Generalize DWARF tests that expect DW_FORM_strp

[Jake-Egan]

Summary

The modified tests were failing on AIX because DWARF on AIX uses inline strings by default, but the tests check for DW_FORM_strp.

This patch removes DW_FORM_strp so both forms will pass the check.

[shchenz]

Maybe we could split this patch into two:
1: for the space and tab
2: for -dwarf-inlined-strings=Disable

The second one looks reasonable to me.
But for the first one, do we know why on AIX it generates a tab instead of space? Is it caused by llvm-dwarfdump on AIX? If so, maybe we need to change llvm-dwarfdump to make it generate space also on AIX?

[dblaikie]

In D112286#3079832, @shchenz wrote:

Maybe we could split this patch into two:
1: for the space and tab
2: for -dwarf-inlined-strings=Disable

The second one looks reasonable to me.
But for the first one, do we know why on AIX it generates a tab instead of space? Is it caused by llvm-dwarfdump on AIX? If so, maybe we need to change llvm-dwarfdump to make it generate space also on AIX?

^ agreed. Would be good to understand why llvm-dwarfdump is doing anything different in terms of whitespace on AIX... I can't think of any reason that should be the case.

[Jake-Egan]

Here is an example for the differing output with file llvm/test/DebugInfo/Generic/varargs.ll

On Linux:

DW_AT_name [DW_FORM_strp]^I( .debug_str[0x00000065] = "a")$

On AIX:

DW_AT_name [DW_FORM_string]^I("a")$

I will create a new patch for the white space issue and convert this patch to add the -dwarf-inlined-strings=Disable option.

[Jake-Egan]

Update patch for -dwarf-inlined-strings=Disable option.

[shchenz]

LGTM. Thanks

[dblaikie]

I think probably generalizing the tests to not care about the string encodings is probably the right thing, rather than forcing indirect strings?

I think these tests are overly specific - they don't look like they should care about the string encoding (& Probably an artifact of older techniques in writing these debug info tests - improvements to the dumper to print strings inline, skip forms, etc)

[dblaikie]

In D112286#3082790, @Jake-Egan wrote:

Here is an example for the differing output with file llvm/test/DebugInfo/Generic/varargs.ll

On Linux:

DW_AT_name [DW_FORM_strp]^I( .debug_str[0x00000065] = "a")$

On AIX:

DW_AT_name [DW_FORM_string]^I("a")$

I will create a new patch for the white space issue and convert this patch to add the -dwarf-inlined-strings=Disable option.

Thanks for showing the difference - I think these can probably most often be tested with CHECK: DW_AT_name {{.*}}"a" is probably fine - or if the tests don't need -v output (doesn't dump forms, etc), it'll probably be simpler, may even be exactly the same output between strp and str encodings.

[Jake-Egan]

In D112286#3085048, @dblaikie wrote:

I think probably generalizing the tests to not care about the string encodings is probably the right thing, rather than forcing indirect strings?

I think these tests are overly specific - they don't look like they should care about the string encoding (& Probably an artifact of older techniques in writing these debug info tests - improvements to the dumper to print strings inline, skip forms, etc)

I agree generalizing the test is probably the better option if DW_FORM_strp isn't necessary for the purpose of the test.

@shchenz I will update the patch if you have no objection.

[shchenz]

In D112286#3086128, @Jake-Egan wrote:

In D112286#3085048, @dblaikie wrote:

I think probably generalizing the tests to not care about the string encodings is probably the right thing, rather than forcing indirect strings?

I think these tests are overly specific - they don't look like they should care about the string encoding (& Probably an artifact of older techniques in writing these debug info tests - improvements to the dumper to print strings inline, skip forms, etc)

I agree generalizing the test is probably the better option if DW_FORM_strp isn't necessary for the purpose of the test.

@shchenz I will update the patch if you have no objection.

Sounds good to me

[Jake-Egan]

Update patch to generalize the tests instead.

[dblaikie]

llvm/test/DebugInfo/Generic/containing-type-extension.ll

8

I'd drop the "= " here (& elsewhere in the patch), like it's done elsewhere

[Jake-Egan]

llvm/test/DebugInfo/Generic/containing-type-extension.ll

8

I was going to do that in a separate patch for the whitespace. I think it's more related to the whitespace issue do you think? The patch would remove both = and spaces.

[dblaikie]

llvm/test/DebugInfo/Generic/containing-type-extension.ll

8

Oh, sure, either way

llvm/test/DebugInfo/Generic/containing-type-extension.ll

 ; RUN: %llc_dwarf -O0 -filetype=obj < %s > %t
 ; RUN: llvm-dwarfdump -v -debug-info %t | FileCheck %s
 
 ; Check that any type can have a vtable holder.
 ; CHECK: [[SP:.*]]: DW_TAG_structure_type
 ; CHECK-NOT: TAG
 ; CHECK: DW_AT_containing_type [DW_FORM_ref4]
-; CHECK: DW_AT_name [DW_FORM_strp] {{.*}}= "vtable")
+; CHECK: DW_AT_name {{.*}}= "vtable")
 
 ; The code doesn't actually matter.
 define i32 @main() #0 !dbg !4 {
 entry:
   %retval = alloca i32, align 4
   store i32 0, i32* %retval
   ret i32 0, !dbg !10
 }

llvm/test/DebugInfo/Generic/debug-label-inline.ll

 ; RUN: llc -O0 -filetype=obj -o - %s | llvm-dwarfdump -v - | FileCheck %s
 ;
 ; Bug 47129
 ; XFAIL: sparc
 ;
 ; CHECK: .debug_info contents:
 ; CHECK: [[LABEL_ORIGIN:0x[0-9a-zA-Z]+]]:{{ *}}DW_TAG_label
-; CHECK-NEXT: DW_AT_name [DW_FORM_strp] {{.*}}"top"
+; CHECK-NEXT: DW_AT_name {{.*}}"top"
 ; CHECK-NEXT: DW_AT_decl_file [DW_FORM_data1] {{.*}}debug-label-inline.c
 ; CHECK-NEXT: DW_AT_decl_line [DW_FORM_data1] {{.*}}8
 ; CHECK: DW_TAG_label
 ; CHECK-NEXT: DW_AT_abstract_origin [DW_FORM_ref4] {{.*}}{[[LABEL_ORIGIN]]} "top"
 ; CHECK-NEXT: DW_AT_low_pc [DW_FORM_addr] {{.*}}{{0x[0-9a-f]+}}
 
 source_filename = "debug-label-inline.c"
 

llvm/test/DebugInfo/Generic/debug-label.ll

 ; RUN: llc -O0 -filetype=obj -o - %s | llvm-dwarfdump -v - | FileCheck %s
 ;
 ; CHECK: .debug_info contents:
 ; CHECK: DW_TAG_label
-; CHECK-NEXT: DW_AT_name [DW_FORM_strp] {{.*}}"top"
+; CHECK-NEXT: DW_AT_name {{.*}}"top"
 ; CHECK-NEXT: DW_AT_decl_file [DW_FORM_data1] {{.*}}debug-label.c
 ; CHECK-NEXT: DW_AT_decl_line [DW_FORM_data1] {{.*}}4
 ; CHECK-NEXT: DW_AT_low_pc [DW_FORM_addr] {{.*}}{{0x[0-9a-f]+}}
 ; CHECK: DW_TAG_label
-; CHECK-NEXT: DW_AT_name [DW_FORM_strp] {{.*}}"done"
+; CHECK-NEXT: DW_AT_name {{.*}}"done"
 ; CHECK-NEXT: DW_AT_decl_file [DW_FORM_data1] {{.*}}debug-label.c
 ; CHECK-NEXT: DW_AT_decl_line [DW_FORM_data1] {{.*}}7
 ; CHECK-NEXT: DW_AT_low_pc [DW_FORM_addr] {{.*}}{{0x[0-9a-f]+}}
-; CHECK-NOT: DW_AT_name [DW_FORM_strp] {{.*}}"top"
+; CHECK-NOT: DW_AT_name {{.*}}"top"
 ;
 ; RUN: llc -O0 -o - %s | FileCheck %s -check-prefix=ASM
 ;
 ; ASM: [[TOP_LOW_PC:[.0-9a-zA-Z]+]]:{{[[:space:]].*}}DEBUG_LABEL: foo:top
 ; ASM: [[DONE_LOW_PC:[.0-9a-zA-Z]+]]:{{[[:space:]].*}}DEBUG_LABEL: foo:done
 ; ASM-LABEL: debug_info
 ; ASM: DW_TAG_label
 ; ASM-NEXT: DW_AT_name

llvm/test/DebugInfo/Generic/tu-composite.ll

 ; RUN: %llc_dwarf -filetype=obj -O0 < %s > %t
 ; RUN: llvm-dwarfdump -v -debug-info %t | FileCheck %s
 ; CHECK: [[TYPE:.*]]: DW_TAG_structure_type
 ; Make sure we correctly handle containing type of a struct being a type identifier.
 ; CHECK-NEXT: DW_AT_containing_type [DW_FORM_ref4]       (cu + {{.*}} => {[[TYPE]]})
-; CHECK-NEXT: DW_AT_name [DW_FORM_strp] {{.*}}= "C")
+; CHECK-NEXT: DW_AT_name {{.*}}= "C")
 
 ; Make sure we correctly handle context of a subprogram being a type identifier.
 ; CHECK: [[SP:.*]]: DW_TAG_subprogram
-; CHECK: DW_AT_name [DW_FORM_strp] {{.*}}= "foo")
+; CHECK: DW_AT_name {{.*}}= "foo")
 ; Make sure we correctly handle containing type of a subprogram being a type identifier.
 ; CHECK: DW_AT_containing_type [DW_FORM_ref4]       (cu + {{.*}} => {[[TYPE]]})
 ; CHECK: DW_TAG_formal_parameter
 ; CHECK: NULL
 ; CHECK: NULL
 
 ; CHECK: [[TYPE2:.*]]: DW_TAG_structure_type
-; CHECK: DW_AT_name [DW_FORM_strp] {{.*}}= "bar")
+; CHECK: DW_AT_name {{.*}}= "bar")
 ; CHECK: DW_TAG_structure_type
-; CHECK: DW_AT_name [DW_FORM_strp] {{.*}}= "D")
+; CHECK: DW_AT_name {{.*}}= "D")
 ; CHECK: DW_TAG_member
-; CHECK: DW_AT_name [DW_FORM_strp] {{.*}}= "a") 
+; CHECK: DW_AT_name {{.*}}= "a") 
 ; Make sure we correctly handle context of a struct being a type identifier.
 ; CHECK: DW_TAG_structure_type
-; CHECK-NEXT: DW_AT_name [DW_FORM_strp] {{.*}}= "Nested")
+; CHECK-NEXT: DW_AT_name {{.*}}= "Nested")
 ; CHECK: DW_TAG_structure_type
-; CHECK-NEXT: DW_AT_name [DW_FORM_strp] {{.*}}= "Nested2")
+; CHECK-NEXT: DW_AT_name {{.*}}= "Nested2")
 ; CHECK-NEXT: DW_AT_declaration [DW_FORM_flag]      (0x01)
 ; CHECK: DW_TAG_structure_type
-; CHECK-NEXT: DW_AT_name [DW_FORM_strp] {{.*}}= "virt<bar>")
+; CHECK-NEXT: DW_AT_name {{.*}}= "virt<bar>")
 ; Make sure we correctly handle type of a template_type being a type identifier.
 ; CHECK: DW_TAG_template_type_parameter
 ; CHECK-NEXT: DW_AT_type [DW_FORM_ref4] (cu + {{.*}} => {[[TYPE2]]}
-; CHECK-NEXT: DW_AT_name [DW_FORM_strp] {{.*}}= "T")
+; CHECK-NEXT: DW_AT_name {{.*}}= "T")
 ; Make sure we correctly handle derived-from of a typedef being a type identifier.
 ; CHECK: DW_TAG_typedef
 ; CHECK-NEXT: DW_AT_type [DW_FORM_ref4] (cu + {{.*}} => {[[TYPE2]]}
-; CHECK: DW_AT_name [DW_FORM_strp] {{.*}}= "baz2")
+; CHECK: DW_AT_name {{.*}}= "baz2")
 ; Make sure we correctly handle derived-from of a pointer type being a type identifier.
 ; CHECK: DW_TAG_pointer_type
 ; CHECK: DW_AT_type [DW_FORM_ref4] (cu + {{.*}} => {[[TYPE]]}
 ; CHECK: DW_TAG_typedef
 ; CHECK-NEXT: DW_AT_type [DW_FORM_ref4] (cu + {{.*}} => {[[TYPE2]]}
-; CHECK: DW_AT_name [DW_FORM_strp] {{.*}}= "baz")
+; CHECK: DW_AT_name {{.*}}= "baz")
 ; Make sure we correctly handle derived-from of an array type being a type identifier.
 ; CHECK: DW_TAG_array_type
 ; CHECK-NEXT: DW_AT_type [DW_FORM_ref4] (cu + {{.*}} => {[[TYPE2]]}
 ; IR generated from clang -g with the following source:
 ; struct C {
 ;   virtual void foo();
 ; };
 ; void C::foo() {