[lld] [lld-macho] Support archives without index (PR #132942)

[Leonard Grey via llvm-commits]

================
@@ -0,0 +1,43 @@
+# REQUIRES: x86
+
+# RUN: rm -rf %t; split-file %s %t
+# RUN: llvm-mc -filetype=obj -triple=x86_64-apple-macos -o %t/main.o %t/main.s
+# RUN: llvm-mc -filetype=obj -triple=x86_64-apple-macos -o %t/lib.o %t/lib.s
+# RUN: llvm-mc -filetype=obj -triple=x86_64-apple-macos -o %t/lib2.o %t/lib2.s
+# RUN: llvm-ar crST %t/lib.a %t/lib.o %t/lib2.o
+# RUN: %lld %t/main.o %t/lib.a -o %t/out
+# RUN: llvm-nm %t/out | FileCheck %s
+
+# CHECK-NOT: T _bar
+# CHECK:     T _foo
+
+## Test that every kind of eager load mechanism still works.
+# RUN: %lld %t/main.o %t/lib.a -all_load -o %t/all_load
+# RUN: llvm-nm %t/all_load | FileCheck %s --check-prefix FORCED-LOAD
+# RUN: %lld %t/main.o -force_load %t/lib.a -o %t/force_load
+# RUN: llvm-nm %t/force_load | FileCheck %s --check-prefix FORCED-LOAD
+# RUN: %lld %t/main.o %t/lib.a -ObjC -o %t/objc
----------------
speednoisemovement wrote:

Relatively cheap if we do it `ObjFile::parseLazy`. The annoying part would be going back and turning every lazy we just created into a defined partway through the iteration. I'm still confused on if this can actually happen though.

https://github.com/llvm/llvm-project/pull/132942