mark and sweep GC without bitvectors works

lib/system/alloc.nim

@@ -319,7 +319,8 @@ iterator allObjects(m: TMemRegion): pointer {.inline.} =
         
         let size = c.size
         var a = cast[TAddress](addr(c.data))
-        while a <% c.acc:
+        let limit = a + c.acc
+        while a <% limit:
           yield cast[pointer](a)
           a = a +% size
       else:

lib/system/gc_ms.nim

@@ -13,7 +13,9 @@
 
 const
   InitialThreshold = 4*1024*1024 # X MB because marking&sweeping is slow
-  withBitvectors = defined(gcUseBitvectors)
+  withBitvectors = defined(gcUseBitvectors) 
+  # bitvectors are significantly faster for GC-bench, but slower for
+  # bootstrapping and use more memory
   rcWhite = 0
   rcGrey = 1   # unused
   rcBlack = 2

todo.txt

@@ -2,9 +2,8 @@ version 0.9.2
 =============
 
 - FFI:
-  * test libffi on windows  --> fowl
+  * test libffi on windows
   * test: times.format with the FFI
-- GC: variation of m&s GC
 - GC: implement simple generational GC
   * first version: mark black in write barrier
   * second version: introduce fake roots instead of marking black
@@ -108,7 +107,8 @@ GC
 Optimizations
 =============
 
-- optimize 'if' with a constant condition
+- optimize 'if' with a constant condition --> necessary in frontend for better
+  dead code elimination
 - escape analysis for string/seq seems to be easy to do too;
   even further write barrier specialization
 - inlining of first class functions