Compare commits

..

13 Commits

Author SHA1 Message Date
LagoLunatic e1da90943c Version v3.0.0-beta.14 2025-08-13 01:45:05 -04:00
LagoLunatic b5713db333 Update dependencies 2025-08-13 01:42:34 -04:00
LagoLunatic 4c3f5e9836 Merge pull request #236 from LagoLunatic/symbol-ctx
Fix context menu not appearing when right clicking the function name in the function diff view
2025-08-11 17:28:26 -04:00
LagoLunatic e9ce02feb0 Disable double tooltip for elided function name in function diff view 2025-08-08 20:36:48 -04:00
LagoLunatic 9a378d85ed Fix context menu on function name in function diff view 2025-08-08 20:35:16 -04:00
LagoLunatic c8ff45f2c8 Merge pull request #234 from LagoLunatic/no-diff-data
Fix data section not showing when there is no section on the other side
2025-08-08 20:32:28 -04:00
LagoLunatic 34e4513c69 Fix data section not showing when there is no section on the other side 2025-08-06 16:09:18 -04:00
Luke Street a015971c20 Use deprecated egui::menu as temp workaround
egui 0.32 refactored menus, and now having a
ComboBox within the menu does not work properly.
Temporarily use the deprecated menu instead.
2025-08-02 13:34:10 -06:00
Luke Street e67d5998b3 Enable PS instructions for any 32-bit PPC ELF
Fixes an issue where ProDG for GameCube objects
were not enabling PS instruction support.
2025-08-02 11:36:35 -06:00
Luke Street 91bc23edfc Fix objdiff-wasm build 2025-08-02 11:32:04 -06:00
Luke Street c9c3b32376 Use let chains (a.k.a. cargo clippy --fix) 2025-08-02 11:27:28 -06:00
Luke Street 0dc123b064 Don't fail on line info parsing; use gimli::RelocateReader
Workaround for #228
2025-08-02 11:27:28 -06:00
Luke Street 1e62d4664c Make function size inference logic arch-specific
For MIPS, account for delay slot nops. For x86,
check for trailing nops (0x90). For PPC, check
for 4-byte 0x00 padding.

Resolves #229
2025-08-02 10:56:26 -06:00
29 changed files with 1052 additions and 917 deletions
Generated
+143 -151
View File
File diff suppressed because it is too large Load Diff
+1 -1
View File
@@ -14,7 +14,7 @@ strip = "debuginfo"
codegen-units = 1
[workspace.package]
version = "3.0.0-beta.12"
version = "3.0.0-beta.14"
authors = ["Luke Street <luke@street.dev>"]
edition = "2024"
license = "MIT OR Apache-2.0"
+5 -6
View File
@@ -248,13 +248,12 @@ fn report_object(
{
continue;
}
if let Some(existing_functions) = &mut existing_functions {
if (symbol.flags.contains(SymbolFlag::Global)
if let Some(existing_functions) = &mut existing_functions
&& (symbol.flags.contains(SymbolFlag::Global)
|| symbol.flags.contains(SymbolFlag::Weak))
&& !existing_functions.insert(symbol.name.clone())
{
continue;
}
&& !existing_functions.insert(symbol.name.clone())
{
continue;
}
let match_percent = symbol_diff.match_percent.unwrap_or_else(|| {
// Support cases where we don't have a target object,
+29 -28
View File
@@ -170,32 +170,31 @@ impl UiView for FunctionDiffUi {
let mut prev_text = None;
let mut prev_margin_text = None;
if self.three_way {
if let Some((obj, symbol_idx, symbol_diff)) =
if self.three_way
&& let Some((obj, symbol_idx, symbol_diff)) =
get_symbol(state.prev_obj.as_ref(), self.prev_sym)
{
let mut text = Text::default();
let rect = content_chunks[4].inner(Margin::new(0, 1));
self.print_sym(
&mut text,
obj,
symbol_idx,
symbol_diff,
&state.diff_obj_config,
rect,
&self.right_highlight,
result,
true,
);
max_width = max_width.max(text.width());
prev_text = Some(text);
{
let mut text = Text::default();
let rect = content_chunks[4].inner(Margin::new(0, 1));
self.print_sym(
&mut text,
obj,
symbol_idx,
symbol_diff,
&state.diff_obj_config,
rect,
&self.right_highlight,
result,
true,
);
max_width = max_width.max(text.width());
prev_text = Some(text);
// Render margin
let mut text = Text::default();
let rect = content_chunks[3].inner(Margin::new(1, 1));
self.print_margin(&mut text, symbol_diff, rect);
prev_margin_text = Some(text);
}
// Render margin
let mut text = Text::default();
let rect = content_chunks[3].inner(Margin::new(1, 1));
self.print_margin(&mut text, symbol_diff, rect);
prev_margin_text = Some(text);
}
let max_scroll_x =
@@ -561,10 +560,12 @@ impl FunctionDiffUi {
let len = label_text.len();
let highlighted =
highlight_kind != HighlightKind::None && *highlight == highlight_kind;
if let Some((cx, cy)) = result.click_xy {
if cx >= sx && cx < sx + len as u16 && cy == sy {
new_highlight = Some(highlight_kind);
}
if let Some((cx, cy)) = result.click_xy
&& cx >= sx
&& cx < sx + len as u16
&& cy == sy
{
new_highlight = Some(highlight_kind);
}
let mut style = Style::new().fg(match segment.color {
DiffTextColor::Normal => Color::Gray,
+6 -1
View File
@@ -62,7 +62,10 @@ config = [
"dep:semver",
"dep:typed-path",
]
dwarf = ["dep:gimli"]
dwarf = [
"dep:gimli",
"dep:typed-arena",
]
serde = [
"dep:pbjson",
"dep:pbjson-build",
@@ -78,6 +81,7 @@ std = [
"prost?/std",
"serde?/std",
"similar?/std",
"typed-arena?/std",
"typed-path?/std",
"dep:filetime",
"dep:memmap2",
@@ -143,6 +147,7 @@ serde_json = { version = "1.0", default-features = false, features = ["alloc"],
# dwarf
gimli = { version = "0.32", default-features = false, features = ["read"], optional = true }
typed-arena = { version = "2.0", default-features = false, optional = true }
# ppc
cwdemangle = { version = "1.0", optional = true }
File diff suppressed because it is too large Load Diff
+47 -10
View File
@@ -15,7 +15,7 @@ use crate::{
diff::{DiffObjConfig, MipsAbi, MipsInstrCategory, display::InstructionPart},
obj::{
InstructionArg, InstructionArgValue, InstructionRef, Relocation, RelocationFlags,
ResolvedInstructionRef, ResolvedRelocation, SymbolFlag, SymbolFlagSet,
ResolvedInstructionRef, ResolvedRelocation, Section, Symbol, SymbolFlag, SymbolFlagSet,
},
};
@@ -140,6 +140,14 @@ impl ArchMips {
})
}
fn default_instruction_flags(&self) -> rabbitizer::InstructionFlags {
match self.isa_extension {
Some(extension) => rabbitizer::InstructionFlags::new_extension(extension),
None => rabbitizer::InstructionFlags::new(IsaVersion::MIPS_III),
}
.with_abi(self.abi)
}
fn instruction_flags(&self, diff_config: &DiffObjConfig) -> rabbitizer::InstructionFlags {
let isa_extension = match diff_config.mips_instr_category {
MipsInstrCategory::Auto => self.isa_extension,
@@ -151,7 +159,7 @@ impl ArchMips {
};
match isa_extension {
Some(extension) => rabbitizer::InstructionFlags::new_extension(extension),
None => rabbitizer::InstructionFlags::new_isa(IsaVersion::MIPS_III, None),
None => rabbitizer::InstructionFlags::new(IsaVersion::MIPS_III),
}
.with_abi(match diff_config.mips_abi {
MipsAbi::Auto => self.abi,
@@ -234,17 +242,16 @@ impl Arch for ArchMips {
object::RelocationFlags::Elf { r_type } => {
if relocation.has_implicit_addend() {
// Check for paired R_MIPS_HI16 and R_MIPS_LO16 relocations.
if let elf::R_MIPS_HI16 | elf::R_MIPS_LO16 = r_type {
if let Some(addend) = self
if let elf::R_MIPS_HI16 | elf::R_MIPS_LO16 = r_type
&& let Some(addend) = self
.paired_relocations
.get(section.index().0)
.and_then(|m| m.get(&address).copied())
{
return Ok(Some(RelocationOverride {
target: RelocationOverrideTarget::Keep,
addend,
}));
}
{
return Ok(Some(RelocationOverride {
target: RelocationOverrideTarget::Keep,
addend,
}));
}
let data = section.data()?;
@@ -331,6 +338,36 @@ impl Arch for ArchMips {
}
flags
}
fn infer_function_size(
&self,
symbol: &Symbol,
section: &Section,
next_address: u64,
) -> Result<u64> {
// Trim any trailing 4-byte zeroes from the end (nops)
let mut new_address = next_address;
while new_address >= symbol.address + 4
&& let Some(data) = section.data_range(new_address - 4, 4)
&& data == [0u8; 4]
{
new_address -= 4;
}
// Check if the last instruction has a delay slot, if so, include the delay slot nop
if new_address + 4 <= next_address
&& new_address >= symbol.address + 4
&& let Some(data) = section.data_range(new_address - 4, 4)
&& let instruction = rabbitizer::Instruction::new(
self.endianness.read_u32_bytes(data.try_into().unwrap()),
Vram::new((new_address - 4) as u32),
self.default_instruction_flags(),
)
&& instruction.opcode().has_delay_slot()
{
new_address += 4;
}
Ok(new_address.saturating_sub(symbol.address))
}
}
fn push_args(
+13 -4
View File
@@ -215,10 +215,10 @@ impl dyn Arch {
// Remove any branch destinations that are outside the function range
for ins in result.iter_mut() {
if let Some(branch_dest) = ins.branch_dest {
if branch_dest < function_start || branch_dest >= function_end {
ins.branch_dest = None;
}
if let Some(branch_dest) = ins.branch_dest
&& (branch_dest < function_start || branch_dest >= function_end)
{
ins.branch_dest = None;
}
}
@@ -406,6 +406,15 @@ pub trait Arch: Send + Sync + Debug {
) -> Vec<ContextItem> {
Vec::new()
}
fn infer_function_size(
&self,
symbol: &Symbol,
_section: &Section,
next_address: u64,
) -> Result<u64> {
Ok(next_address.saturating_sub(symbol.address))
}
}
pub fn new_arch(object: &object::File) -> Result<Box<dyn Arch>> {
+18 -20
View File
@@ -514,14 +514,14 @@ pub fn ppc_data_flow_analysis(
}
fn get_string_data(obj: &Object, symbol_index: usize, offset: Simm) -> Option<&str> {
if let Some(sym) = obj.symbols.get(symbol_index) {
if sym.name.starts_with("@stringBase") && offset.0 != 0 {
if let Some(data) = obj.symbol_data(symbol_index) {
let bytes = &data[offset.0 as usize..];
if let Ok(Ok(str)) = CStr::from_bytes_until_nul(bytes).map(|x| x.to_str()) {
return Some(str);
}
}
if let Some(sym) = obj.symbols.get(symbol_index)
&& sym.name.starts_with("@stringBase")
&& offset.0 != 0
&& let Some(data) = obj.symbol_data(symbol_index)
{
let bytes = &data[offset.0 as usize..];
if let Ok(Ok(str)) = CStr::from_bytes_until_nul(bytes).map(|x| x.to_str()) {
return Some(str);
}
}
None
@@ -577,19 +577,17 @@ fn generate_flow_analysis_result(
let registers = register_state_at.get(index as usize).unwrap_or(&default_register_state);
if let (powerpc::Opcode::Addi, Argument::GPR(rel), Argument::Simm(offset)) =
(ins.op, args[1], args[2])
&& let RegisterContent::Symbol(sym_index) = registers[rel]
&& let Some(str) = get_string_data(obj, sym_index, offset)
{
if let RegisterContent::Symbol(sym_index) = registers[rel] {
if let Some(str) = get_string_data(obj, sym_index, offset) {
// Show the string constant in the analysis result
let formatted = format!("\"{str}\"");
analysis_result.set_argument_value_at_address(
ins_address,
2,
FlowAnalysisValue::Text(clamp_text_length(formatted, 20)),
);
// Don't continue, we want to show the stringbase value as well
}
}
// Show the string constant in the analysis result
let formatted = format!("\"{str}\"");
analysis_result.set_argument_value_at_address(
ins_address,
2,
FlowAnalysisValue::Text(clamp_text_length(formatted, 20)),
);
// Don't continue, we want to show the stringbase value as well
}
let is_store = is_store_instruction(ins.op);
+49 -32
View File
@@ -20,7 +20,7 @@ use crate::{
},
obj::{
FlowAnalysisResult, InstructionRef, Object, Relocation, RelocationFlags,
ResolvedInstructionRef, ResolvedRelocation, Symbol, SymbolFlag, SymbolFlagSet,
ResolvedInstructionRef, ResolvedRelocation, Section, Symbol, SymbolFlag, SymbolFlagSet,
},
};
@@ -66,7 +66,9 @@ impl ArchPpc {
if file.is_64() {
powerpc::Extension::Ppc64 | powerpc::Extension::AltiVec
} else {
powerpc::Extension::AltiVec.into()
// Gekko/Broadway objects often use the EF_PPC_EMB flag,
// but ProDG in particular does not emit it.
powerpc::Extensions::gekko_broadway()
}
}
};
@@ -455,6 +457,22 @@ impl Arch for ArchPpc {
}
out
}
fn infer_function_size(
&self,
symbol: &Symbol,
section: &Section,
mut next_address: u64,
) -> Result<u64> {
// Trim any trailing 4-byte zeroes from the end (padding)
while next_address >= symbol.address + 4
&& let Some(data) = section.data_range(next_address - 4, 4)
&& data == [0u8; 4]
{
next_address -= 4;
}
Ok(next_address.saturating_sub(symbol.address))
}
}
impl ArchPpc {
@@ -850,44 +868,43 @@ fn generate_fake_pool_relocations_for_function(
break;
}
}
if let Some(branch_dest) = branch_dest {
if branch_dest >= func_address as u32
&& (branch_dest - func_address as u32) < code.len() as u32
{
let dest_offset_into_func = branch_dest - func_address as u32;
let dest_code_slice = &code[dest_offset_into_func as usize..];
match ins.op {
Opcode::Bc => {
// Conditional branch.
// Add the branch destination to the queue to do later.
if let Some(branch_dest) = branch_dest
&& branch_dest >= func_address as u32
&& (branch_dest - func_address as u32) < code.len() as u32
{
let dest_offset_into_func = branch_dest - func_address as u32;
let dest_code_slice = &code[dest_offset_into_func as usize..];
match ins.op {
Opcode::Bc => {
// Conditional branch.
// Add the branch destination to the queue to do later.
ins_iters_with_gpr_state.push((
InsIter::new(dest_code_slice, branch_dest, extensions),
gpr_pool_relocs.clone(),
));
// Then continue on with the current iterator.
}
Opcode::B => {
if simplified.mnemonic != "bl" {
// Unconditional branch.
// Add the branch destination to the queue.
ins_iters_with_gpr_state.push((
InsIter::new(dest_code_slice, branch_dest, extensions),
gpr_pool_relocs.clone(),
));
// Then continue on with the current iterator.
// Break out of the current iterator so we can do the newly added one.
break;
}
Opcode::B => {
if simplified.mnemonic != "bl" {
// Unconditional branch.
// Add the branch destination to the queue.
ins_iters_with_gpr_state.push((
InsIter::new(dest_code_slice, branch_dest, extensions),
gpr_pool_relocs.clone(),
));
// Break out of the current iterator so we can do the newly added one.
break;
}
}
_ => unreachable!(),
}
_ => unreachable!(),
}
}
if let Opcode::Bcctr = ins.op {
if simplified.mnemonic == "bctr" {
// Unconditional branch to count register.
// Likely a jump table.
gpr_state_at_bctr.insert(cur_addr, gpr_pool_relocs.clone());
}
if let Opcode::Bcctr = ins.op
&& simplified.mnemonic == "bctr"
{
// Unconditional branch to count register.
// Likely a jump table.
gpr_state_at_bctr.insert(cur_addr, gpr_pool_relocs.clone());
}
// Then handle keeping track of which GPR contains which pool relocation.
+47 -1
View File
@@ -11,7 +11,7 @@ use object::{Endian as _, Object as _, ObjectSection as _, elf, pe};
use crate::{
arch::{Arch, RelocationOverride, RelocationOverrideTarget},
diff::{DiffObjConfig, X86Formatter, display::InstructionPart},
obj::{InstructionRef, Relocation, RelocationFlags, ResolvedInstructionRef},
obj::{InstructionRef, Relocation, RelocationFlags, ResolvedInstructionRef, Section, Symbol},
};
#[derive(Debug)]
@@ -303,6 +303,52 @@ impl Arch for ArchX86 {
fn data_reloc_size(&self, flags: RelocationFlags) -> usize {
self.reloc_size(flags).unwrap_or(1)
}
fn infer_function_size(
&self,
symbol: &Symbol,
section: &Section,
next_address: u64,
) -> Result<u64> {
let Ok(size) = (next_address - symbol.address).try_into() else {
return Ok(next_address.saturating_sub(symbol.address));
};
let Some(code) = section.data_range(symbol.address, size) else {
return Ok(0);
};
// Decode instructions to find the last non-NOP instruction
let mut decoder = self.decoder(code, symbol.address);
let mut instruction = Instruction::default();
let mut new_address = 0;
let mut reloc_iter = section.relocations.iter().peekable();
'outer: while decoder.can_decode() {
let address = decoder.ip();
while let Some(reloc) = reloc_iter.peek() {
match reloc.address.cmp(&address) {
Ordering::Less => {
reloc_iter.next();
}
Ordering::Equal => {
// If the instruction starts at a relocation, it's inline data
let reloc_size = self.reloc_size(reloc.flags).with_context(|| {
format!("Unsupported inline x86 relocation {:?}", reloc.flags)
})?;
if decoder.set_position(decoder.position() + reloc_size).is_ok() {
new_address = address + reloc_size as u64;
decoder.set_ip(new_address);
continue 'outer;
}
}
Ordering::Greater => break,
}
}
decoder.decode_out(&mut instruction);
if instruction.mnemonic() != iced_x86::Mnemonic::Nop {
new_address = instruction.next_ip();
}
}
Ok(new_address.saturating_sub(symbol.address))
}
}
struct InstructionFormatterOutput<'a> {
+26
View File
@@ -127,6 +127,28 @@ fn diff_data_relocs_for_range<'left, 'right>(
diffs
}
pub fn no_diff_data_section(obj: &Object, section_idx: usize) -> Result<SectionDiff> {
let section = &obj.sections[section_idx];
let len = section.data.len();
let data = &section.data[0..len];
let data_diff =
vec![DataDiff { data: data.to_vec(), kind: DataDiffKind::None, len, ..Default::default() }];
let mut reloc_diffs = Vec::new();
for reloc in section.relocations.iter() {
let reloc_len = obj.arch.data_reloc_size(reloc.flags);
let range = reloc.address as usize..reloc.address as usize + reloc_len;
reloc_diffs.push(DataRelocationDiff {
reloc: reloc.clone(),
kind: DataDiffKind::None,
range,
});
}
Ok(SectionDiff { match_percent: Some(0.0), data_diff, reloc_diff: reloc_diffs })
}
/// Compare the data sections of two object files.
pub fn diff_data_section(
left_obj: &Object,
@@ -415,6 +437,10 @@ pub fn diff_generic_section(
))
}
pub fn no_diff_bss_section() -> Result<SectionDiff> {
Ok(SectionDiff { match_percent: Some(0.0), data_diff: vec![], reloc_diff: vec![] })
}
/// Compare the addresses and sizes of each symbol in the BSS sections.
pub fn diff_bss_section(
left_obj: &Object,
+7 -7
View File
@@ -385,13 +385,13 @@ pub fn symbol_context(obj: &Object, symbol_index: usize) -> Vec<ContextItem> {
if let Some(name) = &symbol.demangled_name {
out.push(ContextItem::Copy { value: name.clone(), label: None });
}
if symbol.section.is_some() {
if let Some(address) = symbol.virtual_address {
out.push(ContextItem::Copy {
value: format!("{address:x}"),
label: Some("virtual address".to_string()),
});
}
if symbol.section.is_some()
&& let Some(address) = symbol.virtual_address
{
out.push(ContextItem::Copy {
value: format!("{address:x}"),
label: Some("virtual address".to_string()),
});
}
out.append(&mut obj.arch.symbol_context(obj, symbol_index));
out
+88 -57
View File
@@ -13,7 +13,7 @@ use crate::{
code::{diff_code, no_diff_code},
data::{
diff_bss_section, diff_bss_symbol, diff_data_section, diff_data_symbol,
diff_generic_section,
diff_generic_section, no_diff_bss_section, no_diff_data_section,
},
},
obj::{InstructionRef, Object, Relocation, SectionKind, Symbol, SymbolFlag},
@@ -288,52 +288,84 @@ pub fn diff_objs(
}
for section_match in section_matches {
if let SectionMatch {
left: Some(left_section_idx),
right: Some(right_section_idx),
section_kind,
} = section_match
{
let (left_obj, left_out) = left.as_mut().unwrap();
let (right_obj, right_out) = right.as_mut().unwrap();
match section_kind {
SectionKind::Code => {
let (left_diff, right_diff) = diff_generic_section(
left_obj,
right_obj,
left_out,
right_out,
left_section_idx,
right_section_idx,
)?;
left_out.sections[left_section_idx] = left_diff;
right_out.sections[right_section_idx] = right_diff;
match section_match {
SectionMatch {
left: Some(left_section_idx),
right: Some(right_section_idx),
section_kind,
} => {
let (left_obj, left_out) = left.as_mut().unwrap();
let (right_obj, right_out) = right.as_mut().unwrap();
match section_kind {
SectionKind::Code => {
let (left_diff, right_diff) = diff_generic_section(
left_obj,
right_obj,
left_out,
right_out,
left_section_idx,
right_section_idx,
)?;
left_out.sections[left_section_idx] = left_diff;
right_out.sections[right_section_idx] = right_diff;
}
SectionKind::Data => {
let (left_diff, right_diff) = diff_data_section(
left_obj,
right_obj,
left_out,
right_out,
left_section_idx,
right_section_idx,
)?;
left_out.sections[left_section_idx] = left_diff;
right_out.sections[right_section_idx] = right_diff;
}
SectionKind::Bss | SectionKind::Common => {
let (left_diff, right_diff) = diff_bss_section(
left_obj,
right_obj,
left_out,
right_out,
left_section_idx,
right_section_idx,
)?;
left_out.sections[left_section_idx] = left_diff;
right_out.sections[right_section_idx] = right_diff;
}
SectionKind::Unknown => unreachable!(),
}
SectionKind::Data => {
let (left_diff, right_diff) = diff_data_section(
left_obj,
right_obj,
left_out,
right_out,
left_section_idx,
right_section_idx,
)?;
left_out.sections[left_section_idx] = left_diff;
right_out.sections[right_section_idx] = right_diff;
}
SectionMatch { left: Some(left_section_idx), right: None, section_kind } => {
let (left_obj, left_out) = left.as_mut().unwrap();
match section_kind {
SectionKind::Code => {}
SectionKind::Data => {
left_out.sections[left_section_idx] =
no_diff_data_section(left_obj, left_section_idx)?;
}
SectionKind::Bss | SectionKind::Common => {
left_out.sections[left_section_idx] = no_diff_bss_section()?;
}
SectionKind::Unknown => unreachable!(),
}
SectionKind::Bss | SectionKind::Common => {
let (left_diff, right_diff) = diff_bss_section(
left_obj,
right_obj,
left_out,
right_out,
left_section_idx,
right_section_idx,
)?;
left_out.sections[left_section_idx] = left_diff;
right_out.sections[right_section_idx] = right_diff;
}
SectionMatch { left: None, right: Some(right_section_idx), section_kind } => {
let (right_obj, right_out) = right.as_mut().unwrap();
match section_kind {
SectionKind::Code => {}
SectionKind::Data => {
right_out.sections[right_section_idx] =
no_diff_data_section(right_obj, right_section_idx)?;
}
SectionKind::Bss | SectionKind::Common => {
right_out.sections[right_section_idx] = no_diff_bss_section()?;
}
SectionKind::Unknown => unreachable!(),
}
SectionKind::Unknown => unreachable!(),
}
SectionMatch { left: None, right: None, .. } => {
// Should not happen
}
}
}
@@ -467,15 +499,15 @@ fn apply_symbol_mappings(
) -> Result<()> {
// If we're selecting a symbol to use as a comparison, mark it as used
// This ensures that we don't match it to another symbol at any point
if let Some(left_name) = &mapping_config.selecting_left {
if let Some(left_symbol) = left.symbol_by_name(left_name) {
left_used.insert(left_symbol);
}
if let Some(left_name) = &mapping_config.selecting_left
&& let Some(left_symbol) = left.symbol_by_name(left_name)
{
left_used.insert(left_symbol);
}
if let Some(right_name) = &mapping_config.selecting_right {
if let Some(right_symbol) = right.symbol_by_name(right_name) {
right_used.insert(right_symbol);
}
if let Some(right_name) = &mapping_config.selecting_right
&& let Some(right_symbol) = right.symbol_by_name(right_name)
{
right_used.insert(right_symbol);
}
// Apply manual symbol mappings
@@ -639,17 +671,16 @@ fn find_symbol(
// If they are at the same address in the same section
if in_symbol.name.starts_with('@')
&& matches!(section_kind, SectionKind::Data | SectionKind::Bss)
{
if let Some((symbol_idx, _)) = unmatched_symbols(obj, used).find(|(_, symbol)| {
&& let Some((symbol_idx, _)) = unmatched_symbols(obj, used).find(|(_, symbol)| {
let Some(section_index) = symbol.section else {
return false;
};
symbol.name.starts_with('@')
&& symbol.address == in_symbol.address
&& obj.sections[section_index].name == section_name
}) {
return Some(symbol_idx);
}
})
{
return Some(symbol_idx);
}
// Match Metrowerks symbol$1234 against symbol$2345
if let Some((prefix, suffix)) = in_symbol.name.split_once('$') {
+109
View File
@@ -0,0 +1,109 @@
use alloc::{borrow::Cow, vec::Vec};
use anyhow::{Context, Result};
use object::{Object as _, ObjectSection as _};
use typed_arena::Arena;
use crate::obj::{Section, SectionKind};
/// Parse line information from DWARF 2+ sections.
pub(crate) fn parse_line_info_dwarf2(
obj_file: &object::File,
sections: &mut [Section],
) -> Result<()> {
let arena_data = Arena::new();
let arena_relocations = Arena::new();
let endian = match obj_file.endianness() {
object::Endianness::Little => gimli::RunTimeEndian::Little,
object::Endianness::Big => gimli::RunTimeEndian::Big,
};
let dwarf = gimli::Dwarf::load(|id: gimli::SectionId| -> Result<_> {
load_file_section(id, obj_file, endian, &arena_data, &arena_relocations)
})
.context("loading DWARF sections")?;
let mut iter = dwarf.units();
if let Some(header) = iter.next().map_err(|e| gimli_error(e, "iterating over DWARF units"))? {
let unit = dwarf.unit(header).map_err(|e| gimli_error(e, "loading DWARF unit"))?;
if let Some(program) = unit.line_program.clone() {
let mut text_sections = sections.iter_mut().filter(|s| s.kind == SectionKind::Code);
let mut lines = text_sections.next().map(|section| &mut section.line_info);
let mut rows = program.rows();
while let Some((_header, row)) =
rows.next_row().map_err(|e| gimli_error(e, "loading program row"))?
{
if let (Some(line), Some(lines)) = (row.line(), &mut lines) {
lines.insert(row.address(), line.get() as u32);
}
if row.end_sequence() {
// The next row is the start of a new sequence, which means we must
// advance to the next .text section.
lines = text_sections.next().map(|section| &mut section.line_info);
}
}
}
}
if iter.next().map_err(|e| gimli_error(e, "checking for next unit"))?.is_some() {
log::warn!("Multiple units found in DWARF data, only processing the first");
}
Ok(())
}
#[derive(Debug, Default)]
struct RelocationMap(object::read::RelocationMap);
impl RelocationMap {
fn add(&mut self, file: &object::File, section: &object::Section) {
for (offset, relocation) in section.relocations() {
if let Err(e) = self.0.add(file, offset, relocation) {
log::error!(
"Relocation error for section {} at offset 0x{:08x}: {}",
section.name().unwrap(),
offset,
e
);
}
}
}
}
impl gimli::read::Relocate for &'_ RelocationMap {
fn relocate_address(&self, offset: usize, value: u64) -> gimli::Result<u64> {
Ok(self.0.relocate(offset as u64, value))
}
fn relocate_offset(&self, offset: usize, value: usize) -> gimli::Result<usize> {
<usize as gimli::ReaderOffset>::from_u64(self.0.relocate(offset as u64, value as u64))
}
}
type Relocate<'a, R> = gimli::RelocateReader<R, &'a RelocationMap>;
fn load_file_section<'input, 'arena, Endian: gimli::Endianity>(
id: gimli::SectionId,
file: &object::File<'input>,
endian: Endian,
arena_data: &'arena Arena<Cow<'input, [u8]>>,
arena_relocations: &'arena Arena<RelocationMap>,
) -> Result<Relocate<'arena, gimli::EndianSlice<'arena, Endian>>> {
let mut relocations = RelocationMap::default();
let data = match file.section_by_name(id.name()) {
Some(ref section) => {
relocations.add(file, section);
section.uncompressed_data()?
}
// Use a non-zero capacity so that `ReaderOffsetId`s are unique.
None => Cow::Owned(Vec::with_capacity(1)),
};
let data_ref = arena_data.alloc(data);
let section = gimli::EndianSlice::new(data_ref, endian);
let relocations = arena_relocations.alloc(relocations);
Ok(Relocate::new(section, relocations))
}
#[inline]
fn gimli_error(e: gimli::Error, context: &str) -> anyhow::Error {
anyhow::anyhow!("gimli error {context}: {e:?}")
}
+2
View File
@@ -1,3 +1,5 @@
#[cfg(feature = "dwarf")]
mod dwarf2;
pub mod read;
pub mod split_meta;
+35 -67
View File
@@ -122,7 +122,7 @@ fn map_symbols(
}
// Infer symbol sizes for 0-size symbols
infer_symbol_sizes(&mut symbols, sections);
infer_symbol_sizes(arch, &mut symbols, sections)?;
Ok((symbols, symbol_indices))
}
@@ -136,7 +136,7 @@ fn is_local_label(symbol: &Symbol) -> bool {
&& LABEL_PREFIXES.iter().any(|p| symbol.name.starts_with(p))
}
fn infer_symbol_sizes(symbols: &mut [Symbol], sections: &[Section]) {
fn infer_symbol_sizes(arch: &dyn Arch, symbols: &mut [Symbol], sections: &[Section]) -> Result<()> {
// Create a sorted list of symbol indices by section
let mut symbols_with_section = Vec::<usize>::with_capacity(symbols.len());
for (i, symbol) in symbols.iter().enumerate() {
@@ -206,18 +206,13 @@ fn infer_symbol_sizes(symbols: &mut [Symbol], sections: &[Section]) {
iter_idx += 1;
};
let section = &sections[section_idx];
let mut next_address =
let next_address =
next_symbol.map(|s| s.address).unwrap_or_else(|| section.address + section.size);
if section.kind == SectionKind::Code {
// For functions, trim any trailing 4-byte zeroes from the end (padding, nops)
while next_address > symbol.address + 4
&& let Some(data) = section.data_range(next_address - 4, 4)
&& data == [0u8; 4]
{
next_address -= 4;
}
}
let new_size = next_address.saturating_sub(symbol.address);
let new_size = if section.kind == SectionKind::Code {
arch.infer_function_size(symbol, section, next_address)?
} else {
next_address.saturating_sub(symbol.address)
};
if new_size > 0 {
let symbol = &mut symbols[symbol_idx];
symbol.size = new_size;
@@ -234,6 +229,7 @@ fn infer_symbol_sizes(symbols: &mut [Symbol], sections: &[Section]) {
}
}
}
Ok(())
}
fn map_sections(
@@ -554,12 +550,11 @@ fn perform_data_flow_analysis(obj: &mut Object, config: &DiffObjConfig) -> Resul
}
// Optional full data flow analysis
if config.analyze_data_flow {
if let Some(flow_result) =
if config.analyze_data_flow
&& let Some(flow_result) =
obj.arch.data_flow_analysis(obj, symbol, code, &section.relocations)
{
generated_flow_results.push((symbol.clone(), flow_result));
}
{
generated_flow_results.push((symbol.clone(), flow_result));
}
}
}
@@ -582,6 +577,28 @@ fn parse_line_info(
obj_data: &[u8],
) -> Result<()> {
// DWARF 1.1
if let Err(e) = parse_line_info_dwarf1(obj_file, sections) {
log::warn!("Failed to parse DWARF 1.1 line info: {e}");
}
// DWARF 2+
#[cfg(feature = "dwarf")]
if let Err(e) = super::dwarf2::parse_line_info_dwarf2(obj_file, sections) {
log::warn!("Failed to parse DWARF 2+ line info: {e}");
}
// COFF
if let object::File::Coff(coff) = obj_file
&& let Err(e) = parse_line_info_coff(coff, sections, section_indices, obj_data)
{
log::warn!("Failed to parse COFF line info: {e}");
}
Ok(())
}
/// Parse .line section from DWARF 1.1 format.
fn parse_line_info_dwarf1(obj_file: &object::File, sections: &mut [Section]) -> Result<()> {
if let Some(section) = obj_file.section_by_name(".line") {
let data = section.uncompressed_data()?;
let mut reader: &[u8] = data.as_ref();
@@ -609,55 +626,6 @@ fn parse_line_info(
}
}
}
// DWARF 2+
#[cfg(feature = "dwarf")]
{
fn gimli_error(e: gimli::Error) -> anyhow::Error { anyhow::anyhow!("DWARF error: {e:?}") }
let dwarf_cow = gimli::DwarfSections::load(|id| {
Ok::<_, gimli::Error>(
obj_file
.section_by_name(id.name())
.and_then(|section| section.uncompressed_data().ok())
.unwrap_or(alloc::borrow::Cow::Borrowed(&[][..])),
)
})
.map_err(gimli_error)?;
let endian = match obj_file.endianness() {
object::Endianness::Little => gimli::RunTimeEndian::Little,
object::Endianness::Big => gimli::RunTimeEndian::Big,
};
let dwarf = dwarf_cow.borrow(|section| gimli::EndianSlice::new(section, endian));
let mut iter = dwarf.units();
if let Some(header) = iter.next().map_err(gimli_error)? {
let unit = dwarf.unit(header).map_err(gimli_error)?;
if let Some(program) = unit.line_program.clone() {
let mut text_sections = sections.iter_mut().filter(|s| s.kind == SectionKind::Code);
let mut lines = text_sections.next().map(|section| &mut section.line_info);
let mut rows = program.rows();
while let Some((_header, row)) = rows.next_row().map_err(gimli_error)? {
if let (Some(line), Some(lines)) = (row.line(), &mut lines) {
lines.insert(row.address(), line.get() as u32);
}
if row.end_sequence() {
// The next row is the start of a new sequence, which means we must
// advance to the next .text section.
lines = text_sections.next().map(|section| &mut section.line_info);
}
}
}
}
if iter.next().map_err(gimli_error)?.is_some() {
log::warn!("Multiple units found in DWARF data, only processing the first");
}
}
// COFF
if let object::File::Coff(coff) = obj_file {
parse_line_info_coff(coff, sections, section_indices, obj_data)?;
}
Ok(())
}
+1 -1
View File
@@ -48,7 +48,7 @@ pub fn align_data_to_4<W: std::io::Write + ?Sized>(
len: usize,
) -> std::io::Result<()> {
const ALIGN_BYTES: &[u8] = &[0; 4];
if len % 4 != 0 {
if !len.is_multiple_of(4) {
writer.write_all(&ALIGN_BYTES[..4 - len % 4])?;
}
Ok(())
@@ -507,116 +507,4 @@ expression: diff.instruction_rows
),
arg_diff: [],
},
InstructionDiffRow {
ins_ref: Some(
InstructionRef {
address: 88,
size: 1,
opcode: 465,
branch_dest: None,
},
),
kind: None,
branch_from: None,
branch_to: None,
arg_diff: [],
},
InstructionDiffRow {
ins_ref: Some(
InstructionRef {
address: 89,
size: 1,
opcode: 465,
branch_dest: None,
},
),
kind: None,
branch_from: None,
branch_to: None,
arg_diff: [],
},
InstructionDiffRow {
ins_ref: Some(
InstructionRef {
address: 90,
size: 1,
opcode: 465,
branch_dest: None,
},
),
kind: None,
branch_from: None,
branch_to: None,
arg_diff: [],
},
InstructionDiffRow {
ins_ref: Some(
InstructionRef {
address: 91,
size: 1,
opcode: 465,
branch_dest: None,
},
),
kind: None,
branch_from: None,
branch_to: None,
arg_diff: [],
},
InstructionDiffRow {
ins_ref: Some(
InstructionRef {
address: 92,
size: 1,
opcode: 465,
branch_dest: None,
},
),
kind: None,
branch_from: None,
branch_to: None,
arg_diff: [],
},
InstructionDiffRow {
ins_ref: Some(
InstructionRef {
address: 93,
size: 1,
opcode: 465,
branch_dest: None,
},
),
kind: None,
branch_from: None,
branch_to: None,
arg_diff: [],
},
InstructionDiffRow {
ins_ref: Some(
InstructionRef {
address: 94,
size: 1,
opcode: 465,
branch_dest: None,
},
),
kind: None,
branch_from: None,
branch_to: None,
arg_diff: [],
},
InstructionDiffRow {
ins_ref: Some(
InstructionRef {
address: 95,
size: 1,
opcode: 465,
branch_dest: None,
},
),
kind: None,
branch_from: None,
branch_to: None,
arg_diff: [],
},
]
@@ -29,11 +29,3 @@ expression: output
[(Address(76), Normal, 5), (Spacing(4), Normal, 0), (Opcode(".dword", 65534), Normal, 10), (BranchDest(41), Normal, 0), (Basic(" ~>"), Rotating(6), 0), (Eol, Normal, 0)]
[(Address(80), Normal, 5), (Spacing(4), Normal, 0), (Opcode(".dword", 65534), Normal, 10), (BranchDest(47), Normal, 0), (Basic(" ~>"), Rotating(7), 0), (Eol, Normal, 0)]
[(Address(84), Normal, 5), (Spacing(4), Normal, 0), (Opcode(".dword", 65534), Normal, 10), (BranchDest(53), Normal, 0), (Basic(" ~>"), Rotating(8), 0), (Eol, Normal, 0)]
[(Address(88), Normal, 5), (Spacing(4), Normal, 0), (Opcode("nop", 465), Normal, 10), (Eol, Normal, 0)]
[(Address(89), Normal, 5), (Spacing(4), Normal, 0), (Opcode("nop", 465), Normal, 10), (Eol, Normal, 0)]
[(Address(90), Normal, 5), (Spacing(4), Normal, 0), (Opcode("nop", 465), Normal, 10), (Eol, Normal, 0)]
[(Address(91), Normal, 5), (Spacing(4), Normal, 0), (Opcode("nop", 465), Normal, 10), (Eol, Normal, 0)]
[(Address(92), Normal, 5), (Spacing(4), Normal, 0), (Opcode("nop", 465), Normal, 10), (Eol, Normal, 0)]
[(Address(93), Normal, 5), (Spacing(4), Normal, 0), (Opcode("nop", 465), Normal, 10), (Eol, Normal, 0)]
[(Address(94), Normal, 5), (Spacing(4), Normal, 0), (Opcode("nop", 465), Normal, 10), (Eol, Normal, 0)]
[(Address(95), Normal, 5), (Spacing(4), Normal, 0), (Opcode("nop", 465), Normal, 10), (Eol, Normal, 0)]

Some files were not shown because too many files have changed in this diff Show More