Skip to content

Commit 5a71656

Browse files
NerixyzNerixyz
committed
[LLDB][NativePDB] Estimate symbol sizes
1 parent c1423f3 commit 5a71656

File tree

3 files changed

+69
-35
lines changed

3 files changed

+69
-35
lines changed

‎lldb/source/Plugins/SymbolFile/NativePDB/SymbolFileNativePDB.cpp‎

Lines changed: 51 additions & 16 deletions
Original file line numberDiff line numberDiff line change
@@ -1130,16 +1130,47 @@ void SymbolFileNativePDB::AddSymbols(Symtab &symtab) {
11301130
if (!section_list)
11311131
return;
11321132

1133-
for (auto pid : m_index->publics().getPublicsTable()) {
1133+
PublicSym32 last_sym;
1134+
size_t last_sym_idx = 0;
1135+
lldb::SectionSP section_sp;
1136+
1137+
// To estimate the size of a symbol, we use the difference to the next symbol.
1138+
// If there's no next symbol or the section/segment changed, the symbol will
1139+
// take the remaining space. The estimate can be too high in case there's
1140+
// padding between symbols. This similar to the algorithm used by the DIA
1141+
// SDK.
1142+
auto finish_last_symbol = [&](const PublicSym32 *next) {
1143+
if (!section_sp)
1144+
return;
1145+
Symbol *last = symtab.SymbolAtIndex(last_sym_idx);
1146+
if (!last)
1147+
return;
1148+
1149+
if (next && last_sym.Segment == next->Segment) {
1150+
assert(last_sym.Offset <= next->Offset);
1151+
last->SetByteSize(next->Offset - last_sym.Offset);
1152+
} else {
1153+
// the last symbol was the last in its section
1154+
assert(section_sp->GetByteSize() >= last_sym.Offset);
1155+
assert(!next || next->Segment > last_sym.Segment);
1156+
last->SetByteSize(section_sp->GetByteSize() - last_sym.Offset);
1157+
}
1158+
};
1159+
1160+
// the address map is sorted by the address of a symbol
1161+
for (auto pid : m_index->publics().getAddressMap()) {
11341162
PdbGlobalSymId global{pid, true};
11351163
CVSymbol sym = m_index->ReadSymbolRecord(global);
11361164
auto kind = sym.kind();
11371165
if (kind != S_PUB32)
11381166
continue;
11391167
PublicSym32 pub =
11401168
llvm::cantFail(SymbolDeserializer::deserializeAs<PublicSym32>(sym));
1169+
finish_last_symbol(&pub);
1170+
1171+
if (!section_sp || last_sym.Segment != pub.Segment)
1172+
section_sp = section_list->FindSectionByID(pub.Segment);
11411173

1142-
auto section_sp = section_list->FindSectionByID(pub.Segment);
11431174
if (!section_sp)
11441175
continue;
11451176

@@ -1148,20 +1179,24 @@ void SymbolFileNativePDB::AddSymbols(Symtab &symtab) {
11481179
(pub.Flags & PublicSymFlags::Code) != PublicSymFlags::None)
11491180
type = eSymbolTypeCode;
11501181

1151-
symtab.AddSymbol(Symbol(/*symID=*/pid,
1152-
/*name=*/pub.Name,
1153-
/*type=*/type,
1154-
/*external=*/true,
1155-
/*is_debug=*/true,
1156-
/*is_trampoline=*/false,
1157-
/*is_artificial=*/false,
1158-
/*section_sp=*/section_sp,
1159-
/*value=*/pub.Offset,
1160-
/*size=*/0,
1161-
/*size_is_valid=*/false,
1162-
/*contains_linker_annotations=*/false,
1163-
/*flags=*/0));
1164-
}
1182+
last_sym_idx =
1183+
symtab.AddSymbol(Symbol(/*symID=*/pid,
1184+
/*name=*/pub.Name,
1185+
/*type=*/type,
1186+
/*external=*/true,
1187+
/*is_debug=*/true,
1188+
/*is_trampoline=*/false,
1189+
/*is_artificial=*/false,
1190+
/*section_sp=*/section_sp,
1191+
/*value=*/pub.Offset,
1192+
/*size=*/0,
1193+
/*size_is_valid=*/false,
1194+
/*contains_linker_annotations=*/false,
1195+
/*flags=*/0));
1196+
last_sym = pub;
1197+
}
1198+
1199+
finish_last_symbol(nullptr);
11651200
}
11661201

11671202
size_t SymbolFileNativePDB::ParseFunctions(CompileUnit &comp_unit) {

‎lldb/test/API/functionalities/multiple-slides/TestMultipleSlides.py‎

Lines changed: 3 additions & 4 deletions
Original file line numberDiff line numberDiff line change
@@ -12,10 +12,6 @@
1212
class MultipleSlidesTestCase(TestBase):
1313
NO_DEBUG_INFO_TESTCASE = True
1414

15-
# The intermediate object main.o is compiled without debug info, but
16-
# a.out is linked with `-gdwarf` on Windows. This creates a PDB.
17-
# However, in the native PDB plugin, the symbols don't have a size.
18-
@expectedFailureWindows
1915
def test_mulitple_slides(self):
2016
"""Test that a binary can be slid multiple times correctly."""
2117
self.build()
@@ -33,10 +29,13 @@ def test_mulitple_slides(self):
3329
first_sym.GetEndAddress().GetOffset()
3430
- first_sym.GetStartAddress().GetOffset()
3531
)
32+
int_size = target.FindFirstType("int").GetByteSize()
33+
self.assertGreaterEqual(first_size, 2048 * int_size)
3634
second_size = (
3735
second_sym.GetEndAddress().GetOffset()
3836
- second_sym.GetStartAddress().GetOffset()
3937
)
38+
self.assertGreaterEqual(second_size, 2048 * int_size)
4039

4140
# View the first element of `first` and `second` while
4241
# they have no load address set.

‎lldb/test/Shell/SymbolFile/NativePDB/symtab.cpp‎

Lines changed: 15 additions & 15 deletions
Original file line numberDiff line numberDiff line change
@@ -42,18 +42,18 @@ int main(int argc, char **argv) {
4242
return ns::a_function() + b.b_func();
4343
}
4444

45-
// CHECK-DAG: Code {{.*}} main
46-
// CHECK-DAG: Code {{.*}} ?b_func@?$B@F@ns@@QEBAHXZ
47-
// CHECK-DAG: Code {{.*}} ?something@A@@QEAAXXZ
48-
// CHECK-DAG: Code {{.*}} ??_GDyn@ns@@UEAAPEAXI@Z
49-
// CHECK-DAG: Code {{.*}} ??2@YAPEAX_K@Z
50-
// CHECK-DAG: Code {{.*}} ??3@YAXPEAX_K@Z
51-
// CHECK-DAG: Code {{.*}} ?static_fn@C@?$B@H@ns@@SAHXZ
52-
// CHECK-DAG: Code {{.*}} ?a_function@ns@@YAHXZ
53-
// CHECK-DAG: Code {{.*}} ?static_fn@C@?$B@_N@ns@@SAHXZ
54-
// CHECK-DAG: Code {{.*}} ??1Dyn@ns@@UEAA@XZ
55-
// CHECK-DAG: Code {{.*}} ??0Dyn@ns@@QEAA@XZ
56-
// CHECK-DAG: Data {{.*}} ?global_int@@3HA
57-
// CHECK-DAG: Data {{.*}} ??_7Dyn@ns@@6B@
58-
// CHECK-DAG: Data {{.*}} ?global_a@@3UA@@A
59-
// CHECK-DAG: Data {{.*}} ?global_c@@3UC@?$B@_J@ns@@A
45+
// CHECK-DAG: Code 0x{{[0-9a-f]+}} 0x{{0*[1-9a-f][0-9a-f]*}} 0x00000000 main
46+
// CHECK-DAG: Code 0x{{[0-9a-f]+}} 0x{{0*[1-9a-f][0-9a-f]*}} 0x00000000 ?b_func@?$B@F@ns@@QEBAHXZ
47+
// CHECK-DAG: Code 0x{{[0-9a-f]+}} 0x{{0*[1-9a-f][0-9a-f]*}} 0x00000000 ?something@A@@QEAAXXZ
48+
// CHECK-DAG: Code 0x{{[0-9a-f]+}} 0x{{0*[1-9a-f][0-9a-f]*}} 0x00000000 ??_GDyn@ns@@UEAAPEAXI@Z
49+
// CHECK-DAG: Code 0x{{[0-9a-f]+}} 0x{{0*[1-9a-f][0-9a-f]*}} 0x00000000 ??2@YAPEAX_K@Z
50+
// CHECK-DAG: Code 0x{{[0-9a-f]+}} 0x{{0*[1-9a-f][0-9a-f]*}} 0x00000000 ??3@YAXPEAX_K@Z
51+
// CHECK-DAG: Code 0x{{[0-9a-f]+}} 0x{{0*[1-9a-f][0-9a-f]*}} 0x00000000 ?static_fn@C@?$B@H@ns@@SAHXZ
52+
// CHECK-DAG: Code 0x{{[0-9a-f]+}} 0x{{0*[1-9a-f][0-9a-f]*}} 0x00000000 ?a_function@ns@@YAHXZ
53+
// CHECK-DAG: Code 0x{{[0-9a-f]+}} 0x{{0*[1-9a-f][0-9a-f]*}} 0x00000000 ?static_fn@C@?$B@_N@ns@@SAHXZ
54+
// CHECK-DAG: Code 0x{{[0-9a-f]+}} 0x{{0*[1-9a-f][0-9a-f]*}} 0x00000000 ??1Dyn@ns@@UEAA@XZ
55+
// CHECK-DAG: Code 0x{{[0-9a-f]+}} 0x{{0*[1-9a-f][0-9a-f]*}} 0x00000000 ??0Dyn@ns@@QEAA@XZ
56+
// CHECK-DAG: Data 0x{{[0-9a-f]+}} 0x{{0*[1-9a-f][0-9a-f]*}} 0x00000000 ?global_int@@3HA
57+
// CHECK-DAG: Data 0x{{[0-9a-f]+}} 0x{{0*[1-9a-f][0-9a-f]*}} 0x00000000 ??_7Dyn@ns@@6B@
58+
// CHECK-DAG: Data 0x{{[0-9a-f]+}} 0x{{0*[1-9a-f][0-9a-f]*}} 0x00000000 ?global_a@@3UA@@A
59+
// CHECK-DAG: Data 0x{{[0-9a-f]+}} 0x{{0*[1-9a-f][0-9a-f]*}} 0x00000000 ?global_c@@3UC@?$B@_J@ns@@A

0 commit comments

Comments
 (0)