# CC - a self-hosting, bootstrappable, minimal C compiler ## Introduction On the never-ending quest of a minimal system I found Swieros and C4 (the C compiler in 4 functions). Inspired and intrigued I started to implement my own. For abaos (a small operating system of mine, also in C) I cloned the minimal C library, so we can build a freestanding version of C4. C4 serves as a test whether my own CC is minimal enough and doesn't use silly functions. Additionally C4 as well as CC are compiled both in a (on Linux) hosted version and a freestanding version. We use a series of compilers like gcc, clang, tcc and pcc to make sure that we are not using silly C constructs. In order to be able to port easily we make almost no use of system calls, the ones we need are: - brk: for malloc/free, change the start address of the heap segment of the process, if the OS only assigns a single static space, then brk results in a NOP. - exit: terminate the process, return does not always work in all combinations (for instance with pcc on Linux). Can be a NOP, we don't require any trickery as atext and we don't use buffering anywhere (for instance flushing stdout on exit). - read/write: read from stdin linearly, write to stdout linearly, this is essentially a model using an input and an output tape. Those two functions must really exist. This basically eliminates the need for a file system which we might not have during early bootstrapping. Similarly we simplify the C language to not use certain features which can cause trouble when bootstrapping: - variable arguments: though simple in principle (just some pointers into the stack if you use a stack for function parameters), it is not typesafe. And the only example in practice it's really heavily used for is in printf-like functions. - preprocessor: it needs a filesystem, we take this outside of the compiler by feeding it an (eventually) concatenated list of \*.c files. - two types: int and char, so we can interpret memory as words or as bytes. ## Local version of C4 The local version of C4 has the following adaoptions and extensions: - switch statement from the switch-and-structs branch, adapted c4 itself to use switch statements instead of if's (as in the switch-and-structs branch) - struct support from switch-and-structs - constants like EOF, EXIT\_SUCCESS, NULL - standard C block comments along to c++ end of line ones - negative enum initializers - do/while loops - more C functions like isspace, getc, strcmp - some simplified functions for printing like putstring, putint, putnl - strict C89 conformance, mainly use standard comment blocks, also removed some warnings - some casts around malloc and memset to fit to non-void freestanding-libc - converted printf to putstring/putint/putnl and some helper functions for error reporting like error() - removed all memory leaks - de-POSIX-ified, no open/read/close, use getchar from stdin only (don't assume the existence of a file system), this also means we had to create sort of an old style tape-file with FS markers to separate the files piped to c4. Note: only too late I discovered that there was a C5 version of the same compiler, which would maybe have served better as a basis. ## Examples ### Running on the host system using the hosts C compiler Compiled in either hosted (host libc) or freestanding (our own libc, currently IA-32 Linux kernel only syscalls): `./build.sh cc hostcc hosted d ./build.sh cc hostcc freestanding d ./cc \< test1.c \> test1.asm` Create a plain binary from the assembly code: `fasm test1.asm test1.bin` Disassemble it to verify it's correctness: `ndisasm -b32 -o1000000h -a test1.bin` You can choose gcc, clang, tcc or pcc as host compiler (hostcc). ### Running on the host in the C4 interpreter Running in C4 interpreter, again, the C4 program can be compiled in hosted or freestanding mode: `./build.sh c4 hostcc hosted d ./build.sh c4 hostcc freestanding d` Here again you can choose the host compiler for compiling C4. Then we have to create the standard input for C4 using: `echo -n -e "\034" \> EOF cat cc.c EOF hello.c | ./c4 cat c4.c EOF cc.c EOF hello.c | ./c4 cat c4.c4 EOF c4.c EOF cc.c EOF hello.c | ./c4` EOF contains the traditional FS (file separator) character in the ASCII character set. Every time c4/c4.c is invoked it reads exacly one input file up to the first FS character (or stops at the end of stdin). We can also use -s, or -d on every level as follows: `cat cc.c EOF hello.c | ./c4 -d` ## References Compiler construction in general: - "Compiler Construction"", Niklaus Wirth - https://github.com/DoctorWkt/acwj: a nice series on building a C compiler, step by step with lots of good explanations - https://www.engr.mun.ca/~theo/Misc/exp\_parsing.htm\#climbing, https://en.wikipedia.org/wiki/Operator-precedence\_parser\#Precedence\_climbing\_method - https://github.com/lotabout/write-a-C-interpreter/blob/master/tutorial/en/, tutorial based on C4 how to build a C interpreter, explains nicely details in C4. C4: - https://github.com/rswier/c4.git, C4 - C in four functions, Robert Swierczek, minimalistic C compiler running on an emulator on the IR, inspiration for this project - https://github.com/rswier/c4/blob/switch-and-structs/c4.c, c4 adaptions to provide switch and structs - https://github.com/EarlGray/c4: a X86 JIT version of c4 - https://github.com/jserv/amacc: based on C4, JIT or native code, for ARM, quite well documented, also very nice list of compiler resources on Github page Other minimal compilers and systems: - http://selfie.cs.uni-salzburg.at/: C\* self-hosting C compiler (also emulator, hypervisor) for RISCV, inspiration for what makes up a minimal C language - http://www.iro.umontreal.ca/~felipe/IFT2030-Automne2002/Complements/tinyc.c, Marc Feeley, really easy and much more readable, meant as educational compiler - https://github.com/rswier/swieros.git: c.c in swieros, Robert Swierczek Assembly: - https://github.com/felipensp/assembly/blob/master/x86/itoa.s, for putint (early debugging keyword) - https://baptiste-wicht.com/posts/2011/11/print-strings-integers-intel-assembly.htm (earldy debugging keyword)