This is an archive of the discontinued LLVM Phabricator instance.

Differential D74879

[llvm-objcopy][test] Improve empty section tests
ClosedPublic

Authored by MaskRay on Feb 19 2020, 4:17 PM.

Details

Reviewers
grimar
jhenderson
rupprecht
• espindola
alexander-shaposhnikov
Commits
rG6503aadcc6c9: [llvm-objcopy][test] Improve empty section tests
Summary

empty-sections.test: add two tests adapted from @jhenderson's https://reviews.llvm.org/D74755#1882221
strip-non-alloc.test: improve. D74755 will change the attribution of an empty section.

They mostly test the behavior of Object.cpp:sectionWithinSegment : how we attribute sections to segments.
ParentSegment can affect some subtle layout decisions.

Diff Detail

Event Timeline

MaskRay created this revision.Feb 19 2020, 4:17 PM
Herald added a reviewer: • espindola. · View Herald TranscriptFeb 19 2020, 4:17 PM
Herald added a reviewer: alexander-shaposhnikov. · View Herald Transcript
Herald added a project: Restricted Project. · View Herald Transcript
Herald added subscribers: llvm-commits, abrachet, emaste. · View Herald Transcript
MaskRay added a child revision: D74755: [llvm-objcopy] Attribute an empty section to a segment ending at its address.Feb 19 2020, 4:20 PM
MaskRay mentioned this in D74755: [llvm-objcopy] Attribute an empty section to a segment ending at its address.Feb 19 2020, 4:34 PM
Harbormaster completed remote builds in B46873: Diff 245551.Feb 19 2020, 4:52 PM
grimar added inline comments.Feb 20 2020, 12:11 AM
llvm/test/tools/llvm-objcopy/ELF/empty-section.test
5

Can you add a comment here and for the test below?

77

[SHF_ALLOC] is one byte off.

MaskRay updated this revision to Diff 246077.Feb 22 2020, 8:37 AM
MaskRay marked 3 inline comments as done.

Improve tests

Harbormaster failed remote builds in B47090: Diff 246077!Feb 22 2020, 8:52 AM
MaskRay updated this revision to Diff 246078.Feb 22 2020, 9:19 AM
MaskRay edited the summary of this revision. (Show Details)

Improve tests

Harbormaster failed remote builds in B47091: Diff 246078!Feb 22 2020, 9:38 AM
jhenderson added inline comments.Feb 25 2020, 1:17 AM
llvm/test/tools/llvm-objcopy/ELF/empty-section.test
6–12

Seems to me like here and below you can get rid of the columns after Off?

39

according to the

llvm/test/tools/llvm-objcopy/ELF/strip-non-alloc.test
11–12

Shouldn't these two be on the same line?

MaskRay updated this revision to Diff 246527.Feb 25 2020, 12:03 PM
MaskRay marked 3 inline comments as done.

Address comments

MaskRay added inline comments.Feb 25 2020, 12:07 PM
llvm/test/tools/llvm-objcopy/ELF/empty-section.test
5

The original test is not very useful. I will replace it with --docnum=2 with segments removed.

Harbormaster completed remote builds in B47228: Diff 246527.Feb 25 2020, 1:34 PM
jhenderson accepted this revision.Feb 26 2020, 1:54 AM

LGTM.

This revision is now accepted and ready to land.Feb 26 2020, 1:54 AM
Closed by commit rG6503aadcc6c9: [llvm-objcopy][test] Improve empty section tests (authored by MaskRay). · Explain WhyFeb 26 2020, 9:00 AM
This revision was automatically updated to reflect the committed changes.

Revision Contents

PathSize
llvm/
test/
tools/
llvm-objcopy/
ELF/
135 lines
50 lines

Diff 246757

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

# RUN: yaml2obj %s -o %t ## .empty is not covered by a segment. Its offset is moved to 0x200 to fill the gap.
# RUN: llvm-objcopy %t %t2 # RUN: yaml2obj --docnum=1 %s -o %t1
# RUN: llvm-readobj --sections %t2 | FileCheck %s # RUN: llvm-objcopy %t1 %t1.out
# RUN: llvm-readelf -S %t1.out | FileCheck --check-prefix=CHECK1 %s
grimarUnsubmitted
Done
ReplyInline Actions

Can you add a comment here and for the test below?

grimar: Can you add a comment here and for the test below?
MaskRayAuthorUnsubmitted
Done
ReplyInline Actions

The original test is not very useful. I will replace it with --docnum=2 with segments removed.

MaskRay: The original test is not very useful. I will replace it with --docnum=2 with segments removed.
!ELF # CHECK1: Name Type Address Off
# CHECK1-NEXT: NULL 0000000000000000 000000
# CHECK1-NEXT: .foo PROGBITS 0000000000000000 000100
# CHECK1-NEXT: .empty PROGBITS 0000000000001000 000200
# CHECK1-NEXT: .baz PROGBITS 0000000000001000 001000
--- !ELF
jhendersonUnsubmitted
Done
ReplyInline Actions

Seems to me like here and below you can get rid of the columns after Off?

jhenderson: Seems to me like here and below you can get rid of the columns after `Off`?
FileHeader: FileHeader:
Class: ELFCLASS64 Class: ELFCLASS64
Data: ELFDATA2LSB Data: ELFDATA2LSB
Type: ET_EXEC Type: ET_EXEC
Machine: EM_X86_64 Machine: EM_X86_64
Sections: Sections:
- Name: .text - Name: gap
Type: Fill
Size: 0xE00
- Name: .foo
Type: SHT_PROGBITS Type: SHT_PROGBITS
Flags: [ SHF_ALLOC, SHF_EXECINSTR ] Flags: [ SHF_ALLOC ]
Address: 0x0 AddressAlign: 0x100
AddressAlign: 0x0000000000001000 Address: 0
Content: "00000000" Size: 0x100
- Name: .empty - Name: .empty
Type: SHT_PROGBITS Type: SHT_PROGBITS
Flags: [ SHF_ALLOC ] Flags: [ SHF_ALLOC ]
Address: 0x1000 Address: 0x1000
AddressAlign: 0x0000000000001000 - Name: .baz
Content: ""
- Name: .data
Type: SHT_PROGBITS Type: SHT_PROGBITS
Flags: [ SHF_ALLOC ] Flags: [ SHF_ALLOC ]
AddressAlign: 0x1000
Address: 0x1000 Address: 0x1000
AddressAlign: 0x0000000000001000 Size: 0x100
Content: "00000000"
## Test that we attribute .empty to the second segment. Its offset is assigned according to
jhendersonUnsubmitted
Done
ReplyInline Actions

according to the

jhenderson: according to the
## the p_offset of the second segment.
# RUN: yaml2obj --docnum=2 %s -o %t2
# RUN: llvm-objcopy %t2 %t2.out
# RUN: llvm-readelf -S -l %t2.out | FileCheck --check-prefix=CHECK2 %s
# CHECK2: Name Type Address Off
# CHECK2-NEXT: NULL 0000000000000000 000000
# CHECK2-NEXT: .foo PROGBITS 0000000000000000 000100
# CHECK2-NEXT: .empty PROGBITS 0000000000001000 001000
# CHECK2-NEXT: .baz PROGBITS 0000000000001000 001000
# CHECK: Name: .text # CHECK2: Type Offset VirtAddr PhysAddr FileSiz MemSiz Flg Align
# CHECK-NEXT: Type: SHT_PROGBITS # CHECK2-NEXT: LOAD 0x000100 0x0000000000000000 0x0000000000000000 0x000100 0x000100 0x100
# CHECK-NEXT: Flags [ # CHECK2-NEXT: LOAD 0x001000 0x0000000000001000 0x0000000000001000 0x000100 0x000100 0x1000
# CHECK-NEXT: SHF_ALLOC
# CHECK-NEXT: SHF_EXECINSTR --- !ELF
# CHECK-NEXT: ] FileHeader:
Class: ELFCLASS64
# CHECK: Name: .empty Data: ELFDATA2LSB
# CHECK-NEXT: Type: SHT_PROGBITS Type: ET_EXEC
# CHECK-NEXT: Flags [ Machine: EM_X86_64
# CHECK-NEXT: SHF_ALLOC Sections:
# CHECK-NEXT: ] - Name: gap
# CHECK-NEXT: Address: 0x1000 Type: Fill
# CHECK-NEXT: Offset: 0x2000 Size: 0xE00
# CHECK-NEXT: Size: 0 - Name: .foo
Type: SHT_PROGBITS
# CHECK: Name: .data Flags: [ SHF_ALLOC ]
# CHECK-NEXT: Type: SHT_PROGBITS AddressAlign: 0x100
# CHECK-NEXT: Flags [ Address: 0
# CHECK-NEXT: SHF_ALLOC Size: 0x100
# CHECK-NEXT: ] - Name: .empty
# CHECK-NEXT: Address: 0x1000 Type: SHT_PROGBITS
# CHECK-NEXT: Offset: 0x2000 Flags: [ SHF_ALLOC ]
# CHECK-NEXT: Size: 4 Address: 0x1000
- Name: .baz
Type: SHT_PROGBITS
Flags: [ SHF_ALLOC ]
grimarUnsubmitted
Done
ReplyInline Actions

[SHF_ALLOC] is one byte off.

grimar: `[SHF_ALLOC]` is one byte off.
AddressAlign: 0x1000
Address: 0x1000
Size: 0x100
ProgramHeaders:
- Type: PT_LOAD
VAddr: 0
PAddr: 0
Align: 0x100
Sections:
- Section: .foo
- Type: PT_LOAD
VAddr: 0x1000
PAddr: 0x1000
Align: 0x1000
Sections:
- Section: .empty
- Section: .baz

llvm/test/tools/llvm-objcopy/ELF/strip-non-alloc.test

# RUN: yaml2obj %s -o %t # RUN: yaml2obj --docnum=1 %s -o %t
# RUN: llvm-objcopy --strip-non-alloc %t %t2 # RUN: llvm-objcopy --strip-non-alloc %t %t.out
# RUN: llvm-readobj --file-headers --sections %t2 | FileCheck %s # RUN: llvm-readobj --file-headers --sections %t.out | FileCheck %s
!ELF # CHECK: SectionHeaderCount: 5
# CHECK: Name: non_alloc_in_segment
# CHECK: Name: .bss
# CHECK: Name: .text
# CHECK: Name: .shstrtab
--- !ELF
FileHeader: FileHeader:
jhendersonUnsubmitted
Done
ReplyInline Actions

Shouldn't these two be on the same line?

jhenderson: Shouldn't these two be on the same line?
Class: ELFCLASS64 Class: ELFCLASS64
Data: ELFDATA2LSB Data: ELFDATA2LSB
Type: ET_REL Type: ET_REL
Machine: EM_X86_64 Machine: EM_X86_64
Sections: Sections:
- Name: non_alloc_in_segment - Name: non_alloc_in_segment
Type: SHT_PROGBITS Type: SHT_PROGBITS
Flags: [ ] Flags: [ ]
Size: 4 Size: 4
- Name: .bss - Name: .bss
Type: SHT_NOBITS Type: SHT_NOBITS
Flags: [ SHF_ALLOC ] Flags: [ SHF_ALLOC ]
- Name: .text - Name: .text
Type: SHT_PROGBITS Type: SHT_PROGBITS
Flags: [ SHF_ALLOC, SHF_EXECINSTR ] Flags: [ SHF_ALLOC, SHF_EXECINSTR ]
- Name: .blarg
Type: SHT_PROGBITS
Flags: [ ]
ProgramHeaders: ProgramHeaders:
# Use an arbitrary segment type to show that the segment type is unimportant. # Use an arbitrary segment type to show that the segment type is unimportant.
- Type: 0x61234567 - Type: 0x61234567
Sections: Sections:
- Section: non_alloc_in_segment - Section: non_alloc_in_segment
# CHECK: SectionHeaderCount: 5 # RUN: yaml2obj --docnum=2 %s -o %t2
# RUN: llvm-objcopy --strip-non-alloc %t2 %t2.out
# RUN: llvm-readobj --file-headers --sections %t2.out | FileCheck --check-prefix=CHECK2 %s
# CHECK: Name: non_alloc_in_segment # CHECK2: SectionHeaderCount: 3
# CHECK: Name: .bss # CHECK2: Name: .text
# CHECK: Name: .text # CHECK2: Name: .shstrtab
# CHECK: Name: .shstrtab
--- !ELF
FileHeader:
Class: ELFCLASS64
Data: ELFDATA2LSB
Type: ET_REL
Machine: EM_X86_64
Sections:
- Name: .text
Type: SHT_PROGBITS
Flags: [ SHF_ALLOC, SHF_EXECINSTR ]
## empty_trailing_non_alloc is considered included by the segment,
## so it will be retained.
- Name: empty_trailing_non_alloc
Type: SHT_PROGBITS
- Name: trailing_non_alloc
Type: SHT_PROGBITS
Content: 00
ProgramHeaders:
- Type: 0x61234567
Sections:
- Section: .text