Remove decomp-permuter (#447)

* Nuke decomp permuter

* Add decomp permuter and mips2c to gitignore
This commit is contained in:
Anghelo Carvajal
2021-11-11 12:46:18 -03:00
committed by GitHub
parent 2e5c142f3b
commit c018e83d36
88 changed files with 2 additions and 12846 deletions
+2
View File
@@ -33,6 +33,8 @@ tools/ido_recomp/* binary
ctx.c
graphs/
*.c.m2c
tools/decomp-permuter/
tools/mips_to_c/
# Assets
*.png
-45
View File
@@ -1,45 +0,0 @@
name: Systray
on:
push:
branches: [ main ]
paths:
- 'src/net/cmd/systray/*'
pull_request:
branches: [ main ]
paths:
- 'src/net/cmd/systray/*'
jobs:
build:
name: Build on ${{ matrix.os }}
runs-on: ${{ matrix.os }}
strategy:
matrix:
include:
- os: ubuntu-16.04
binary: permuter-systray-linux
- os: windows-latest
binary: permuter-systray.exe
- os: macos-latest
binary: permuter-systray-macos
steps:
- uses: actions/checkout@main
- name: Install gtk3
if: ${{ matrix.os == 'ubuntu-16.04' }}
run: sudo apt-get install libgtk-3-dev libappindicator3-dev
- name: Setup Go environment
uses: actions/setup-go@v2.1.3
- name: Build
run: go build -o ${{ matrix.binary }} -ldflags "-s -w" tray.go
working-directory: src/net/cmd/systray/
- name: Upload artifact
uses: actions/upload-artifact@v2
with:
name: ${{ matrix.binary }}
path: src/net/cmd/systray/${{ matrix.binary }}
-11
View File
@@ -1,11 +0,0 @@
*.o
*.s
*.c
*.py[cod]
.mypy_cache/
.cache/
__pycache__/
!test/*.c
/nonmatchings
.vscode/
pah.conf
-12
View File
@@ -1,12 +0,0 @@
; DO NOT EDIT (unless you know what you are doing)
;
; This subdirectory is a git "subrepo", and this file is maintained by the
; git-subrepo command. See https://github.com/git-commands/git-subrepo#readme
;
[subrepo]
remote = https://github.com/simonlindholm/decomp-permuter.git
branch = main
commit = a20bac9422b6d8adbf7c06473c2ae3c3fee16be5
parent = 2668eec556c01fa2f4c16a203c93c208dc03e639
method = merge
cmdver = 0.4.3
@@ -1,6 +0,0 @@
repos:
- repo: https://github.com/psf/black
rev: 20.8b1
hooks:
- id: black
language_version: python3.6
-21
View File
@@ -1,21 +0,0 @@
MIT License
Copyright (c) 2019 Simon Lindholm and contributors
Permission is hereby granted, free of charge, to any person obtaining a copy
of this software and associated documentation files (the "Software"), to deal
in the Software without restriction, including without limitation the rights
to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
copies of the Software, and to permit persons to whom the Software is
furnished to do so, subject to the following conditions:
The above copyright notice and this permission notice shall be included in all
copies or substantial portions of the Software.
THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
SOFTWARE.
-120
View File
@@ -1,120 +0,0 @@
# Decomp permuter
Automatically permutes C files to better match a target binary. The permuter has two modes of operation:
- Random: purely at random, introduce temporary variables for values, change types, put statements on the same line...
- Manual: test all combinations of user-specified variations, using macros like `PERM_GENERAL(a = b ? c : d;, if (b) a = c; else a = d;)` to try both specified alternatives.
The modes can also be combined, by using the `PERM_RANDOMIZE` macro.
[<img src="https://asciinema.org/a/232846.svg" height="300">](https://asciinema.org/a/232846)
The main target for the tool is MIPS code compiled by old compilers (IDO, possibly GCC).
Getting it to work on other architectures shouldn't be too hard, however.
https://github.com/laqieer/decomp-permuter-arm has an ARM port.
## Usage
`./permuter.py directory/` runs the permuter; see below for the meaning of the directory.
Pass `-h` to see possible flags. `-j` is suggested (enables multi-threaded mode).
You'll first need to install a couple of prerequisites: `python3 -m pip install pycparser pynacl toml` (also `dataclasses` if on Python 3.6 or below)
The permuter expects as input one or more directory containing:
- a .c file with a single function,
- a .o file to match,
- a .sh file that compiles the .c file.
For projects with a properly configured makefile, you should be able to set these up by running
```
./import.py <path/to/file.c> <path/to/file.s>
```
where file.c contains the function to be permuted, and file.s is its assembly in a self-contained file.
Otherwise, see USAGE.md for more details.
For projects using Ninja instead of Make, add a `permuter_settings.toml` in the root or `tools/` directory of the project:
```toml
build_system = "ninja"
```
Then `import.py` should work as expected if `build.ninja` is at the root of the project.
The .c file may be modified with any of the following macros which affect manual permutation:
- `PERM_GENERAL(a, b, ...)` expands to any of `a`, `b`, ...
- `PERM_VAR(a, b)` sets the meta-variable `a` to `b`, `PERM_VAR(a)` expands to the meta-variable `a`.
- `PERM_RANDOMIZE(code)` expands to `code`, but allows randomization within that region. Multiple regions may be specified.
- `PERM_LINESWAP(lines)` expands to a permutation of the ordered set of non-whitespace lines (split by `\n`). Each line must contain zero or more complete C statements. (For incomplete statements use `PERM_LINESWAP_TEXT`, which is slower because it has to repeatedly parse C code.)
- `PERM_INT(lo, hi)` expands to an integer between `lo` and `hi` (which must be constants).
- `PERM_IGNORE(code)` expands to `code`, without passing it through the C parser library (pycparser)/randomizer. This can be used to avoid parse errors for non-standard C, e.g. `asm` blocks.
- `PERM_PRETEND(code)` expands to `code` for the purpose of the C parser/randomizer, but gets removed afterwards. This can be used together with `PERM_IGNORE` to enable the permuter to deal with input it isn't designed for (e.g. inline functions, C++, non-code).
- `PERM_ONCE([key,] code)` expands to either `code` or to nothing, such that each unique key gets expanded exactly once. `key` defaults to `code`. For example, `PERM_ONCE(a;) b; PERM_ONCE(a;)` expands to either `a; b;` or `b; a;`.
Arguments are split by a commas, exluding commas inside parenthesis. `(,)` is a special escape sequence that resolves to `,`.
Nested macros are allowed, so e.g.
```
PERM_VAR(delayed, )
PERM_GENERAL(stmt;, PERM_VAR(delayed, stmt;))
...
PERM_VAR(delayed)
```
is an alternative way of writing `PERM_ONCE`.
## permuter@home
The permuter supports a distributed mode, where people can donate processor power to your permuter runs to speed them up.
To use this, pass `-J` when running `permuter.py` and follow the instructions.
You will need to be granted access by someone who is already connected to a permuter network.
To allow others to use your computer for permuter runs, do the following:
- install Docker (used for sandboxing and to ensure a consistent environment)
- if on Linux, add yourself to the Docker group: `sudo usermod -aG docker $USER`
- install required packages: `python3 -m pip install docker`
- open a terminal, and run `./pah.py run-server` to start the server.
There are a few required arguments (e.g. how many cores to use), see `--help` for more details.
Please be aware that being in the Docker group implies (password-less) sudo rights.
You can avoid that for your personal account by running the permuter under a separate user.
Unfortunately, there is currently no way to run a sandboxed permuter server without sudo rights. 😢
Anyone who is granted access to permuter@home can run a server.
To set up a new permuter network, see [src/net/controller/README.md](./src/net/controller/README.md).
## FAQ
**What do the scores mean?** The scores are computed by taking diffs of objdump'd .o
files, and giving different penalties for lines that are the same/use the same
instruction/are reordered/don't match at all. 0 means the function matches fully.
Stack positions are ignored unless --stack-diffs is passed (but beware that the
permuter is currently quite bad at resolving stack differences). For more details,
see scorer.py. It's far from a perfect system, and should probably be tweaked to
look at e.g. the register diff graph.
**What sort of non-matchings are the permuter good at?** It's generally best towards
the end, when mostly regalloc changes remain. If there are reorderings or functional
changes, it's often easy to resolve those by hand, and neither the scorer nor the
randomizer tends to play well with them.
**Should I use this instead of trying to match code by hand?** No, but it can be a good
complement. PERM macros can be used to quickly test lots of variations of a function at
once, in cases where there are interactions between several parts of a function.
The randomization mode often finds lots of nonsensical changes that improve regalloc
"by accident"; it's up to you to pick out the ones that look sensible. If none do,
it can still be useful to know which parts of the function need to be changed to get the
code nearer to matching. Having made one of the improvements, and the function can then be
permuted again, to find further possible improvements.
## Helping out
There's tons of room for helping out with the permuter!
Many more randomization passes could be added, the scoring function is far from optimal,
the permuter could be made easier to use, etc. etc. The GitHub Issues list has some ideas.
Ideally, `mypy permuter.py` and `./run-tests.sh` should succeed with no errors, and files
formatted with `black`. To setup a pre-commit hook for black, run:
```
pip install pre-commit black
pre-commit install
```
PRs that skip this are still welcome, however.
-25
View File
@@ -1,25 +0,0 @@
This file describes how to manually set up a directory for use with the permuter.
**You probably don't need to do this!** In normal circumstances, `./import.py`
does all this for you. See README.md for more details.
* create a directory that will contain all of the input files for the invokation
* put a compile command into `<dir>/compile.sh` (see e.g. `compile_example.sh`; it will be invoked as `./compile.sh input.c -o output.o`)
* `gcc -E -P -I header_dir -D'__attribute__(x)=' orig_c_file.c > <dir>/base.c`
* `python3 strip_other_fns.py <dir>/base.c func_name`
* put asm for `func_name` into `<dir>/target.s`, with the following header:
```asm
.set noat
.set noreorder
.set gp=64
.macro glabel label
.global \label
.type \label, @function
\label:
.endm
```
* `mips-linux-gnu-as -march=vr4300 -mabi=32 <dir>/target.s -o <dir>/target.o`
* optional sanity checks:
- `<dir>/compile.sh <dir>/base.c -o <dir>/base.o`
- `./diff.sh <dir>/target.o <dir>/base.o`
* `./permuter.py <dir>`
-2
View File
@@ -1,2 +0,0 @@
#!/bin/bash
mips-linux-gnu-gcc -O2 "$@"
-17
View File
@@ -1,17 +0,0 @@
#!/bin/bash
if [[ $# < 2 ]]; then
echo "Usage: $0 orig.o new.o [flags]"
exit 1
fi
if [ ! -f $1 -o ! -f $2 ]; then
echo Source files not readable
exit 1
fi
INPUT1="$1"
INPUT2="$2"
shift
shift
wdiff -n <(python3 ./src/objdump.py "$INPUT1" "$@") <(python3 ./src/objdump.py "$INPUT2" "$@") | colordiff | less -Ric
File diff suppressed because it is too large Load Diff
-27
View File
@@ -1,27 +0,0 @@
[mypy]
check_untyped_defs = True
disallow_any_generics = False
disallow_incomplete_defs = True
disallow_subclassing_any = True
disallow_untyped_calls = True
disallow_untyped_decorators = True
disallow_untyped_defs = True
no_implicit_optional = True
warn_redundant_casts = True
warn_return_any = True
warn_unused_ignores = True
mypy_path = stubs
python_version = 3.7
files = import.py, pah.py, permuter.py, src/net/evaluator.py
[mypy-nacl.*]
ignore_missing_imports = True
[mypy-pystray.*]
ignore_missing_imports = True
[mypy-docker.*]
ignore_missing_imports = True
[mypy-PIL.*]
ignore_missing_imports = True
-4
View File
@@ -1,4 +0,0 @@
#!/usr/bin/env python3
from src.net.cmd.main import main
main()
-5
View File
@@ -1,5 +0,0 @@
#!/usr/bin/env python3
from src.main import main
if __name__ == "__main__":
main()
@@ -1,9 +0,0 @@
# Optional configuration file for import.py. Put it in the root or in tools/
# of the repo you are importing from.
build_system = "ninja"
[preserve_macros]
"g[DS]P.*" = "void"
"gDma.*" = "void"
"_SHIFTL" = "unsigned int"
-3
View File
@@ -1,3 +0,0 @@
#!/bin/sh
python3 -m unittest discover -s test/
# python3 -m pytest test/
-13
View File
@@ -1,13 +0,0 @@
#!/bin/bash
if [[ $1 < 2 ]]; then
echo "Usage: $0 output_dir"
echo "Ex: $0 nonmatchings/func_80000000"
exit 1
fi
if [[ ! -d $1 ]]; then
echo "Argument must be a directory"
exit 1
fi
find $1 -name score.txt -exec echo -n {}\ \; -exec cat {} \; | sort -rnk2
-303
View File
@@ -1,303 +0,0 @@
"""Functions and classes for dealing with types in a C AST.
They make a number of simplifying assumptions:
- const and volatile doesn't matter.
- arithmetic promotes all int-like types to 'int'.
- no two variables can have the same name, even across functions.
For the purposes of the randomizer these restrictions are acceptable."""
from dataclasses import dataclass, field
from typing import Union, Dict, Set, List
from pycparser import c_ast
from pycparser.c_ast import ArrayDecl, TypeDecl, PtrDecl, FuncDecl, IdentifierType
Type = Union[PtrDecl, ArrayDecl, TypeDecl, FuncDecl]
SimpleType = Union[PtrDecl, TypeDecl]
StructUnion = Union[c_ast.Struct, c_ast.Union]
@dataclass
class TypeMap:
typedefs: Dict[str, Type] = field(default_factory=dict)
fn_ret_types: Dict[str, Type] = field(default_factory=dict)
var_types: Dict[str, Type] = field(default_factory=dict)
struct_defs: Dict[str, StructUnion] = field(default_factory=dict)
def basic_type(name: Union[str, List[str]]) -> TypeDecl:
names = [name] if isinstance(name, str) else name
idtype = IdentifierType(names=names)
return TypeDecl(declname=None, quals=[], type=idtype)
def pointer(type: Type) -> Type:
return PtrDecl(quals=[], type=type)
def resolve_typedefs(type: Type, typemap: TypeMap) -> Type:
while (
isinstance(type, TypeDecl)
and isinstance(type.type, IdentifierType)
and len(type.type.names) == 1
and type.type.names[0] in typemap.typedefs
):
type = typemap.typedefs[type.type.names[0]]
return type
def pointer_decay(type: Type, typemap: TypeMap) -> SimpleType:
real_type = resolve_typedefs(type, typemap)
if isinstance(real_type, ArrayDecl):
return PtrDecl(quals=[], type=real_type.type)
if isinstance(real_type, FuncDecl):
return PtrDecl(quals=[], type=type)
if isinstance(real_type, TypeDecl) and isinstance(real_type.type, c_ast.Enum):
return basic_type("int")
assert not isinstance(
type, (ArrayDecl, FuncDecl)
), "resolve_typedefs can't hide arrays/functions"
return type
def get_decl_type(decl: c_ast.Decl) -> Type:
"""For a Decl that declares a variable (and not just a struct/union/enum),
return its type."""
assert decl.name is not None
assert isinstance(decl.type, (PtrDecl, ArrayDecl, FuncDecl, TypeDecl))
return decl.type
def deref_type(type: Type, typemap: TypeMap) -> Type:
type = resolve_typedefs(type, typemap)
assert isinstance(type, (ArrayDecl, PtrDecl)), "dereferencing non-pointer"
return type.type
def struct_member_type(struct: StructUnion, field_name: str, typemap: TypeMap) -> Type:
if not struct.decls:
assert (
struct.name in typemap.struct_defs
), f"Accessing field {field_name} of undefined struct {struct.name}"
struct = typemap.struct_defs[struct.name]
assert struct.decls, "struct_defs never points to an incomplete type"
for decl in struct.decls:
if isinstance(decl, c_ast.Decl):
if decl.name == field_name:
return get_decl_type(decl)
if decl.name == None and isinstance(decl.type, (c_ast.Struct, c_ast.Union)):
try:
return struct_member_type(decl.type, field_name, typemap)
except AssertionError:
pass
assert False, f"No field {field_name} in struct {struct.name}"
def expr_type(node: c_ast.Node, typemap: TypeMap) -> Type:
def rec(sub_expr: c_ast.Node) -> Type:
return expr_type(sub_expr, typemap)
if isinstance(node, c_ast.Assignment):
return rec(node.lvalue)
if isinstance(node, c_ast.StructRef):
lhs_type = rec(node.name)
if node.type == "->":
lhs_type = deref_type(lhs_type, typemap)
struct_type = resolve_typedefs(lhs_type, typemap)
assert isinstance(struct_type, TypeDecl)
assert isinstance(
struct_type.type, (c_ast.Struct, c_ast.Union)
), f"struct deref of non-struct {struct_type.declname}"
return struct_member_type(struct_type.type, node.field.name, typemap)
if isinstance(node, c_ast.Cast):
return node.to_type.type
if isinstance(node, c_ast.Constant):
if node.type == "string":
return pointer(basic_type("char"))
if node.type == "char":
return basic_type("int")
return basic_type(node.type.split(" "))
if isinstance(node, c_ast.ID):
return typemap.var_types[node.name]
if isinstance(node, c_ast.UnaryOp):
if node.op in ["p++", "p--", "++", "--"]:
return rec(node.expr)
if node.op == "&":
return pointer(rec(node.expr))
if node.op == "*":
subtype = rec(node.expr)
return deref_type(subtype, typemap)
if node.op in ["-", "+"]:
subtype = pointer_decay(rec(node.expr), typemap)
if allowed_basic_type(subtype, typemap, ["double"]):
return basic_type("double")
if allowed_basic_type(subtype, typemap, ["float"]):
return basic_type("float")
if node.op in ["sizeof", "-", "+", "~", "!"]:
return basic_type("int")
assert False, f"unknown unary op {node.op}"
if isinstance(node, c_ast.BinaryOp):
lhs_type = pointer_decay(rec(node.left), typemap)
rhs_type = pointer_decay(rec(node.right), typemap)
if node.op in [">>", "<<"]:
return lhs_type
if node.op in ["<", "<=", ">", ">=", "==", "!=", "&&", "||"]:
return basic_type("int")
if node.op in "&|^%":
return basic_type("int")
real_lhs = resolve_typedefs(lhs_type, typemap)
real_rhs = resolve_typedefs(rhs_type, typemap)
if node.op in "+-":
lptr = isinstance(real_lhs, PtrDecl)
rptr = isinstance(real_rhs, PtrDecl)
if lptr or rptr:
if lptr and rptr:
assert node.op != "+", "pointer + pointer"
return basic_type("int")
if lptr:
return lhs_type
assert node.op == "+", "int - pointer"
return rhs_type
if node.op in "*/+-":
assert isinstance(real_lhs, TypeDecl)
assert isinstance(real_rhs, TypeDecl)
assert isinstance(real_lhs.type, IdentifierType)
assert isinstance(real_rhs.type, IdentifierType)
if "double" in real_lhs.type.names + real_rhs.type.names:
return basic_type("double")
if "float" in real_lhs.type.names + real_rhs.type.names:
return basic_type("float")
return basic_type("int")
if isinstance(node, c_ast.FuncCall):
expr = node.name
if isinstance(expr, c_ast.ID):
if expr.name not in typemap.fn_ret_types:
raise Exception(f"Called function {expr.name} is missing a prototype")
return typemap.fn_ret_types[expr.name]
else:
fptr_type = resolve_typedefs(rec(expr), typemap)
if isinstance(fptr_type, PtrDecl):
fptr_type = fptr_type.type
fptr_type = resolve_typedefs(fptr_type, typemap)
assert isinstance(fptr_type, FuncDecl), "call to non-function"
return fptr_type.type
if isinstance(node, c_ast.ExprList):
return rec(node.exprs[-1])
if isinstance(node, c_ast.ArrayRef):
subtype = rec(node.name)
return deref_type(subtype, typemap)
if isinstance(node, c_ast.TernaryOp):
return rec(node.iftrue)
assert False, f"Unknown expression node type: {node}"
def decayed_expr_type(expr: c_ast.Node, typemap: TypeMap) -> SimpleType:
return pointer_decay(expr_type(expr, typemap), typemap)
def same_type(
type1: Type, type2: Type, typemap: TypeMap, allow_similar: bool = False
) -> bool:
while True:
type1 = resolve_typedefs(type1, typemap)
type2 = resolve_typedefs(type2, typemap)
if isinstance(type1, ArrayDecl) and isinstance(type2, ArrayDecl):
type1 = type1.type
type2 = type2.type
continue
if isinstance(type1, PtrDecl) and isinstance(type2, PtrDecl):
type1 = type1.type
type2 = type2.type
continue
if isinstance(type1, TypeDecl) and isinstance(type2, TypeDecl):
sub1 = type1.type
sub2 = type2.type
if isinstance(sub1, c_ast.Struct) and isinstance(sub2, c_ast.Struct):
return sub1.name == sub2.name
if isinstance(sub1, c_ast.Union) and isinstance(sub2, c_ast.Union):
return sub1.name == sub2.name
if (
allow_similar
and isinstance(sub1, (IdentifierType, c_ast.Enum))
and isinstance(sub2, (IdentifierType, c_ast.Enum))
):
# All int-ish types are similar (except void, but whatever)
return True
if isinstance(sub1, c_ast.Enum) and isinstance(sub2, c_ast.Enum):
return sub1.name == sub2.name
if isinstance(sub1, IdentifierType) and isinstance(sub2, IdentifierType):
return sorted(sub1.names) == sorted(sub2.names)
return False
def allowed_basic_type(
type: SimpleType, typemap: TypeMap, allowed_types: List[str]
) -> bool:
"""Check if a type resolves to a basic type with one of the allowed_types
keywords in it."""
base_type = resolve_typedefs(type, typemap)
if not isinstance(base_type, c_ast.TypeDecl):
return False
if not isinstance(base_type.type, c_ast.IdentifierType):
return False
if all(x not in base_type.type.names for x in allowed_types):
return False
return True
def build_typemap(ast: c_ast.FileAST) -> TypeMap:
ret = TypeMap()
for item in ast.ext:
if isinstance(item, c_ast.Typedef):
ret.typedefs[item.name] = item.type
if isinstance(item, c_ast.FuncDef):
assert item.decl.name is not None, "cannot define anonymous function"
assert isinstance(item.decl.type, FuncDecl)
ret.fn_ret_types[item.decl.name] = item.decl.type.type
if isinstance(item, c_ast.Decl) and isinstance(item.type, FuncDecl):
assert item.name is not None, "cannot define anonymous function"
ret.fn_ret_types[item.name] = item.type.type
defined_function_decls: Set[c_ast.Decl] = set()
class Visitor(c_ast.NodeVisitor):
def visit_Struct(self, struct: c_ast.Struct) -> None:
if struct.decls and struct.name is not None:
ret.struct_defs[struct.name] = struct
# Do not visit decls of this struct
def visit_Union(self, union: c_ast.Union) -> None:
if union.decls and union.name is not None:
ret.struct_defs[union.name] = union
# Do not visit decls of this union
def visit_Decl(self, decl: c_ast.Decl) -> None:
if decl.name is not None:
ret.var_types[decl.name] = get_decl_type(decl)
if not isinstance(decl.type, FuncDecl) or decl in defined_function_decls:
# Do not visit declarations in parameter lists of functions
# other than our own.
self.visit(decl.type)
def visit_Enumerator(self, enumerator: c_ast.Enumerator) -> None:
ret.var_types[enumerator.name] = basic_type("int")
def visit_FuncDef(self, fn: c_ast.FuncDef) -> None:
if fn.decl.name is not None:
ret.var_types[fn.decl.name] = get_decl_type(fn.decl)
defined_function_decls.add(fn.decl)
self.generic_visit(fn)
Visitor().visit(ast)
return ret
def set_decl_name(decl: c_ast.Decl) -> None:
name = decl.name
assert name is not None
type = get_decl_type(decl)
while not isinstance(type, TypeDecl):
type = type.type
type.declname = name
File diff suppressed because it is too large Load Diff

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