maekawatoshiki / qcc Goto Github PK

View Code? Open in Web Editor NEW

43.0 6.0 2.0 244 KB

qcc is a small toy C compiler

License: MIT License

Makefile 0.63% C++ 85.12% C 14.07% Shell 0.17%

compiler llvm cpp11 cpp toy

qcc's Introduction

QCC

qcc is a small toy compiler for C language.
waiting for pull-req!

FEATURES

qcc generates bitcode from C source code with LLVM. so we have to convert bitcode to native.
the shell script './qcc.sh' does it. or do yourself such as below:

$ ./qcc c.c -o c.bc # qcc generates c.bc
$ llc-3.8 c.bc # c.bc -> c.s
$ clang c.s # c.s -> a.out

BUILD

used tools: clang, llvm-3.8
if you are using Ubuntu or Debian, maybe you can see what you have to apt-get install if you look at Dockerfile.

$ make
$ make test
$ ./qcc.sh <SOURCE CODE>

TRY ON DOCKER

follow below

docker build -t="qcc" .
docker run -it "qcc" .

qcc's People

Contributors

Stargazers

Watchers

Forkers

kariya-mitsuru airog

qcc's Issues

SEGV on global string

QCC runs into SEGV if any string literal exists in global scope.

Sample code:

const char *s = "Hello World";

Stack trace from clang address sanitizer:

    #0 0x102e06970 in llvm::IRBuilderBase::CreateGlobalString(llvm::StringRef, llvm::Twine const&, unsigned int) (qcc+0x10016b970)
    #1 0x102cb189f in Codegen::statement(StringAST*) (qcc+0x10001689f)
    #2 0x102ca55d3 in Codegen::statement(AST*) (qcc+0x10000a5d3)
    #3 0x102cad87c in Codegen::statement(VarDeclarationAST*) (qcc+0x10001287c)
    #4 0x102ca55c4 in Codegen::statement(AST*) (qcc+0x10000a5c4)
    #5 0x102ca4ea0 in Codegen::run(std::__1::vector<AST*, std::__1::allocator<AST*> >, std::__1::basic_string<char, std::__1::char_traits<char>, std::__1::allocator<char> >, bool) (qcc+0x100009ea0)
    #6 0x102d09f70 in QCC::run(std::__1::basic_string<char, std::__1::char_traits<char>, std::__1::allocator<char> >) (qcc+0x10006ef70)
    #7 0x102d0a68f in QCC::run() (qcc+0x10006f68f)
    #8 0x102cf5ecd in main (qcc+0x10005aecd)
    #9 0x7fffac57d254 in start (libdyld.dylib+0x5254)

Migration to newer LLVM

I heard you were planning to migrate LLVM 3.8 on #8 (comment).

How is it going?

Since homebrew (macOS community-driven package manager) only support LLVM 3.7 and higher now, I cannot try qcc.

More accurate implementation for array subscript operator

According to C standard, this code is well-formed but QCC hasn't support some part yet.

int a[] = {1,2,3,4,5};
a[0]; // of course this has no problem.
0[a]; // this is also good, though QCC hasn't supported yet.
'\0'[a] // even this is well-formed (at least under ASCII), though QCC hasn't supported yet.

According to C89, subscript operator E1[E2] is identical to (*(E1+(E2))).
According to C99 and C11, it is identical to (*((E1)+(E2))).

Replacing array expression might be needed.

Support function-like macros with no arguments

FreeBSD's sys/cdefs.h (needed by stdio.h) contains a macro:

#define __unreachable() ((void)0)

This is currently not supported and QCC claims expected ','.

FYI function-like macros with no argument is sometimes preferred to regular macros.
See http://stackoverflow.com/questions/4251005/ .

Escape sequences aren't handled correctly

Any escape sequences (e.g. \n,\r) are output as typical characters.

Test code:

#include <stdio.h>
int main(){
  printf("hello world\n");
}

Expected output:

hello world

Actual output:

hello world\n

maekawatoshiki / qcc Goto Github PK

qcc's Introduction

QCC

FEATURES

BUILD

TRY ON DOCKER

qcc's People

Contributors

Stargazers

Watchers

Forkers

qcc's Issues

SEGV on global string

Migration to newer LLVM

More accurate implementation for array subscript operator

Support function-like macros with no arguments

Escape sequences aren't handled correctly

Recommend Projects

React

Vue.js

Typescript

TensorFlow

Django

Laravel

D3

Recommend Topics

javascript

web

server

Machine learning

Visualization

Game

Recommend Org

Facebook

Microsoft

Google

Alibaba

D3

Tencent