fix: DebugInfo alignment and offset for aggregate types (#1329)

mhasel · volsa · web-flow · commit 9e55aed6ba15 · 2024-10-07T21:15:26.000+02:00
Arrays and Strings now use the inner type to calculate running-offset and alignment. Offsets and sizes now align with a comparable `C` program compiled with clang: ```c #include <stdbool.h> typedef struct { short a; bool b; char s[81]; long long ll; short arr[11]; short j; } mainProg_t; int main() { return 0; } ``` generated DI: ```llvm !14 = !DIDerivedType(tag: DW_TAG_typedef, name: "mainProg_t", file: !1, line: 10, baseType: !15) !15 = distinct !DICompositeType(tag: DW_TAG_structure_type, file: !1, line: 3, size: 960, elements: !16) !16 = !{!17, !19, !21, !26, !28, !32} !17 = !DIDerivedType(tag: DW_TAG_member, name: "a", scope: !15, file: !1, line: 4, baseType: !18, size: 16) !18 = !DIBasicType(name: "short", size: 16, encoding: DW_ATE_signed) !19 = !DIDerivedType(tag: DW_TAG_member, name: "b", scope: !15, file: !1, line: 5, baseType: !20, size: 8, offset: 16) !20 = !DIBasicType(name: "_Bool", size: 8, encoding: DW_ATE_boolean) !21 = !DIDerivedType(tag: DW_TAG_member, name: "s", scope: !15, file: !1, line: 6, baseType: !22, size: 648, offset: 24) !22 = !DICompositeType(tag: DW_TAG_array_type, baseType: !23, size: 648, elements: !24) !23 = !DIBasicType(name: "char", size: 8, encoding: DW_ATE_signed_char) !24 = !{!25} !25 = !DISubrange(count: 81) !26 = !DIDerivedType(tag: DW_TAG_member, name: "ll", scope: !15, file: !1, line: 7, baseType: !27, size: 64, offset: 704) !27 = !DIBasicType(name: "long long", size: 64, encoding: DW_ATE_signed) !28 = !DIDerivedType(tag: DW_TAG_member, name: "arr", scope: !15, file: !1, line: 8, baseType: !29, size: 176, offset: 768) !29 = !DICompositeType(tag: DW_TAG_array_type, baseType: !18, size: 176, elements: !30) !30 = !{!31} !31 = !DISubrange(count: 11) !32 = !DIDerivedType(tag: DW_TAG_member, name: "j", scope: !15, file: !1, line: 9, baseType: !18, size: 16, offset: 944) ``` ```llvm PROGRAM prog VAR a : INT := 420; b : BOOL := TRUE; s : STRING := 'Hello, world!'; ll : LINT := 69; arr : ARRAY[0..10] OF INT; j : INT; END_VAR END_PROGRAM FUNCTION main : DINT END_FUNCTION ``` generated DI: ```llvm !3 = !DICompositeType(tag: DW_TAG_structure_type, name: "prog", scope: !2, file: !2, line: 11, size: 960, align: 64, flags: DIFlagPublic, elements: !4, identifier: "prog") !4 = !{!5, !7, !9, !14, !16, !20} !5 = !DIDerivedType(tag: DW_TAG_member, name: "a", scope: !2, file: !2, line: 13, baseType: !6, size: 16, align: 16, flags: DIFlagPublic) !6 = !DIBasicType(name: "INT", size: 16, encoding: DW_ATE_signed, flags: DIFlagPublic) !7 = !DIDerivedType(tag: DW_TAG_member, name: "b", scope: !2, file: !2, line: 14, baseType: !8, size: 8, align: 8, offset: 16, flags: DIFlagPublic) !8 = !DIBasicType(name: "BOOL", size: 8, encoding: DW_ATE_boolean, flags: DIFlagPublic) !9 = !DIDerivedType(tag: DW_TAG_member, name: "s", scope: !2, file: !2, line: 15, baseType: !10, size: 648, align: 8, offset: 24, flags: DIFlagPublic) !10 = !DICompositeType(tag: DW_TAG_array_type, baseType: !11, size: 648, align: 8, elements: !12) !11 = !DIBasicType(name: "char", size: 8, encoding: DW_ATE_UTF, flags: DIFlagPublic) !12 = !{!13} !13 = !DISubrange(count: 81, lowerBound: 0) !14 = !DIDerivedType(tag: DW_TAG_member, name: "ll", scope: !2, file: !2, line: 16, baseType: !15, size: 64, align: 64, offset: 704, flags: DIFlagPublic) !15 = !DIBasicType(name: "LINT", size: 64, encoding: DW_ATE_signed, flags: DIFlagPublic) !16 = !DIDerivedType(tag: DW_TAG_member, name: "arr", scope: !2, file: !2, line: 17, baseType: !17, size: 176, align: 16, offset: 768, flags: DIFlagPublic) !17 = !DICompositeType(tag: DW_TAG_array_type, baseType: !6, size: 176, align: 16, elements: !18) !18 = !{!19} !19 = !DISubrange(count: 11, lowerBound: 0) !20 = !DIDerivedType(tag: DW_TAG_member, name: "j", scope: !2, file: !2, line: 18, baseType: !6, size: 16, align: 16, offset: 944, flags: DIFlagPublic) ``` Running an example program with the above struct with initialized values now also prints the expected values when running `info locals` in gdb. ![image](https://github.com/user-attachments/assets/ae328259-1c4f-4c77-97b1-78963e69eafc) --------- Co-authored-by: Volkan Sagcan <Volkan.SAGCAN@bachmann.info>
diff --git a/src/codegen/tests/debug_tests/snapshots/rusty__codegen__tests__debug_tests__expression_debugging__aggregate_return_value_variable_in_function.snap b/src/codegen/tests/debug_tests/snapshots/rusty__codegen__tests__debug_tests__expression_debugging__aggregate_return_value_variable_in_function.snap
@@ -49,7 +49,7 @@ attributes #2 = { argmemonly nofree nounwind willreturn }
 !9 = !DILocation(line: 3, column: 12, scope: !4)
 !10 = !DILocalVariable(name: "myFunc", scope: !4, file: !5, line: 2, type: !11)
 !11 = !DIDerivedType(tag: DW_TAG_pointer_type, name: "__ref_to_STRING", baseType: !12, size: 64, align: 64, dwarfAddressSpace: 1)
-!12 = !DICompositeType(tag: DW_TAG_array_type, baseType: !13, size: 80, align: 64, elements: !14)
+!12 = !DICompositeType(tag: DW_TAG_array_type, baseType: !13, size: 648, align: 8, elements: !14)
 !13 = !DIBasicType(name: "char", size: 8, encoding: DW_ATE_UTF, flags: DIFlagPublic)
 !14 = !{!15}
 !15 = !DISubrange(count: 81, lowerBound: 0)
diff --git a/src/codegen/tests/debug_tests/snapshots/rusty__codegen__tests__debug_tests__expression_debugging__array_size_correctly_set_in_dwarf_info.snap b/src/codegen/tests/debug_tests/snapshots/rusty__codegen__tests__debug_tests__expression_debugging__array_size_correctly_set_in_dwarf_info.snap
@@ -40,8 +40,8 @@ attributes #1 = { argmemonly nofree nounwind willreturn writeonly }
 !7 = !{null}
 !8 = !{}
 !9 = !DILocation(line: 6, column: 8, scope: !4)
-!10 = !DILocalVariable(name: "a", scope: !4, file: !5, line: 4, type: !11, align: 64)
-!11 = !DICompositeType(tag: DW_TAG_array_type, baseType: !12, size: 2048, align: 64, elements: !13)
+!10 = !DILocalVariable(name: "a", scope: !4, file: !5, line: 4, type: !11, align: 32)
+!11 = !DICompositeType(tag: DW_TAG_array_type, baseType: !12, size: 2048, align: 32, elements: !13)
 !12 = !DIBasicType(name: "DINT", size: 32, encoding: DW_ATE_signed, flags: DIFlagPublic)
 !13 = !{!14}
 !14 = !DISubrange(count: 64, lowerBound: 1)
diff --git a/src/codegen/tests/debug_tests/snapshots/rusty__codegen__tests__debug_tests__expression_debugging__string_size_correctly_set_in_dwarf_info.snap b/src/codegen/tests/debug_tests/snapshots/rusty__codegen__tests__debug_tests__expression_debugging__string_size_correctly_set_in_dwarf_info.snap
@@ -13,7 +13,7 @@ source_filename = "<internal>"
 !0 = !DIGlobalVariableExpression(var: !1, expr: !DIExpression())
 !1 = distinct !DIGlobalVariable(name: "a", scope: !2, file: !2, line: 3, type: !3, isLocal: false, isDefinition: true)
 !2 = !DIFile(filename: "<internal>", directory: "")
-!3 = !DICompositeType(tag: DW_TAG_array_type, baseType: !4, size: 64, align: 64, elements: !5)
+!3 = !DICompositeType(tag: DW_TAG_array_type, baseType: !4, size: 520, align: 8, elements: !5)
 !4 = !DIBasicType(name: "char", size: 8, encoding: DW_ATE_UTF, flags: DIFlagPublic)
 !5 = !{!6}
 !6 = !DISubrange(count: 65, lowerBound: 0)
diff --git a/src/codegen/tests/snapshots/rusty__codegen__tests__debug_tests__global_var_array_added_to_debug_info.snap b/src/codegen/tests/snapshots/rusty__codegen__tests__debug_tests__global_var_array_added_to_debug_info.snap
@@ -15,19 +15,19 @@ source_filename = "<internal>"
 !0 = !DIGlobalVariableExpression(var: !1, expr: !DIExpression())
 !1 = distinct !DIGlobalVariable(name: "a", scope: !2, file: !2, line: 3, type: !3, isLocal: false, isDefinition: true)
 !2 = !DIFile(filename: "<internal>", directory: "")
-!3 = !DICompositeType(tag: DW_TAG_array_type, baseType: !4, size: 352, align: 64, elements: !5)
+!3 = !DICompositeType(tag: DW_TAG_array_type, baseType: !4, size: 352, align: 32, elements: !5)
 !4 = !DIBasicType(name: "DINT", size: 32, encoding: DW_ATE_signed, flags: DIFlagPublic)
 !5 = !{!6}
 !6 = !DISubrange(count: 11, lowerBound: 0)
 !7 = !DIGlobalVariableExpression(var: !8, expr: !DIExpression())
 !8 = distinct !DIGlobalVariable(name: "b", scope: !2, file: !2, line: 4, type: !9, isLocal: false, isDefinition: true)
-!9 = !DICompositeType(tag: DW_TAG_array_type, baseType: !4, size: 3520, align: 64, elements: !10)
+!9 = !DICompositeType(tag: DW_TAG_array_type, baseType: !4, size: 3520, align: 32, elements: !10)
 !10 = !{!6, !11}
 !11 = !DISubrange(count: 10, lowerBound: 11)
 !12 = !DIGlobalVariableExpression(var: !13, expr: !DIExpression())
 !13 = distinct !DIGlobalVariable(name: "c", scope: !2, file: !2, line: 5, type: !14, isLocal: false, isDefinition: true)
-!14 = !DICompositeType(tag: DW_TAG_array_type, baseType: !15, size: 3520, align: 64, elements: !5)
-!15 = !DICompositeType(tag: DW_TAG_array_type, baseType: !4, size: 320, align: 64, elements: !16)
+!14 = !DICompositeType(tag: DW_TAG_array_type, baseType: !15, size: 3520, align: 32, elements: !5)
+!15 = !DICompositeType(tag: DW_TAG_array_type, baseType: !4, size: 320, align: 32, elements: !16)
 !16 = !{!11}
 !17 = !{i32 2, !"Dwarf Version", i32 5}
 !18 = !{i32 2, !"Debug Info Version", i32 3}
diff --git a/src/codegen/tests/snapshots/rusty__codegen__tests__debug_tests__global_var_string_added_to_debug_info.snap b/src/codegen/tests/snapshots/rusty__codegen__tests__debug_tests__global_var_string_added_to_debug_info.snap
@@ -14,13 +14,13 @@ source_filename = "<internal>"
 !0 = !DIGlobalVariableExpression(var: !1, expr: !DIExpression())
 !1 = distinct !DIGlobalVariable(name: "a", scope: !2, file: !2, line: 3, type: !3, isLocal: false, isDefinition: true)
 !2 = !DIFile(filename: "<internal>", directory: "")
-!3 = !DICompositeType(tag: DW_TAG_array_type, baseType: !4, size: 80, align: 64, elements: !5)
+!3 = !DICompositeType(tag: DW_TAG_array_type, baseType: !4, size: 648, align: 8, elements: !5)
 !4 = !DIBasicType(name: "char", size: 8, encoding: DW_ATE_UTF, flags: DIFlagPublic)
 !5 = !{!6}
 !6 = !DISubrange(count: 81, lowerBound: 0)
 !7 = !DIGlobalVariableExpression(var: !8, expr: !DIExpression())
 !8 = distinct !DIGlobalVariable(name: "b", scope: !2, file: !2, line: 4, type: !9, isLocal: false, isDefinition: true)
-!9 = !DICompositeType(tag: DW_TAG_array_type, baseType: !10, size: 160, align: 64, elements: !5)
+!9 = !DICompositeType(tag: DW_TAG_array_type, baseType: !10, size: 1296, align: 16, elements: !5)
 !10 = !DIBasicType(name: "wchar", size: 16, encoding: DW_ATE_UTF, flags: DIFlagPublic)
 !11 = !{i32 2, !"Dwarf Version", i32 5}
 !12 = !{i32 2, !"Debug Info Version", i32 3}
diff --git a/src/codegen/tests/snapshots/rusty__codegen__tests__debug_tests__global_var_struct_added_to_debug_info.snap b/src/codegen/tests/snapshots/rusty__codegen__tests__debug_tests__global_var_struct_added_to_debug_info.snap
@@ -23,8 +23,8 @@ source_filename = "<internal>"
 !6 = !DIBasicType(name: "DINT", size: 32, encoding: DW_ATE_signed, flags: DIFlagPublic)
 !7 = !DIDerivedType(tag: DW_TAG_member, name: "b", scope: !2, file: !2, line: 4, baseType: !8, size: 64, align: 64, offset: 64, flags: DIFlagPublic)
 !8 = !DIBasicType(name: "LREAL", size: 64, encoding: DW_ATE_float, flags: DIFlagPublic)
-!9 = !DIDerivedType(tag: DW_TAG_member, name: "c", scope: !2, file: !2, line: 5, baseType: !10, size: 352, align: 64, offset: 128, flags: DIFlagPublic)
-!10 = !DICompositeType(tag: DW_TAG_array_type, baseType: !6, size: 352, align: 64, elements: !11)
+!9 = !DIDerivedType(tag: DW_TAG_member, name: "c", scope: !2, file: !2, line: 5, baseType: !10, size: 352, align: 32, offset: 128, flags: DIFlagPublic)
+!10 = !DICompositeType(tag: DW_TAG_array_type, baseType: !6, size: 352, align: 32, elements: !11)
 !11 = !{!12}
 !12 = !DISubrange(count: 11, lowerBound: 0)
 !13 = !DIGlobalVariableExpression(var: !14, expr: !DIExpression())
@@ -69,8 +69,8 @@ attributes #0 = { nofree nosync nounwind readnone speculatable willreturn }
 !6 = !DIBasicType(name: "DINT", size: 32, encoding: DW_ATE_signed, flags: DIFlagPublic)
 !7 = !DIDerivedType(tag: DW_TAG_member, name: "b", scope: !2, file: !2, line: 4, baseType: !8, size: 64, align: 64, offset: 64, flags: DIFlagPublic)
 !8 = !DIBasicType(name: "LREAL", size: 64, encoding: DW_ATE_float, flags: DIFlagPublic)
-!9 = !DIDerivedType(tag: DW_TAG_member, name: "c", scope: !2, file: !2, line: 5, baseType: !10, size: 352, align: 64, offset: 128, flags: DIFlagPublic)
-!10 = !DICompositeType(tag: DW_TAG_array_type, baseType: !6, size: 352, align: 64, elements: !11)
+!9 = !DIDerivedType(tag: DW_TAG_member, name: "c", scope: !2, file: !2, line: 5, baseType: !10, size: 352, align: 32, offset: 128, flags: DIFlagPublic)
+!10 = !DICompositeType(tag: DW_TAG_array_type, baseType: !6, size: 352, align: 32, elements: !11)
 !11 = !{!12}
 !12 = !DISubrange(count: 11, lowerBound: 0)
 !13 = !{i32 2, !"Dwarf Version", i32 5}
@@ -114,8 +114,8 @@ declare !dbg !25 void @__init_mystruct(%myStruct*)
 !6 = !DIBasicType(name: "DINT", size: 32, encoding: DW_ATE_signed, flags: DIFlagPublic)
 !7 = !DIDerivedType(tag: DW_TAG_member, name: "b", scope: !2, file: !2, line: 4, baseType: !8, size: 64, align: 64, offset: 64, flags: DIFlagPublic)
 !8 = !DIBasicType(name: "LREAL", size: 64, encoding: DW_ATE_float, flags: DIFlagPublic)
-!9 = !DIDerivedType(tag: DW_TAG_member, name: "c", scope: !2, file: !2, line: 5, baseType: !10, size: 352, align: 64, offset: 128, flags: DIFlagPublic)
-!10 = !DICompositeType(tag: DW_TAG_array_type, baseType: !6, size: 352, align: 64, elements: !11)
+!9 = !DIDerivedType(tag: DW_TAG_member, name: "c", scope: !2, file: !2, line: 5, baseType: !10, size: 352, align: 32, offset: 128, flags: DIFlagPublic)
+!10 = !DICompositeType(tag: DW_TAG_array_type, baseType: !6, size: 352, align: 32, elements: !11)
 !11 = !{!12}
 !12 = !DISubrange(count: 11, lowerBound: 0)
 !13 = !DIGlobalVariableExpression(var: !14, expr: !DIExpression())
diff --git a/src/datalayout.rs b/src/datalayout.rs
@@ -85,6 +85,7 @@ impl MemoryLocation {
         MemoryLocation(Bytes(value))
     }
 
+    // see https://en.wikipedia.org/wiki/Data_structure_alignment#Computing_padding
     pub fn align_to(self, align: Bytes) -> Self {
         let align = align.value() - 1;
         MemoryLocation::new((self.value() + align) & !align)
diff --git a/src/typesystem.rs b/src/typesystem.rs
@@ -621,7 +621,7 @@ impl DataTypeInformation {
             DataTypeInformation::String { size, encoding } => size
                 .as_int_value(index)
                 .map(|size| encoding.get_bytes_per_char() * size as u32)
-                .map(Bytes::from_bits)
+                .map(Bytes::new)
                 .unwrap(),
             DataTypeInformation::Struct { members, .. } => members
                 .iter()
@@ -669,15 +669,13 @@ impl DataTypeInformation {
         let type_layout = index.get_type_layout();
         match self {
             DataTypeInformation::Array { inner_type_name, .. } => {
-                let inner_type = index.get_type_information_or_void(inner_type_name);
-                if inner_type.get_alignment(index) > type_layout.i64 {
-                    type_layout.v128
-                } else {
-                    type_layout.v64
-                }
+                index.get_type_information_or_void(inner_type_name).get_alignment(index)
             }
             DataTypeInformation::Struct { .. } => type_layout.aggregate,
-            DataTypeInformation::String { .. } => type_layout.v64, //Strings are arrays
+            DataTypeInformation::String { encoding, .. } => match encoding {
+                StringEncoding::Utf8 => Bytes::new(1),
+                StringEncoding::Utf16 => Bytes::new(2),
+            },
             DataTypeInformation::Pointer { .. } => type_layout.p64,
             DataTypeInformation::Integer { size, semantic_size, .. } => {
                 if let Some(1) = semantic_size {

Original file line number	Diff line number	Diff line change
`@@ -85,6 +85,7 @@ impl MemoryLocation {`
`85`	`85`	`MemoryLocation(Bytes(value))`
`86`	`86`	`}`
`87`	`87`
	`88`	`+ // see https://en.wikipedia.org/wiki/Data_structure_alignment#Computing_padding`
`88`	`89`	`pub fn align_to(self, align: Bytes) -> Self {`
`89`	`90`	`let align = align.value() - 1;`
`90`	`91`	`MemoryLocation::new((self.value() + align) & !align)`