[ELF] - Improve the test cases for notes sections.

[grimar]

Summary

This is a follow up for D46623.

It cleans the existent test case and adds one more to
test the notes better.

[emaste]

note-noalloc2.s is close to my original test, I think that's reasonable.

For the note-noalloc.s change we've lost the check that the non-alloc note still has a section. I think we should keep that.

test/ELF/note-noalloc.s

2

aww

[grimar]

In D46690#1094388, @emaste wrote:

note-noalloc2.s is close to my original test, I think that's reasonable.

For the note-noalloc.s change we've lost the check that the non-alloc note still has a section. I think we should keep that.

See.

I am checking that PT_NOTE has a 16 bytes size (so that is including only "a" section).
I am not sure why do we want to check that we have a output section for non-a. Currently, we can have the following behavior of LLD:

1. If we have note.x and note y, where first is "a" and second is non-a. We will have 2 sections in the output.

2. we can have 2 .note sections, one of which is "a" and the second is non-a. LLD will combine them into single .note by name, I think.

And we will emit the PT_NOTE for both.

Your change changed how PT_NOTES are emitted., it did not change how we emit the output sections.
So I would not add tests for output notes probably as 1+2 shows we have no special rules for notes right now I think.

[grimar]

After the second look, I think it is ok to add check here that we still keep non-a section and keep its SHT_NOTE flag for safety though. I'll update the patch.

[grimar]

• Addressed review comments.

[emaste]

In D46690#1094394, @grimar wrote:

After the second look, I think it is ok to add check here that we still keep non-a section and keep its SHT_NOTE flag for safety though. I'll update the patch.

Yes, a safety check that we don't accidentally break this in the future was my main concern. As @ruiu said in the other review it seems the ghc folks are using the non-alloc note effectively as a .comment section and I'd like to confirm we maintain that functionality.

test/ELF/note-noalloc.s

17

its

[grimar]

Thanks for the review, Ed!

test/ELF/note-noalloc.s

 // REQUIRES: x86
-// RUN: llvm-mc -filetype=obj -triple=x86_64-unknown-freebsd %s -o %t.o
+// RUN: llvm-mc -filetype=obj -triple=x86_64-pc-linux %s -o %t.o
 // RUN: ld.lld %t.o -o %t -shared
 // RUN: llvm-readobj -program-headers -sections %t | FileCheck %s
 
-// PR37361: A note without SHF_ALLOC should not create a PT_NOTE program
-// header (but should have a SHT_NOTE section).
+// PR37361: A note without SHF_ALLOC should not be included into a PT_NOTE program header.
 
-// CHECK: Name: .note.tag
-// CHECK: Type: SHT_NOTE
-// CHECK: Name: .debug.ghc-link-info
-// CHECK: Type: SHT_NOTE
-// CHECK-NOT: Type: SHT_NOTE
+// CHECK:      Section {
+// CHECK:        Index:
+// CHECK:        Name: .note.a
+// CHECK-NEXT:   Type: SHT_NOTE
+// CHECK-NEXT:   Flags [
+// CHECK-NEXT:     SHF_ALLOC
+// CHECK-NEXT:   ]
+// CHECK-NEXT:   Address: 0x[[ADDR:.*]]
 
-// CHECK: Type: PT_NOTE
-// CHECK-NOT: Type: PT_NOTE
+// Check we still emit the non-alloc SHT_NOTE section and keep it's type.
+
+// CHECK:        Name: .note.b
+// CHECK-NEXT:   Type: SHT_NOTE
+// CHECK-NEXT:   Flags [
+// CHECK-NEXT:   ]
+
+// CHECK:      ProgramHeader {
+// CHECK:        Type: PT_NOTE
+// CHECK-NEXT:   Offset:
+// CHECK-NEXT:   VirtualAddress: 0x[[ADDR]]
+// CHECK-NEXT:   PhysicalAddress: 0x245
+// CHECK-NEXT:   FileSize: 16
+// CHECK-NEXT:   MemSize: 16
+// CHECK-NOT:  PT_NOTE
 
-        .section        .note.tag,"a",@note
-        .quad 1234
+.section        .note.a,"a",@note
+.quad 1
+.quad 2
 
-        .section        .debug.ghc-link-info,"",@note
-        .quad 5678
+.section        .note.b,"",@note
+.quad 3

test/ELF/note-noalloc2.s

+# REQUIRES: x86
+# RUN: llvm-mc -filetype=obj -triple=x86_64-pc-linux %s -o %t.o
+# RUN: ld.lld %t.o -o %t
+# RUN: llvm-readobj -program-headers %t | FileCheck %s
+
+## Check we do not create a PT_NOTE segment for non-allocatable note section.
+
+# CHECK-NOT:  PT_NOTE
+
+.section  .note,"",@note
+.quad 0