[llvm-libtool-darwin] Add support for -arch_only

[sameerarora101]

Summary

Add support for -arch_only option for llvm-libtool-darwin. This diff
also adds support for accepting universal files as input and flattening
them to create the required static library. Supports input universal
files contaning both Mach-O object files or archives.

Differences from cctools' libtool:

• -arch_only can be specified multiple times
• archives containing universal files are considered invalid (libtool allows such archives)

Depends on D84209.

[sameerarora101]

Resolving clang-tidy issues

[sameerarora101]

Rebase

[sameerarora101]

Add documentation for -arch_only option.

[jhenderson]

llvm/docs/CommandGuide/llvm-libtool-darwin.rst

64–67 ↗

(On Diff #281738)

I should have been paying more attention to this before, but it would be good to have the options listed in alphabetical order, I think.

llvm/test/tools/llvm-libtool-darwin/cpu-subtype-matching.test

23

Nit: delete extra blank line.

llvm/test/tools/llvm-libtool-darwin/universal-file-flattening.test

90–91

Nit: here and in EMPTY-ARCH, add spaces to make the first line's pattern line up with the one below.

llvm/tools/llvm-libtool-darwin/llvm-libtool-darwin.cpp

108

Error messages should generally start with a lower-case letter (i.e. "Invalid" -> "invalid")

114

Given that there's no File parameter, perhaps thsi should be "a file's"

117

uint32_t for these, to match the return type of getCPUTypeFromArchitecture.

124–126

Rather than a comment that will be forgotten about and/or ignored, could you not just add the constant?

223

Nit: delete blank line here.

230

Don't use auto here, please.

246

Why are you throwing away this error rahter than reporting it earlier?

[sameerarora101]

llvm/docs/CommandGuide/llvm-libtool-darwin.rst

64–67 ↗

(On Diff #281738)

Ok, I have ordered them alphabetically now.

llvm/tools/llvm-libtool-darwin/llvm-libtool-darwin.cpp

124–126

Is it as simple as adding a constant? Nothing else needs to be done?

Here is a potential place where I think the constant should go (under MachO.h):

enum CPUSubTypeARM {
  CPU_SUBTYPE_ARM_ALL = 0,
  CPU_SUBTYPE_ARM_V4T = 5,
  CPU_SUBTYPE_ARM_V6 = 6,
  CPU_SUBTYPE_ARM_V5 = 7,
  CPU_SUBTYPE_ARM_V5TEJ = 7,
  CPU_SUBTYPE_ARM_XSCALE = 8,
  CPU_SUBTYPE_ARM_V7 = 9,
  //  unused  ARM_V7F     = 10,
  CPU_SUBTYPE_ARM_V7S = 11,
  CPU_SUBTYPE_ARM_V7K = 12,
  CPU_SUBTYPE_ARM_V6M = 14,
  CPU_SUBTYPE_ARM_V7M = 15,
  CPU_SUBTYPE_ARM_V7EM = 16
};

enum CPUSubTypeARM64 {
  CPU_SUBTYPE_ARM64_ALL = 0,
  CPU_SUBTYPE_ARM64E = 2,
};

enum CPUSubTypeARM64_32 { CPU_SUBTYPE_ARM64_32_V8 = 1 };

However, I am also unsure of the RHS value that needs to be assigned to the constant. Any ideas?

246

So a universal member can either be a MachOObjectFile or an Archive - so we need to try casting the member in both ways. First I try casting it as an object file. If that gives an error, it might be the case that the member is an Archive. Now, if I succeed on casting the member as an Archive, then I can throw away the error that I received from casting the member as an object file as the member was an archive and should not have been casted as an object file in first place.

llvm-lipo also seems to be following this design (under printBinaryArchs):

for (const auto &O : UO->objects()) {
      Expected<std::unique_ptr<MachOObjectFile>> MachOObjOrError =
          O.getAsObjectFile();
      if (MachOObjOrError) {
        OS << Slice(MachOObjOrError->get()).getArchString() << " ";
        continue;
      }
      Expected<std::unique_ptr<Archive>> ArchiveOrError = O.getAsArchive();
      if (ArchiveOrError) {
        consumeError(MachOObjOrError.takeError());
        OS << Slice(ArchiveOrError->get()).getArchString() << " ";
        continue;
      }
      consumeError(ArchiveOrError.takeError());
      reportError(Binary->getFileName(), MachOObjOrError.takeError());
    }

[sameerarora101]

Alphabetically ordering docs.

[sameerarora101]

llvm/tools/llvm-libtool-darwin/llvm-libtool-darwin.cpp

124–126

Added diff for constant V8 here: D85041

[jhenderson]

llvm/tools/llvm-libtool-darwin/llvm-libtool-darwin.cpp

108

We tend to avoid explicit new lines in error messages. They can make it non-obvious that output belongs to the error message, when stderr and stdout are piped to the same location, which can be confusing. You can probably just have this as a single sentence (i.e. "invalid architecture 'foo': valid architecture names are ..."). Note that I also added the ArchitectureName to the message, since that gives a little more context as to what is wrong.

124–126

I think your other patch is fine. It may be beneficial to look at the history of the file you're modifiying to see what else got changed when other subtypes were added.

246

Okay, makes a degree of sense. consumeError calls should really have a comment explaining why it's safe to throw them away though, so please add one.

[sameerarora101]

llvm/tools/llvm-libtool-darwin/llvm-libtool-darwin.cpp

124–126

Yup, thanks. Furthermore, I'll also wait on D85041 to get landed so that I can make modifications here.

[sameerarora101]

Invalid architecture error msg and consumeError comment.

[jhenderson]

All other changes look fine. Just waiting on D85041 then.

[sameerarora101]

Rebase.

[sameerarora101]

Rebase after adding constant V8.

[sameerarora101]

I think I need to add V8 support under llvm/include/llvm/TextAPI/MachO/Architecture.def here:

///
/// ARM64 architectures sorted by cpu sub type id.
///
ARCHINFO(arm64, MachO::CPU_TYPE_ARM64, MachO::CPU_SUBTYPE_ARM64_ALL, 64)
ARCHINFO(arm64e, MachO::CPU_TYPE_ARM64, MachO::CPU_SUBTYPE_ARM64E, 64)

What should be the first element for that tuple in case of V8 (arm64_v8 or arm64v8 or maybe something else) ? Thanks.

[smeenai]

In D84770#2204147, @sameerarora101 wrote:

I think I need to add V8 support under llvm/include/llvm/TextAPI/MachO/Architecture.def here:

///
/// ARM64 architectures sorted by cpu sub type id.
///
ARCHINFO(arm64, MachO::CPU_TYPE_ARM64, MachO::CPU_SUBTYPE_ARM64_ALL, 64)
ARCHINFO(arm64e, MachO::CPU_TYPE_ARM64, MachO::CPU_SUBTYPE_ARM64E, 64)

What should be the first element for that tuple in case of V8 (arm64_v8 or arm64v8 or maybe something else) ? Thanks.

Hmm, what necessitates the TAPI patch? arm64_v8 seems fine to me, though @ributzka and @cishida should weigh in there.

[sameerarora101]

In D84770#2208826, @smeenai wrote:

In D84770#2204147, @sameerarora101 wrote:

I think I need to add V8 support under llvm/include/llvm/TextAPI/MachO/Architecture.def here:

///
/// ARM64 architectures sorted by cpu sub type id.
///
ARCHINFO(arm64, MachO::CPU_TYPE_ARM64, MachO::CPU_SUBTYPE_ARM64_ALL, 64)
ARCHINFO(arm64e, MachO::CPU_TYPE_ARM64, MachO::CPU_SUBTYPE_ARM64E, 64)

What should be the first element for that tuple in case of V8 (arm64_v8 or arm64v8 or maybe something else) ? Thanks.

Hmm, what necessitates the TAPI patch? arm64_v8 seems fine to me, though @ributzka and @cishida should weigh in there.

Converting a yaml file with Subtype V8 to an object file requires this patch. I was trying to add a test for V8, but my yaml was incorrectly parsed as an object file with unknown architecture.

[cishida]

In D84770#2210049, @sameerarora101 wrote:

In D84770#2208826, @smeenai wrote:

In D84770#2204147, @sameerarora101 wrote:

I think I need to add V8 support under llvm/include/llvm/TextAPI/MachO/Architecture.def here:

///
/// ARM64 architectures sorted by cpu sub type id.
///
ARCHINFO(arm64, MachO::CPU_TYPE_ARM64, MachO::CPU_SUBTYPE_ARM64_ALL, 64)
ARCHINFO(arm64e, MachO::CPU_TYPE_ARM64, MachO::CPU_SUBTYPE_ARM64E, 64)

What should be the first element for that tuple in case of V8 (arm64_v8 or arm64v8 or maybe something else) ? Thanks.

Hmm, what necessitates the TAPI patch? arm64_v8 seems fine to me, though @ributzka and @cishida should weigh in there.

Converting a yaml file with Subtype V8 to an object file requires this patch. I was trying to add a test for V8, but my yaml was incorrectly parsed as an object file with unknown architecture.

Ah, yes reading in tbd files as input will require this addition. I have no strong opinion, but I prefer for arm64v8 as will cause less downstream disruption.

[jhenderson]

LGTM, but please wait for other stakeholders.

[smeenai]

LGTM.

Let's commit this one as-is, and you can add arm64v8 to TextAPI and add the corresponding test here as a follow-up.

llvm/tools/llvm-libtool-darwin/llvm-libtool-darwin.cpp

131

Nit: you don't need braces, since both sides are simple return expressions (the first one is multi-line, but I think it's pretty clear it's part of the same statement).

[sameerarora101]

In D84770#2214390, @smeenai wrote:

LGTM.

Let's commit this one as-is, and you can add arm64v8 to TextAPI and add the corresponding test here as a follow-up.

Got it. I also realized yesterday that adding arm64v8 to TextAPI is not the only thing necessary for supporting v8 tests. We would have to update llvm/lib/Object/MachOObjectFile.cpp file as well to be able to recognize the Triple() correctly. I saw your response on D85334. It makes more sense to add the test after arm64e/arm64v8 support is there (as separate patch only). Thanks, I'll commit this now

[sameerarora101]

Remove redundant braces.

llvm/test/tools/llvm-libtool-darwin/cpu-subtype-matching.test

+## This test checks that the CPU subtype matching logic is handled correctly.
+
+# RUN: yaml2obj %s --docnum=1 -o %t.armv6
+# RUN: yaml2obj %s --docnum=2 -o %t.armv7
+
+# RUN: llvm-libtool-darwin -static -o %t.lib %t.armv6 %t.armv7 -arch_only armv7
+
+## Check that only armv7 binary is present:
+# RUN: llvm-ar t %t.lib | \
+# RUN:   FileCheck %s --check-prefix=ARM-NAMES --implicit-check-not={{.}} -DPREFIX=%basename_t.tmp
+
+# ARM-NAMES: [[PREFIX]].armv7
+
+## Check that only armv7 symbol is present:
+# RUN: llvm-nm --print-armap %t.lib | \
+# RUN:   FileCheck %s --check-prefix=ARM-SYMBOLS -DPREFIX=%basename_t.tmp --match-full-lines
+
+# ARM-SYMBOLS:      Archive map
+# ARM-SYMBOLS-NEXT: _armv7 in [[PREFIX]].armv7
+# ARM-SYMBOLS-EMPTY:
+
+
+## armv6.yaml
+## CPUTYPE:    CPU_TYPE_ARM
+## CPUSUBTYPE: CPU_SUBTYPE_ARM_V6
+--- !mach-o
+FileHeader:
+  magic:           0xFEEDFACE
+  cputype:         0x0000000C
+  cpusubtype:      0x00000006
+  filetype:        0x00000001
+  ncmds:           2
+  sizeofcmds:      148
+  flags:           0x00002000
+LoadCommands:
+  - cmd:             LC_SEGMENT
+    cmdsize:         124
+    segname:         ''
+    vmaddr:          0
+    vmsize:          24
+    fileoff:         296
+    filesize:        24
+    maxprot:         7
+    initprot:        7
+    nsects:          1
+    flags:           0
+    Sections:
+      - sectname:        __text
+        segname:         __TEXT
+        addr:            0x0000000000000000
+        size:            24
+        offset:          0x00000128
+        align:           2
+        reloff:          0x00000000
+        nreloc:          0
+        flags:           0x80000400
+        reserved1:       0x00000000
+        reserved2:       0x00000000
+        reserved3:       0x00000000
+        content:         04D04DE208009FE500008DE504D08DE21EFF2FE100000000
+  - cmd:             LC_SYMTAB
+    cmdsize:         24
+    symoff:          328
+    nsyms:           1
+    stroff:          340
+    strsize:         8
+LinkEditData:
+  NameList:
+    - n_strx:          1
+      n_type:          0x0F
+      n_sect:          1
+      n_desc:          0
+      n_value:         0
+  StringTable:
+    - ''
+    - _armv6
+    - ''
+...
+
+## armv7.yaml
+## CPUTYPE:    CPU_TYPE_ARM
+## CPUSUBTYPE: CPU_SUBTYPE_ARM_V7
+--- !mach-o
+FileHeader:
+  magic:           0xFEEDFACE
+  cputype:         0x0000000C
+  cpusubtype:      0x00000009
+  filetype:        0x00000001
+  ncmds:           2
+  sizeofcmds:      148
+  flags:           0x00002000
+LoadCommands:
+  - cmd:             LC_SEGMENT
+    cmdsize:         124
+    segname:         ''
+    vmaddr:          0
+    vmsize:          10
+    fileoff:         280
+    filesize:        10
+    maxprot:         7
+    initprot:        7
+    nsects:          1
+    flags:           0
+    Sections:
+      - sectname:        __text
+        segname:         __TEXT
+        addr:            0x0000000000000000
+        size:            10
+        offset:          0x00000118
+        align:           1
+        reloff:          0x00000000
+        nreloc:          0
+        flags:           0x80000400
+        reserved1:       0x00000000
+        reserved2:       0x00000000
+        reserved3:       0x00000000
+        content:         81B00020009001B07047
+  - cmd:             LC_SYMTAB
+    cmdsize:         24
+    symoff:          292
+    nsyms:           1
+    stroff:          304
+    strsize:         8
+LinkEditData:
+  NameList:
+    - n_strx:          1
+      n_type:          0x0F
+      n_sect:          1
+      n_desc:          8
+      n_value:         0
+  StringTable:
+    - ''
+    - _armv7
+    - ''
+...
+
+# RUN: yaml2obj %s --docnum=3 -o %t.x86_64
+# RUN: yaml2obj %s --docnum=4 -o %t.x86_64_h
+
+# RUN: llvm-libtool-darwin -static -o %t.lib %t.x86_64 %t.x86_64_h -arch_only x86_64
+
+## Check that only x86_64 binary is present:
+# RUN: llvm-ar t %t.lib | \
+# RUN:   FileCheck %s --check-prefix=X86-NAMES --implicit-check-not={{.}} -DPREFIX=%basename_t.tmp
+
+# X86-NAMES: [[PREFIX]].x86_64
+
+## Check that only x86_64 symbol is present:
+# RUN: llvm-nm --print-armap %t.lib | \
+# RUN:   FileCheck %s --check-prefix=X86-SYMBOLS -DPREFIX=%basename_t.tmp --match-full-lines
+
+# X86-SYMBOLS:      Archive map
+# X86-SYMBOLS-NEXT: _x86_64 in [[PREFIX]].x86_64
+# X86-SYMBOLS-EMPTY:
+
+## x86_64.yaml
+## CPUTYPE:    CPU_TYPE_X86_64
+## CPUSUBTYPE: CPU_SUBTYPE_X86_64_ALL
+--- !mach-o
+FileHeader:
+  magic:           0xFEEDFACF
+  cputype:         0x01000007
+  cpusubtype:      0x00000003
+  filetype:        0x00000001
+  ncmds:           2
+  sizeofcmds:      176
+  flags:           0x00002000
+  reserved:        0x00000000
+LoadCommands:
+  - cmd:             LC_SEGMENT_64
+    cmdsize:         152
+    segname:         ''
+    vmaddr:          0
+    vmsize:          15
+    fileoff:         312
+    filesize:        15
+    maxprot:         7
+    initprot:        7
+    nsects:          1
+    flags:           0
+    Sections:
+      - sectname:        __text
+        segname:         __TEXT
+        addr:            0x0000000000000000
+        size:            15
+        offset:          0x00000138
+        align:           4
+        reloff:          0x00000000
+        nreloc:          0
+        flags:           0x80000400
+        reserved1:       0x00000000
+        reserved2:       0x00000000
+        reserved3:       0x00000000
+        content:         554889E531C0C745FC000000005DC3
+  - cmd:             LC_SYMTAB
+    cmdsize:         24
+    symoff:          328
+    nsyms:           1
+    stroff:          344
+    strsize:         8
+LinkEditData:
+  NameList:
+    - n_strx:          1
+      n_type:          0x0F
+      n_sect:          1
+      n_desc:          0
+      n_value:         0
+  StringTable:
+    - ''
+    - _x86_64
+    - ''
+...
+
+## x86_64h.yaml
+## CPUTYPE:    CPU_TYPE_X86_64
+## CPUSUBTYPE: CPU_SUBTYPE_X86_64_H
+--- !mach-o
+FileHeader:
+  magic:           0xFEEDFACF
+  cputype:         0x01000007
+  cpusubtype:      0x00000008
+  filetype:        0x00000001
+  ncmds:           2
+  sizeofcmds:      176
+  flags:           0x00002000
+  reserved:        0x00000000
+LoadCommands:
+  - cmd:             LC_SEGMENT_64
+    cmdsize:         152
+    segname:         ''
+    vmaddr:          0
+    vmsize:          15
+    fileoff:         312
+    filesize:        15
+    maxprot:         7
+    initprot:        7
+    nsects:          1
+    flags:           0
+    Sections:
+      - sectname:        __text
+        segname:         __TEXT
+        addr:            0x0000000000000000
+        size:            15
+        offset:          0x00000138
+        align:           4
+        reloff:          0x00000000
+        nreloc:          0
+        flags:           0x80000400
+        reserved1:       0x00000000
+        reserved2:       0x00000000
+        reserved3:       0x00000000
+        content:         554889E531C0C745FC000000005DC3
+  - cmd:             LC_SYMTAB
+    cmdsize:         24
+    symoff:          328
+    nsyms:           1
+    stroff:          344
+    strsize:         8
+LinkEditData:
+  NameList:
+    - n_strx:          1
+      n_type:          0x0F
+      n_sect:          1
+      n_desc:          0
+      n_value:         0
+  StringTable:
+    - ''
+    - _x86_64_h
+    - ''
+...

llvm/test/tools/llvm-libtool-darwin/universal-file-flattening.test

+## This test checks that a universal file is flattened correctly.
+
+# RUN: yaml2obj %s -o %t-universal.o
+# RUN: yaml2obj %S/Inputs/input1.yaml -o %t-input1.o
+# RUN: yaml2obj %S/Inputs/input2.yaml -o %t-input2.o
+
+# RUN: llvm-libtool-darwin -static -o %t.lib %t-universal.o -arch_only arm64
+
+## Check that the binary is present:
+# RUN: llvm-ar t %t.lib | \
+# RUN:   FileCheck %s --check-prefix=CHECK-NAMES --implicit-check-not={{.}} -DPREFIX=%basename_t.tmp
+
+# CHECK-NAMES: [[PREFIX]]-universal.o
+
+## Check that symbols are present:
+# RUN: llvm-nm --print-armap %t.lib | \
+# RUN:   FileCheck %s --check-prefix=CHECK-SYMBOLS -DPREFIX=%basename_t.tmp --match-full-lines
+
+# CHECK-SYMBOLS:      Archive map
+# CHECK-SYMBOLS-NEXT: _arm64 in [[PREFIX]]-universal.o
+# CHECK-SYMBOLS-EMPTY:
+
+## Check that the output archive is in Darwin format:
+# RUN: llvm-objdump --macho --archive-headers %t.lib | \
+# RUN:   FileCheck %s --check-prefix=FORMAT -DPREFIX=%basename_t.tmp -DARCHIVE=%t.lib
+
+# FORMAT:      Archive : [[ARCHIVE]]
+# FORMAT-NEXT: __.SYMDEF
+# FORMAT-NEXT: [[PREFIX]]-universal.o
+# FORMAT-NOT:  {{.}}
+
+## Passing both a universal file and an object file:
+# RUN: llvm-libtool-darwin -static -o %t.lib %t-universal.o %t-input1.o -arch_only x86_64
+# RUN: llvm-ar t %t.lib | \
+# RUN:   FileCheck %s --check-prefix=BOTH-NAMES --implicit-check-not={{.}} -DPREFIX=%basename_t.tmp
+# RUN: llvm-nm --print-armap %t.lib | \
+# RUN:   FileCheck %s --check-prefix=BOTH-SYMBOLS -DPREFIX=%basename_t.tmp --match-full-lines
+
+# BOTH-NAMES:      [[PREFIX]]-universal.o
+# BOTH-NAMES-NEXT: [[PREFIX]]-input1.o
+
+# BOTH-SYMBOLS:      Archive map
+# BOTH-SYMBOLS-NEXT: _x86_64 in [[PREFIX]]-universal.o
+# BOTH-SYMBOLS-NEXT: _symbol1 in [[PREFIX]]-input1.o
+# BOTH-SYMBOLS-EMPTY:
+
+## Passing both a universal file and an object file but filtering out the object file:
+# RUN: llvm-libtool-darwin -static -o %t.lib %t-universal.o %t-input1.o -arch_only arm64
+# RUN: llvm-ar t %t.lib | \
+# RUN:   FileCheck %s --check-prefix=CHECK-NAMES --implicit-check-not={{.}} -DPREFIX=%basename_t.tmp
+# RUN: llvm-nm --print-armap %t.lib | \
+# RUN:   FileCheck %s --check-prefix=CHECK-SYMBOLS -DPREFIX=%basename_t.tmp --match-full-lines
+
+## Universal file containing an archive:
+# RUN: rm -f %t.ar
+# RUN: llvm-ar cr %t.ar %t-input1.o %t-input2.o
+# RUN: llvm-lipo %t.ar -create -output %t-fat-with-archive.o
+# RUN: llvm-libtool-darwin -static -o %t.lib %t-fat-with-archive.o -arch_only x86_64
+# RUN: llvm-ar t %t.lib | \
+# RUN:   FileCheck %s --check-prefix=ARCHIVE-NAMES --implicit-check-not={{.}} -DPREFIX=%basename_t.tmp
+# RUN: llvm-nm --print-armap %t.lib | \
+# RUN:   FileCheck %s --check-prefix=ARCHIVE-SYMBOLS -DPREFIX=%basename_t.tmp --match-full-lines
+
+# ARCHIVE-NAMES:      [[PREFIX]]-input1.o
+# ARCHIVE-NAMES-NEXT: [[PREFIX]]-input2.o
+
+# ARCHIVE-SYMBOLS:      Archive map
+# ARCHIVE-SYMBOLS-NEXT: _symbol1 in [[PREFIX]]-input1.o
+# ARCHIVE-SYMBOLS-NEXT: _symbol2 in [[PREFIX]]-input2.o
+# ARCHIVE-SYMBOLS-EMPTY:
+
+## Allow arch_only to be specified more than once (pick the last one):
+# RUN: llvm-libtool-darwin -static -o %t.lib %t-universal.o -arch_only arm64 -arch_only x86_64
+# RUN: llvm-ar t %t.lib | \
+# RUN:   FileCheck %s --check-prefix=DOUBLE-NAMES --implicit-check-not={{.}} -DPREFIX=%basename_t.tmp
+# RUN: llvm-nm --print-armap %t.lib | \
+# RUN:   FileCheck %s --check-prefix=DOUBLE-SYMBOLS -DPREFIX=%basename_t.tmp --match-full-lines
+
+# DOUBLE-NAMES: [[PREFIX]]-universal.o
+
+# DOUBLE-SYMBOLS:      Archive map
+# DOUBLE-SYMBOLS-NEXT: _x86_64 in [[PREFIX]]-universal.o
+# DOUBLE-SYMBOLS-EMPTY:
+
+## Invalid architecture:
+# RUN: not llvm-libtool-darwin -static -o %t.lib %t-universal.o -arch_only arch101 2>&1 | \
+# RUN:   FileCheck %s --check-prefix=INVALID-ARCH
+
+# INVALID-ARCH: Invalid architecture: arch101
+# INVALID-ARCH-NEXT: Valid architecture names are
+
+## Empty architecture:
+# RUN: not llvm-libtool-darwin -static -o %t.lib %t-universal.o -arch_only "" 2>&1 | \
+# RUN:   FileCheck %s --check-prefix=EMPTY-ARCH
+
+# EMPTY-ARCH: Invalid architecture:
+# EMPTY-ARCH-NEXT: Valid architecture names are
+
+## Missing architecture:
+# RUN: not llvm-libtool-darwin -static -o %t.lib %t-universal.o -arch_only ppc 2>&1 | \
+# RUN:   FileCheck %s --check-prefix=MISSING-ARCH
+
+# MISSING-ARCH: error: no library created (no object files in input files matching -arch_only ppc)
+
+## arch_only missing argument:
+# RUN: not llvm-libtool-darwin -static -o %t.lib %t-universal.o -arch_only 2>&1 | \
+# RUN:   FileCheck %s --check-prefix=REQUIRE-ARCH
+
+# REQUIRE-ARCH: for the --arch_only option: requires a value!
+
+## x86_64-arm64-universal.yaml
+--- !fat-mach-o
+FatHeader:
+  magic:           0xCAFEBABE
+  nfat_arch:       2
+FatArchs:
+  - cputype:         0x01000007
+    cpusubtype:      0x00000003
+    offset:          0x0000000000001000
+    size:            352
+    align:           12
+  - cputype:         0x0100000C
+    cpusubtype:      0x00000000
+    offset:          0x0000000000004000
+    size:            384
+    align:           14
+Slices:
+  - !mach-o
+    FileHeader:
+      magic:           0xFEEDFACF
+      cputype:         0x01000007
+      cpusubtype:      0x00000003
+      filetype:        0x00000001
+      ncmds:           2
+      sizeofcmds:      176
+      flags:           0x00002000
+      reserved:        0x00000000
+    LoadCommands:
+      - cmd:             LC_SEGMENT_64
+        cmdsize:         152
+        segname:         ''
+        vmaddr:          0
+        vmsize:          15
+        fileoff:         312
+        filesize:        15
+        maxprot:         7
+        initprot:        7
+        nsects:          1
+        flags:           0
+        Sections:
+          - sectname:        __text
+            segname:         __TEXT
+            addr:            0x0000000000000000
+            size:            15
+            offset:          0x00000138
+            align:           4
+            reloff:          0x00000000
+            nreloc:          0
+            flags:           0x80000400
+            reserved1:       0x00000000
+            reserved2:       0x00000000
+            reserved3:       0x00000000
+            content:         554889E531C0C745FC000000005DC3
+      - cmd:             LC_SYMTAB
+        cmdsize:         24
+        symoff:          328
+        nsyms:           1
+        stroff:          344
+        strsize:         8
+    LinkEditData:
+      NameList:
+        - n_strx:          1
+          n_type:          0x0F
+          n_sect:          1
+          n_desc:          0
+          n_value:         0
+      StringTable:
+        - ''
+        - _x86_64
+        - ''
+  - !mach-o
+    FileHeader:
+      magic:           0xFEEDFACF
+      cputype:         0x0100000C
+      cpusubtype:      0x00000000
+      filetype:        0x00000001
+      ncmds:           2
+      sizeofcmds:      176
+      flags:           0x00002000
+      reserved:        0x00000000
+    LoadCommands:
+      - cmd:             LC_SEGMENT_64
+        cmdsize:         152
+        segname:         ''
+        vmaddr:          0
+        vmsize:          24
+        fileoff:         312
+        filesize:        24
+        maxprot:         7
+        initprot:        7
+        nsects:          1
+        flags:           0
+        Sections:
+          - sectname:        __text
+            segname:         __TEXT
+            addr:            0x0000000000000000
+            size:            24
+            offset:          0x00000138
+            align:           2
+            reloff:          0x00000000
+            nreloc:          0
+            flags:           0x80000400
+            reserved1:       0x00000000
+            reserved2:       0x00000000
+            reserved3:       0x00000000
+            content:         FF4300D1FF0F00B908008052E00308AAFF430091C0035FD6
+      - cmd:             LC_SYMTAB
+        cmdsize:         24
+        symoff:          336
+        nsyms:           2
+        stroff:          368
+        strsize:         16
+    LinkEditData:
+      NameList:
+        - n_strx:          7
+          n_type:          0x0E
+          n_sect:          1
+          n_desc:          0
+          n_value:         0
+        - n_strx:          1
+          n_type:          0x0F
+          n_sect:          1
+          n_desc:          0
+          n_value:         0
+      StringTable:
+        - ''
+        - _arm64
+        - ltmp0
+        - ''
+        - ''
+        - ''
+...

llvm/tools/llvm-libtool-darwin/CMakeLists.txt

 set(LLVM_LINK_COMPONENTS
   BinaryFormat
   Object
   Support
+  TextAPI
   )
 
 add_llvm_tool(llvm-libtool-darwin
   llvm-libtool-darwin.cpp
 )

llvm/tools/llvm-libtool-darwin/LLVMBuild.txt

 ;
 ;   http://llvm.org/docs/LLVMBuild.html
 ;
 ;===------------------------------------------------------------------------===;
 [component_0]
 type = Tool
 name = llvm-libtool-darwin
 parent = Tools
-required_libraries = BinaryFormat Object Support
+required_libraries = BinaryFormat Object Support TextAPI

llvm/tools/llvm-libtool-darwin/llvm-libtool-darwin.cpp

 //===-- llvm-libtool-darwin.cpp - a tool for creating libraries -----------===//
 //
 // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
 // See https://llvm.org/LICENSE.txt for license information.
 // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
 //
 //===----------------------------------------------------------------------===//
 //
 // A utility for creating static and dynamic libraries for Darwin.
 //
 //===----------------------------------------------------------------------===//
 
 #include "llvm/BinaryFormat/Magic.h"
 #include "llvm/Object/ArchiveWriter.h"
 #include "llvm/Object/MachO.h"
+#include "llvm/Object/MachOUniversal.h"
 #include "llvm/Object/ObjectFile.h"
 #include "llvm/Support/CommandLine.h"
 #include "llvm/Support/InitLLVM.h"
 #include "llvm/Support/LineIterator.h"
 #include "llvm/Support/WithColor.h"
+#include "llvm/TextAPI/MachO/Architecture.h"
 
 using namespace llvm;
 using namespace llvm::object;
 
 static bool Deterministic; // Updated by 'D' and 'U' modifiers.
 
 cl::OptionCategory LibtoolCategory("llvm-libtool-darwin Options");
 
 static cl::opt<std::string> OutputFile("o", cl::desc("Specify output filename"),
                                        cl::value_desc("filename"), cl::Required,
                                        cl::cat(LibtoolCategory));
 
 static cl::list<std::string> InputFiles(cl::Positional,
                                         cl::desc("<input files>"),
                                         cl::ZeroOrMore,
                                         cl::cat(LibtoolCategory));
 
+static cl::opt<std::string> ArchType(
+    "arch_only", cl::desc("Specify architecture type for output library"),
+    cl::value_desc("arch_type"), cl::ZeroOrMore, cl::cat(LibtoolCategory));
+
 enum class Operation { Static };
 
 static cl::opt<Operation> LibraryOperation(
     cl::desc("Library Type: "),
     cl::values(
         clEnumValN(Operation::Static, "static",
                    "Produce a statically linked library from the input files")),
     cl::Required, cl::cat(LibtoolCategory));
if (!DirName.empty())
       sys::path::append(Path, DirName, Line);
     else
       sys::path::append(Path, Line);
     InputFiles.push_back(Twine(Path).str());
   }
   return Error::success();
 }
 
-static Error verifyMachOObject(const NewArchiveMember &Member) {
+static Error validateArchitectureName(StringRef ArchitectureName) {
+  if (!MachOObjectFile::isValidArch(ArchitectureName)) {
+    std::string Buf;
+    raw_string_ostream OS(Buf);
+    for (StringRef arch : MachOObjectFile::getValidArchs())
+      OS << arch << " ";
+
+    return createStringError(
+        std::errc::invalid_argument,
+        "Invalid architecture: %s\nValid architecture names are %s",
+        ArchitectureName.str().c_str(), OS.str().c_str());
+  }
+  return Error::success();
+}
+
+// Check that File's architecture [FileCPUType, FileCPUSubtype]
+// matches the architecture specified under -arch_only flag.
+static bool acceptFileArch(uint32_t FileCPUType, uint32_t FileCPUSubtype) {
+  unsigned int ArchCPUType, ArchCPUSubtype;
+  std::tie(ArchCPUType, ArchCPUSubtype) = MachO::getCPUTypeFromArchitecture(
+      MachO::getArchitectureFromName(ArchType));
+
+  if (ArchCPUType != FileCPUType)
+    return false;
+
+  // Currently LLVM doesn't have the constant MachO::CPU_SUBTYPE_ARM64_V8.
+  // When the constant is added, following modification needs to be done for the
+  // case of MachO::CPU_TYPE_ARM64 below:
+  // if (ArchCPUType == MachO::CPU_TYPE_ARM64) {
+  //   if (ArchCPUSubtype == MachO::CPU_SUBTYPE_ARM64_ALL)
+  //     return FileCPUSubtype == MachO::CPU_SUBTYPE_ARM64_ALL ||
+  //            FileCPUSubtype == MachO::CPU_SUBTYPE_ARM64_V8;
+  //   else
+  //      return ArchCPUSubtype == FileCPUSubtype;
+  // }
+  switch (ArchCPUType) {
+  case MachO::CPU_TYPE_ARM:
+  case MachO::CPU_TYPE_ARM64:
+  case MachO::CPU_TYPE_ARM64_32:
+  case MachO::CPU_TYPE_X86_64:
+    return ArchCPUSubtype == FileCPUSubtype;
+  default:
+    return true;
+  }
+}
+
+static Error verifyAndAddMachOObject(std::vector<NewArchiveMember> &Members,
+                                     NewArchiveMember Member) {
   auto MBRef = Member.Buf->getMemBufferRef();
   Expected<std::unique_ptr<object::ObjectFile>> ObjOrErr =
       object::ObjectFile::createObjectFile(MBRef);
 
   // Throw error if not a valid object file.
   if (!ObjOrErr)
     return createFileError(Member.MemberName, ObjOrErr.takeError());
 
   // Throw error if not in Mach-O format.
   if (!isa<object::MachOObjectFile>(**ObjOrErr))
     return createStringError(std::errc::invalid_argument,
                              "'%s': format not supported",
                              Member.MemberName.data());
 
+  auto O = dyn_cast<MachOObjectFile>(ObjOrErr->get());
+  uint32_t FileCPUType, FileCPUSubtype;
+  std::tie(FileCPUType, FileCPUSubtype) = MachO::getCPUTypeFromArchitecture(
+      MachO::getArchitectureFromName(O->getArchTriple().getArchName()));
+
+  // If -arch_only is specified then skip this file if it doesn't match
+  // the architecture specified.
+  if (!ArchType.empty() && !acceptFileArch(FileCPUType, FileCPUSubtype)) {
+    return Error::success();
+  }
+
+  Members.push_back(std::move(Member));
   return Error::success();
 }
 
 static Error addChildMember(std::vector<NewArchiveMember> &Members,
                             const object::Archive::Child &M) {
   Expected<NewArchiveMember> NMOrErr =
       NewArchiveMember::getOldMember(M, Deterministic);
   if (!NMOrErr)
     return NMOrErr.takeError();
 
-  // Verify that Member is a Mach-O object file.
-  if (Error E = verifyMachOObject(*NMOrErr))
+  if (Error E = verifyAndAddMachOObject(Members, std::move(*NMOrErr)))
     return E;
 
-  Members.push_back(std::move(*NMOrErr));
   return Error::success();
 }
 
-static Error
-addMember(std::vector<NewArchiveMember> &Members, StringRef FileName,
-          std::vector<std::unique_ptr<MemoryBuffer>> &ArchiveBuffers) {
-  Expected<NewArchiveMember> NMOrErr =
-      NewArchiveMember::getFile(FileName, Deterministic);
-  if (!NMOrErr)
-    return createFileError(FileName, NMOrErr.takeError());
-
-  // For regular archives, use the basename of the object path for the member
-  // name.
-  NMOrErr->MemberName = sys::path::filename(NMOrErr->MemberName);
-
-  // Flatten archives.
-  if (identify_magic(NMOrErr->Buf->getBuffer()) == file_magic::archive) {
-    Expected<std::unique_ptr<Archive>> LibOrErr =
-        object::Archive::create(NMOrErr->Buf->getMemBufferRef());
-    if (!LibOrErr)
-      return createFileError(FileName, LibOrErr.takeError());
-    object::Archive &Lib = **LibOrErr;
-
-    Error Err = Error::success();
-    for (const object::Archive::Child &Child : Lib.children(Err))
-      if (Error E = addChildMember(Members, Child))
-        return createFileError(FileName, std::move(E));
-    if (Err)
-      return createFileError(FileName, std::move(Err));
+static Error processArchive(std::vector<NewArchiveMember> &Members,
+                            object::Archive &Lib, StringRef FileName) {
 
-    // Update vector ArchiveBuffers with the MemoryBuffers to transfer
-    // ownership.
-    ArchiveBuffers.push_back(std::move(NMOrErr->Buf));
-    return Error::success();
-  }
+  return Error::success();
+}
+
+static Error
+addArchiveMembers(std::vector<NewArchiveMember> &Members,
+                  std::vector<std::unique_ptr<MemoryBuffer>> &ArchiveBuffers,
+                  NewArchiveMember NM, StringRef FileName) {
+  Expected<std::unique_ptr<Archive>> LibOrErr =
+      object::Archive::create(NM.Buf->getMemBufferRef());
+  if (!LibOrErr)
+    return createFileError(FileName, LibOrErr.takeError());
+
+  if (Error E = processArchive(Members, **LibOrErr, FileName))
+    return E;
+
+  ArchiveBuffers.push_back(std::move(NM.Buf));
 
-  // Verify that Member is a Mach-O object file.
-  if (Error E = verifyMachOObject(*NMOrErr))
+static Error addUniversalMembers(
+    std::vector<NewArchiveMember> &Members,
+    std::vector<std::unique_ptr<MemoryBuffer>> &UniversalBuffers,
+    NewArchiveMember NM, StringRef FileName) {
+
+  Expected<std::unique_ptr<MachOUniversalBinary>> BinaryOrErr =
+      MachOUniversalBinary::create(NM.Buf->getMemBufferRef());
+  if (!BinaryOrErr)
+    return createFileError(FileName, BinaryOrErr.takeError());
+
+  auto UO = BinaryOrErr->get();
+  for (const auto &O : UO->objects()) {
+
+    Expected<std::unique_ptr<MachOObjectFile>> MachOObjOrErr =
+        O.getAsObjectFile();
+    if (MachOObjOrErr) {
+      NewArchiveMember NewMember =
+          NewArchiveMember(MachOObjOrErr->get()->getMemoryBufferRef());
+      NewMember.MemberName = sys::path::filename(NewMember.MemberName);
+
+      if (Error E = verifyAndAddMachOObject(Members, std::move(NewMember)))
+        return E;
+      continue;
+    }
+
+    Expected<std::unique_ptr<Archive>> ArchiveOrError = O.getAsArchive();
+    if (ArchiveOrError) {
+      consumeError(MachOObjOrErr.takeError());
+      if (Error E = processArchive(Members, **ArchiveOrError, FileName))
         return E;
+      continue;
+    }
+
+    Error CombinedError =
+        joinErrors(ArchiveOrError.takeError(), MachOObjOrErr.takeError());
+    return createFileError(FileName, std::move(CombinedError));
+  }
 
-  Members.push_back(std::move(*NMOrErr));
+  // Update vector UniversalBuffers with the MemoryBuffers to transfer
+  // ownership.
+  UniversalBuffers.push_back(std::move(NM.Buf));
+  return Error::success();
+}
+
+static Error addMember(std::vector<NewArchiveMember> &Members,
+                       std::vector<std::unique_ptr<MemoryBuffer>> &FileBuffers,
+                       StringRef FileName) {
+  Expected<NewArchiveMember> NMOrErr =
+      NewArchiveMember::getFile(FileName, Deterministic);
+  if (!NMOrErr)
+    return createFileError(FileName, NMOrErr.takeError());
+
+  // For regular archives, use the basename of the object path for the member
+  // name.
+  NMOrErr->MemberName = sys::path::filename(NMOrErr->MemberName);
+  file_magic Magic = identify_magic(NMOrErr->Buf->getBuffer());
+
+  // Flatten archives.
+  if (Magic == file_magic::archive)
+    return addArchiveMembers(Members, FileBuffers, std::move(*NMOrErr),
+                             FileName);
+
+  // Flatten universal files.
+  if (Magic == file_magic::macho_universal_binary)
+    return addUniversalMembers(Members, FileBuffers, std::move(*NMOrErr),
+                               FileName);
+
+  if (Error E = verifyAndAddMachOObject(Members, std::move(*NMOrErr)))
+    return E;
   return Error::success();
 }
 
 static Error createStaticLibrary() {
   std::vector<NewArchiveMember> NewMembers;
-  std::vector<std::unique_ptr<MemoryBuffer>> ArchiveBuffers;
-  for (StringRef Member : InputFiles)
-    if (Error E = addMember(NewMembers, Member, ArchiveBuffers))
+  std::vector<std::unique_ptr<MemoryBuffer>> FileBuffers;
+  for (StringRef FileName : InputFiles)
+    if (Error E = addMember(NewMembers, FileBuffers, FileName))
       return E;
 
+  if (NewMembers.empty() && !ArchType.empty())
+    return createStringError(std::errc::invalid_argument,
+                             "no library created (no object files in input "
+                             "files matching -arch_only %s)",
+                             ArchType.c_str());
+
   if (Error E = writeArchive(OutputFile, NewMembers,
                              /*WriteSymtab=*/true,
                              /*Kind=*/object::Archive::K_DARWIN, Deterministic,
                              /*Thin=*/false))
     return E;
   return Error::success();
 }
 
 static Error parseCommandLine(int Argc, char **Argv) {
   cl::ParseCommandLineOptions(Argc, Argv, "llvm-libtool-darwin\n");
   if (DeterministicOption && NonDeterministicOption)
     return createStringError(std::errc::invalid_argument,
                              "cannot specify both -D and -U flags");
   else
     Deterministic = NonDeterministicOption ? false : true;
 
   if (!FileList.empty())
     if (Error E = processFileList())
       return E;
 
   if (InputFiles.empty())
     return createStringError(std::errc::invalid_argument,
                              "no input files specified");
 
+  if (ArchType.getNumOccurrences())
+    if (Error E = validateArchitectureName(ArchType))
+      return E;
+
   return Error::success();
 }
 
 int main(int Argc, char **Argv) {
   InitLLVM X(Argc, Argv);
   cl::HideUnrelatedOptions({&LibtoolCategory, &ColorCategory});
   if (Error E = parseCommandLine(Argc, Argv)) {
     WithColor::defaultErrorHandler(std::move(E));