('<Q', 0x40000000); input.seek(32); input.write(bytes)"
+# RUN: not llvm-objcopy %t2.o 2>&1 | FileCheck %s
+
+--- !ELF
+FileHeader:
+  Class:   ELFCLASS64
+  Data:    ELFDATA2LSB
+  Type:    ET_EXEC
+  Machine: EM_X86_64
+ProgramHeaders:
+  - Type: PT_LOAD
+
+# CHECK: error: program headers longer than binary

test/tools/llvm-objcopy/ELF/invalid-e_shoff.test

+## If the e_shoff field is set to a value either past the end of the file, or
+## such that e_shoff + e_shnum * sizeof(Elf_Shdr) is past the end of the file,
+## we should emit an error. This test checks that the emitted error is sensible.
+
+# RUN: yaml2obj %s -o %t.o
+# RUN: cp %t.o %t2.o
+
+## Sanity check that the section header table is at offset 64:
+# RUN: llvm-readobj --file-headers %t.o | FileCheck %s --check-prefix=VALIDATE
+# VALIDATE: SectionHeaderOffset: 0x40{{$}}
+
+## Truncate the file to end before the section header table ends.
+# RUN: %python -c "with open('%/t.o', 'r+') as input: input.truncate(65)"
+# RUN: not llvm-objcopy %t.o 2>&1 | FileCheck %s -DINPUT=%t.o
+
+## Set the e_shoff field to a value much larger than the object file size.
+# RUN: %python -c "with open('%/t2.o', 'r+') as input: import struct; bytes = struct.pack('<Q', 0x40000000); input.seek(40); input.write(bytes)"
+# RUN: not llvm-objcopy %t2.o 2>&1 | FileCheck %s -DINPUT=%t2.o
+
+--- !ELF
+FileHeader:
+  Class:   ELFCLASS64
+  Data:    ELFDATA2LSB
+  Type:    ET_EXEC
+  Machine: EM_X86_64
+Sections:
+  - Name: .foo
+    Type: SHT_PROGBITS
+
+# CHECK: error: '[[INPUT]]': section header table goes past the end of the file

test/tools/llvm-objcopy/ELF/linked-section.test

+## This test checks that if a section has a sh_link field, and one or more
+## sections are removed such that the target section index changes, then
+## llvm-objcopy correctly updates the sh_link field.
+
+# RUN: yaml2obj %s -o %t.o
+# RUN: llvm-objcopy --remove-section .foo %t.o %t2.o
+# RUN: llvm-readobj --section-headers %t2.o | FileCheck %s
+
+--- !ELF
+FileHeader:
+  Class:   ELFCLASS64
+  Data:    ELFDATA2LSB
+  Type:    ET_EXEC
+  Machine: EM_X86_64
+Sections:
+  - Name: .foo
+    Type: SHT_PROGBITS
+  - Name: .bar
+    Type: SHT_PROGBITS
+  - Name: .baz
+    Type: SHT_PROGBITS
+    Link: .bar
+
+# CHECK:      Index: 1
+# CHECK-NEXT: Name: .bar
+# CHECK:      Name: .baz
+## Find the next "Link" line, then check the value is exactly the index of .bar.
+# CHECK:      Link
+# CHECK-SAME:     : 1{{$}}

test/tools/llvm-objcopy/ELF/overlapping-sections-in-segments.test

+## This test shows that llvm-objcopy does not baulk at overlapping sections
+## where such sections are within segments.
+## These don't really make sense, but the tool should still handle invalid
+## inputs somehow.
+# RUN: yaml2obj %s -o %t.o
+
+## First, check that the section header table appears immediately after the program
+## header table.
+# RUN: llvm-readobj --file-headers %t.o | FileCheck %s --check-prefix=SHDRS-OFFSET
+# SHDRS-OFFSET: SectionHeaderOffset: 0x78{{$}}
+
+## Binary edit the section header sh_offset field of the second section to
+## overlap the first one.
+# RUN: %python -c "with open('%/t.o', 'r+') as input: import struct; bytes = struct.pack('<Q', 0x1001); input.seek(272); input.write(bytes)"
+
+## Sanity check that the binary editing modified the correct field.
+# RUN: llvm-readobj --section-headers %t.o | FileCheck %s
+
+## Check that the contents are as expected before the copy.
+# RUN: llvm-readobj -x .first -x .second %t.o | FileCheck %s --check-prefix=CONTENTS
+
+## Now check that the section contents are still correct after the copy and show that
+## llvm-objcopy hasn't "unoverlapped" these sections because they are in segments.
+# RUN: llvm-objcopy %t.o %t2.o
+# RUN: llvm-readobj --section-headers %t2.o | FileCheck %s
+# RUN: llvm-readobj -x .first -x .second %t2.o | FileCheck %s --check-prefix=CONTENTS
+
+# CHECK: Name: .first
+# CHECK: Offset: 0x1000
+# CHECK: Name: .second
+# CHECK: Offset: 0x1001
+
+# CONTENTS:      Hex dump of section '.first':
+# CONTENTS-NEXT: 0x00000000 01234567
+# CONTENTS-NEXT: Hex dump of section '.second':
+# CONTENTS-NEXT: 0x00000000 23456789
+
+--- !ELF
+FileHeader:
+  Class:   ELFCLASS64
+  Data:    ELFDATA2LSB
+  Type:    ET_EXEC
+  Machine: EM_X86_64
+Sections:
+  - Name: .first
+    Type: SHT_PROGBITS
+    Content: '01234567'
+    AddressAlign: 0x1000
+  - Name: .second
+    Type: SHT_PROGBITS
+    Content: '89abcdef'
+ProgramHeaders:
+  - Type:     PT_LOAD
+    FileSize: 5
+    Offset:   0x1000
+    Align:    0x1000

test/tools/llvm-objcopy/ELF/overlapping-sections.test

+## This test shows that llvm-objcopy does not baulk at overlapping sections.
+## These don't really make sense, but the tool should still handle invalid
+## inputs somehow.
+# RUN: yaml2obj %s -o %t.o
+
+## First, check that the section header table appears immediately after the ELF
+## header.
+# RUN: llvm-readobj --file-headers %t.o | FileCheck %s --check-prefix=SHDRS-OFFSET
+# SHDRS-OFFSET: SectionHeaderOffset: 0x40{{$}}
+
+## Binary edit the section header sh_offset field of the second section to
+## overlap the first one.
+# RUN: %python -c "with open('%/t.o', 'r+') as input: import struct; bytes = struct.pack('<Q', 0x1001); input.seek(216); input.write(bytes)"
+
+## Sanity check that the binary editing modified the correct field.
+# RUN: llvm-readobj --section-headers %t.o | FileCheck %s --check-prefix=VALIDATE
+
+# VALIDATE: Name: .first
+# VALIDATE: Offset: 0x1000
+# VALIDATE: Name: .second
+# VALIDATE: Offset: 0x1001
+
+## Check that the contents are as expected before the copy.
+# RUN: llvm-readobj -x .first -x .second %t.o | FileCheck %s --check-prefix=CONTENTS
+
+## Now check that the section contents are still correct after the copy.
+## Also characterize the behavior of llvm-objcopy in that it "unoverlaps" the
+## two sections.
+# RUN: llvm-objcopy %t.o %t2.o
+# RUN: llvm-readobj --section-headers %t2.o | FileCheck %s
+# RUN: llvm-readobj -x .first -x .second %t2.o | FileCheck %s --check-prefix=CONTENTS
+
+# CHECK: Name: .first
+# CHECK: Offset: 0x1000
+# CHECK: Name: .second
+# CHECK: Offset: 0x1004
+
+# CONTENTS:      Hex dump of section '.first':
+# CONTENTS-NEXT: 0x00000000 01234567
+# CONTENTS-NEXT: Hex dump of section '.second':
+# CONTENTS-NEXT: 0x00000000 23456789
+
+--- !ELF
+FileHeader:
+  Class:   ELFCLASS64
+  Data:    ELFDATA2LSB
+  Type:    ET_EXEC
+  Machine: EM_X86_64
+Sections:
+  - Name: .first
+    Type: SHT_PROGBITS
+    Content: '01234567'
+    AddressAlign: 0x1000
+  - Name: .second
+    Type: SHT_PROGBITS
+    Content: '89abcdef'

test/tools/llvm-objcopy/ELF/rename-section-duplicate-names.test

+## When multiple sections have the same name, and --rename-section is specified,
+## GNU objcopy renames all these sections. This test shows that llvm-objcopy
+## does the same.
+
+## Note: we have to do this test in two stages because yaml2obj cannot handle
+## multiple sections with the same name. This has the benefit of showing that
+## we can rename to an already existing name.
+# RUN: yaml2obj %s -o %t.o
+
+## First make two sections with the same name...
+# RUN: llvm-objcopy --rename-section=.foo=.bar %t.o %t2.o
+## ... then rename them both.
+# RUN: llvm-objcopy --rename-section=.bar=.baz %t2.o %t3.o
+# RUN: llvm-readobj --sections %t3.o | FileCheck %s
+
+--- !ELF
+FileHeader:
+  Class:   ELFCLASS64
+  Data:    ELFDATA2LSB
+  Type:    ET_REL
+  Machine: EM_X86_64
+Sections:
+  - Name:  .foo
+    Type:  SHT_PROGBITS
+    Flags: [ SHF_ALLOC ]
+  - Name:  .bar
+    Type:  SHT_PROGBITS
+
+# CHECK: Name: .baz
+# CHECK: Name: .baz

test/tools/llvm-objcopy/ELF/rename-section-empty-string.test

+## This test verifies that llvm-objcopy can handle renaming a section to and
+## from an empty string.
+# RUN: yaml2obj %s -o %t.o
+
+# RUN: llvm-objcopy %t.o %t2.o --rename-section=.foo=
+# RUN: llvm-readobj --sections %t2.o | FileCheck %s --check-prefix=TO
+
+# RUN: llvm-objcopy %t2.o %t3.o --rename-section==.bar
+# RUN: llvm-readobj --sections %t3.o | FileCheck %s --check-prefix=FROM
+
+--- !ELF
+FileHeader:
+  Class:   ELFCLASS64
+  Data:    ELFDATA2LSB
+  Type:    ET_REL
+  Machine: EM_X86_64
+Sections:
+  - Name: .foo
+    Type: SHT_PROGBITS
+
+# TO:        Index: 1
+# TO-NEXT:   Name:  (0)
+# FROM:      Index: 1
+# FROM-NEXT: Name: .bar (

test/tools/llvm-objcopy/ELF/shared-strtab-shstrtab.s

+# REQUIRES: x86-registered-target
+
+## It is possible for the section header table and symbol table to share the
+## same string table for storing section and symbol names. This test shows that
+## under various circumstances, the names are still correct after llvm-objcopy
+## has copied such an object file, and that the name table is still shared.
+## This test uses the assembler rather than yaml2obj because yaml2obj generates
+## separate string tables, whereas the assembler shares them.
+
+# RUN: llvm-mc -filetype=obj -triple x86_64-pc-linux %s -o %t.o
+## Sanity check that the string tables are shared:
+# RUN: llvm-readobj --section-headers %t.o \
+# RUN:   | FileCheck %s --check-prefix=VALIDATE --implicit-check-not=.shstrtab
+
+# VALIDATE: Name: .strtab
+
+## Case 1: basic copy.
+# RUN: llvm-objcopy %t.o %t.basic
+# RUN: llvm-readobj --section-headers --symbols %t.basic \
+# RUN:   | FileCheck %s --check-prefix=BASIC --implicit-check-not=.shstrtab
+
+# BASIC: Sections [
+# BASIC: Name: .foo (
+# BASIC: Name: .strtab (
+# BASIC: Symbols [
+# BASIC: Name: foo (
+
+## Case 2: renaming a section.
+# RUN: llvm-objcopy %t.o %t.rename-section --rename-section .foo=.oof
+# RUN: llvm-readobj --section-headers --symbols %t.rename-section \
+# RUN:   | FileCheck %s --check-prefix=SECTION-RENAME --implicit-check-not=.shstrtab
+
+# SECTION-RENAME: Sections [
+# SECTION-RENAME: Name: .oof (
+# SECTION-RENAME: Name: .strtab (
+# SECTION-RENAME: Symbols [
+# SECTION-RENAME: Name: foo (
+
+## Case 3: renaming a symbol.
+# RUN: llvm-objcopy %t.o %t.redefine-symbol --redefine-sym foo=oof
+# RUN: llvm-readobj --section-headers --symbols %t.redefine-symbol \
+# RUN:   | FileCheck %s --check-prefix=SYMBOL-RENAME --implicit-check-not=.shstrtab
+
+# SYMBOL-RENAME: Sections [
+# SYMBOL-RENAME: Name: .foo (
+# SYMBOL-RENAME: Name: .strtab (
+# SYMBOL-RENAME: Symbols [
+# SYMBOL-RENAME: Name: oof (
+
+## Case 4: removing a section.
+# RUN: llvm-objcopy %t.o %t.remove-section -R .foo
+# RUN: llvm-readobj --section-headers --symbols %t.remove-section \
+# RUN:   | FileCheck %s --check-prefix=SECTION-REMOVE --implicit-check-not=.shstrtab --implicit-check-not=.foo
+
+# SECTION-REMOVE: Sections [
+# SECTION-REMOVE: Name: .strtab (
+# SECTION-REMOVE: Symbols [
+# SECTION-REMOVE: Name: foo (
+
+## Case 5: removing a symbol.
+# RUN: llvm-objcopy %t.o %t.remove-symbol -N foo
+# RUN: llvm-readobj --section-headers --symbols %t.remove-symbol \
+# RUN:   | FileCheck %s --check-prefix=SYMBOL-REMOVE --implicit-check-not=.shstrtab --implicit-check-not=foo
+
+# SYMBOL-REMOVE: Sections [
+# SYMBOL-REMOVE: Name: .foo (
+# SYMBOL-REMOVE: Name: .strtab (
+# SYMBOL-REMOVE: Symbols [
+
+## Case 6: adding a section.
+# RUN: llvm-objcopy %t.o %t.add-section --add-section .bar=%s
+# RUN: llvm-readobj --section-headers --symbols %t.add-section \
+# RUN:   | FileCheck %s --check-prefix=SECTION-ADD --implicit-check-not=.shstrtab
+
+# SECTION-ADD: Sections [
+# SECTION-ADD: Name: .foo (
+# SECTION-ADD: Name: .strtab (
+# SECTION-ADD: Name: .bar (
+# SECTION-ADD: Symbols [
+# SECTION-ADD: Name: foo (
+
+## Case 7: adding a symbol.
+# RUN: llvm-objcopy %t.o %t.add-symbol --add-symbol bar=0x1234
+# RUN: llvm-readobj --section-headers --symbols %t.add-symbol \
+# RUN:   | FileCheck %s --check-prefix=SYMBOL-ADD --implicit-check-not=.shstrtab
+
+# SYMBOL-ADD: Sections [
+# SYMBOL-ADD: Name: .foo (
+# SYMBOL-ADD: Name: .strtab (
+# SYMBOL-ADD: Symbols [
+# SYMBOL-ADD: Name: foo (
+# SYMBOL-ADD: Name: bar (
+
+## Case 8: removing all symbols.
+# RUN: llvm-objcopy %t.o %t.strip-all --strip-all
+# RUN: llvm-readobj --section-headers --symbols %t.strip-all \
+# RUN:   | FileCheck %s --check-prefix=STRIP-ALL --implicit-check-not=.shstrtab
+
+# STRIP-ALL:      Sections [
+# STRIP-ALL:      Name: .foo (
+# STRIP-ALL:      Name: .strtab (
+# STRIP-ALL:      Symbols [
+# STRIP-ALL-NEXT: ]
+
+.section .foo,"a",@progbits
+foo = 0x4321

test/tools/llvm-objcopy/ELF/strip-sections.test

 # RUN: yaml2obj %s > %t
 # RUN: llvm-objcopy --strip-sections %t %t2
 # RUN: llvm-readobj --file-headers --program-headers %t2 | FileCheck %s
 # RUN: od -t x1 -j 4096 %t2 | FileCheck %s --check-prefix=DATA
 
+## Sanity check the DATA-NOT line by showing that "fe ed fa ce" appears
+## if --strip-sections is not specified.
+# RUN: llvm-objcopy %t %t3
+# RUN: od -t x1 -j 4096 %t3 | FileCheck %s --check-prefix=VALIDATE
+
 !ELF
 FileHeader:
   Class:           ELFCLASS64
   Data:            ELFDATA2LSB
   Type:            ET_EXEC
   Machine:         EM_X86_64
 Sections:
   - Name:            .text
     Type:            SHT_PROGBITS
     Flags:           [ SHF_ALLOC, SHF_EXECINSTR ]
     AddressAlign:    0x0000000000001000
     Content:         "DEADBEEF"
   - Name:            .non_alloc_in_segment
     Type:            SHT_PROGBITS
     Flags:           [ ]
     Content:         "CAFEBABE"
+  - Name:            .non_alloc_not_in_segment
+    Type:            SHT_PROGBITS
+    Flags:           [ ]
+    Content:         "FEEDFACE"
 ProgramHeaders:
 - Type: PT_LOAD
   Flags: [ PF_X, PF_R ]
   Sections:
     - Section: .text
     - Section: .non_alloc_in_segment
 
 #DATA: 0010000 de ad be ef ca fe ba be
+#DATA-NOT: fe ed fa ce
+
+#VALIDATE: 0010000 de ad be ef ca fe ba be fe ed fa ce
 
 #CHECK: ElfHeader {
 #CHECK-NEXT:  Ident {
 #CHECK-NEXT:     Magic: (7F 45 4C 46)
 #CHECK-NEXT:     Class: 64-bit (0x2)
 #CHECK-NEXT:     DataEncoding: LittleEndian (0x1)
 #CHECK-NEXT:     FileVersion: 1
 #CHECK-NEXT:     OS/ABI: SystemV (0x0)