Add HACKING.md.

Author: rui314

HACKING.md

@@ -0,0 +1,31 @@
+I accept small patches, but because this is my hobby project to
+learn about compilers, it's unlikely to accept large patches.
+That's in practice not going to be an issue, because
+if you are writing a large patch, that is your hobby project.
+You want to hack in your forked repository rather than
+taking time to send pull requests.
+
+# Memory management
+
+No memory management is a memory management scheme in 8cc.
+Memory regions allocated using malloc are never freed
+until the process terminates. That has greatly simplified
+the code and the APIs because 1) you can write code as if
+garbage collector were present, and 2) that design
+decision has eliminated use-after-free bugs entirely.
+
+Modern computers have gigs of memory. 8cc consumes
+only about 100MB to compile 10K lines of C source file.
+Compiler is not a long-running process.
+It will never run out of memory unless you give an
+unrealistically large source file.
+
+If we really need to free memory, we could use Boehm garbage
+collector. I don't see that need at this moment though.
+
+# Backend
+
+Backend is being rewritten. Once it's done, the current backend
+code will be discarded. The new backend models after the LLVM IR
+because the IR looks to be designed well. That's not going to be
+the same, though.

vector.c

@@ -3,25 +3,6 @@
 
 /*
  * Vectors are containers of void pointers that can change in size.
- *
- * Note on memory management:
- *
- * No memory management is a memory management scheme in 8cc.
- * Memory regions allocated using malloc are never freed
- * until the process terminates. That has greatly simplified
- * the code and the API because 1) when you want to write
- * a function returning a new heap-allocated value, you can
- * just malloc a new value and return that, and 2) that design
- * decision has eliminated use-after-free bugs entirely.
- *
- * Modern computers have gigs of memory. 8cc consumes
- * only about 100MB to compile 10K lines of C source file.
- * Compiler is not a long-running process.
- * It will never run out of memory unless you give an
- * unrealistically large source file.
- *
- * If we really need to free memory, we could use Boehm garbage
- * collector. I don't see that need at this moment though.
  */
 
 #include <assert.h>