Fixed up some things I borked along the way with lists and loops.

Minor edits.

dev

@@ -44,7 +44,13 @@ function meson_build {
             rm -rf ${1}
         fi
         log Creating meson target ${1}
-        meson setup ${1}
+
+        if [[ ${1} -ne "debug" ]] ; then
+            SETUP_ARGS=-Dbuildtype=release
+        else
+            SETUP_ARGS=-Dbuildtype=debugoptimized
+        fi
+        meson setup ${SETUP_ARGS} ${1}
     fi
     cd ${1}
 

include/con4m.h

@@ -9,13 +9,30 @@
 
 // #define C4M_GC_ALL_OFF
 // #define C4M_GC_ALL_ON
-// #define C4M_VM_DEBUG
 // #define C4M_TYPE_LOG
 
-#ifndef C4M_NO_DEV_MODE
+// When this is on, the `debug` instruction will run.
+// Note that the debug instruction is not even generated unless
+// C4M_DEV is on.
+
+// #define C4M_VM_DEBUG
+// #define C4M_VM_DEBUG_DEFAULT true
+
+#ifdef C4M_NO_DEV_MODE
+#undef C4M_DEV
+#else
 #define C4M_DEV
 #endif
 
+#if defined(C4M_VM_DEBUG)
+#if !defined(C4M_DEV)
+#error "Cannot debug VM when C4M_DEV_MODE is set."
+#endif
+#if !defined(C4M_VM_DEBUG_DEFAULT)
+#define C4M_VM_DEBUG_DEFAULT false
+#endif
+#endif
+
 #ifdef C4M_TRACE_GC
 #ifndef C4M_GC_STATS
 #define C4M_GC_STATS
@@ -49,7 +66,7 @@
 #include "con4m/color.h"
 
 // Basic "exclusive" (i.e., single threaded) list.
-#include "con4m/xlist.h"
+#include "con4m/list.h"
 
 // Type system API.
 #include "con4m/type.h"

include/con4m/datatypes/vm.h

@@ -337,6 +337,7 @@ typedef enum : uint8_t {
     C4M_ZNop           = 0xFF,
 #ifdef C4M_DEV
     C4M_ZPrint = 0xFD,
+    C4M_ZDebug = 0xFC,
 #endif
 } c4m_zop_t;
 

include/con4m/list.h

@@ -0,0 +1,23 @@
+#pragma once
+
+#include "con4m.h"
+
+extern void       *c4m_list_get(c4m_list_t *, int64_t, bool *);
+extern void        c4m_list_append(c4m_list_t *list, void *item);
+extern void        c4m_list_add_if_unique(c4m_list_t *list,
+                                          void       *item,
+                                          bool (*fn)(void *, void *));
+extern void       *c4m_list_pop(c4m_list_t *list);
+extern void        c4m_list_plus_eq(c4m_list_t *, c4m_list_t *);
+extern c4m_list_t *c4m_list_plus(c4m_list_t *, c4m_list_t *);
+extern bool        c4m_list_set(c4m_list_t *, int64_t, void *);
+extern c4m_list_t *c4m_list(c4m_type_t *);
+extern int64_t     c4m_list_len(const c4m_list_t *);
+extern c4m_list_t *c4m_list_get_slice(c4m_list_t *, int64_t, int64_t);
+extern void        c4m_list_set_slice(c4m_list_t *,
+                                      int64_t,
+                                      int64_t,
+                                      c4m_list_t *);
+extern bool        c4m_list_contains(c4m_list_t *, c4m_obj_t);
+extern c4m_list_t *c4m_list_copy(c4m_list_t *);
+extern c4m_list_t *c4m_list_shallow_copy(c4m_list_t *);

include/con4m/string.h

@@ -13,7 +13,7 @@ extern c4m_utf8_t         *c4m_utf8_repeat(c4m_codepoint_t, int64_t);
 extern c4m_utf32_t        *c4m_utf32_repeat(c4m_codepoint_t, int64_t);
 extern c4m_utf32_t        *_c4m_str_strip(const c4m_str_t *s, ...);
 extern c4m_str_t          *_c4m_str_truncate(const c4m_str_t *s, int64_t, ...);
-extern c4m_utf32_t        *_c4m_str_join(const c4m_list_t *,
+extern c4m_utf32_t        *_c4m_str_join(c4m_list_t *,
                                          const c4m_str_t *,
                                          ...);
 extern c4m_utf8_t         *c4m_str_from_int(int64_t n);
@@ -24,7 +24,7 @@ extern c4m_utf8_t         *c4m_rich(c4m_utf8_t *, c4m_utf8_t *style);
 extern c4m_codepoint_t     c4m_index(const c4m_str_t *, int64_t);
 extern bool                c4m_str_can_coerce_to(c4m_type_t *, c4m_type_t *);
 extern c4m_obj_t           c4m_str_coerce_to(const c4m_str_t *, c4m_type_t *);
-extern c4m_list_t        *c4m_str_xsplit(c4m_str_t *, c4m_str_t *);
+extern c4m_list_t         *c4m_str_xsplit(c4m_str_t *, c4m_str_t *);
 extern struct flexarray_t *c4m_str_fsplit(c4m_str_t *, c4m_str_t *);
 extern bool                c4m_str_starts_with(const c4m_str_t *,
                                                const c4m_str_t *);
@@ -102,8 +102,8 @@ c4m_to_cstring(c4m_str_t *s)
     return s->data;
 }
 
-extern c4m_list_t *c4m_u8_map(const c4m_list_t *);
-extern bool         c4m_str_eq(c4m_str_t *, c4m_str_t *);
+extern c4m_list_t *c4m_u8_map(c4m_list_t *);
+extern bool        c4m_str_eq(c4m_str_t *, c4m_str_t *);
 
 extern const uint64_t c4m_pmap_str[2];
 

meson.build

@@ -35,7 +35,6 @@ else
   using_osx = false
   link_args = []
   c_args    = c_args + ['-D_GNU_SOURCE']
-
 endif
 
 exe_link_args = link_args + ['-flto', '-w']
@@ -54,6 +53,10 @@ if cc.links(fpty_code, name: 'forkpty_check')
   add_project_arguments('-DHAVE_PTY_H', language: 'c')
 endif
 
+if (get_option('buildtype') == 'release')
+  add_project_arguments([ '-O2' ], language: 'c')
+endif
+
 c4m_src  = ['src/con4m/style.c',
             'src/con4m/colors.c',
             'src/con4m/breaks.c', 

src/con4m/collect.c

@@ -671,10 +671,10 @@ c4m_collect_arena(c4m_arena_t *from_space)
     gstr = c4m_cstr_format("{}", c4m_box_double(u));
     gstr = c4m_str_slice(gstr, 0, 5);
 
-    c4m_printf("[b]Collect utilization[/]: [em]{}[/]% garbage",
+    c4m_printf("[b]Collect utilization[/]: [em]{}%[/] [i]garbage",
                gstr);
 
-    c4m_printf("[b]Average allocation size:[/] [em]{:,} bytes",
+    c4m_printf("[b]Average allocation size:[/] [em]{:,}[/] bytes",
                c4m_box_u64((c4m_total_words * 8) / c4m_total_allocs));
 
 #endif

src/con4m/compiler/codegen.c

@@ -1,6 +1,16 @@
 #define C4M_USE_INTERNAL_API
 #include "con4m.h"
 
+#ifdef C4M_DEV
+#define gen_debug(debug_arg) \
+    emit(ctx, C4M_ZDebug, c4m_kw("arg", c4m_ka(debug_arg)))
+#define debug_label(cstr) \
+    gen_label(ctx, c4m_new_utf8(cstr))
+#else
+#define gen_debug(debug_arg)
+#define debug_label(cstr)
+#endif
+
 // This is used in nested if/else only.
 typedef struct {
     c4m_jump_info_t *targets;
@@ -31,6 +41,7 @@ typedef struct {
     int                   instruction_counter;
     int                   current_stack_offset;
     int                   max_stack_size;
+    int                   module_patch_loc;
     bool                  lvalue;
     assign_type_t         assign_method;
     c4m_symbol_t         *retsym;
@@ -796,6 +807,8 @@ gen_module(gen_ctx *ctx)
     int num_params = gen_parameter_checks(ctx);
 
     emit(ctx, C4M_ZModuleEnter, c4m_kw("arg", c4m_ka(num_params)));
+    ctx->module_patch_loc = ctx->instruction_counter;
+    emit(ctx, C4M_ZMoveSp, c4m_kw("arg", c4m_ka(0)));
     gen_kids(ctx);
 }
 
@@ -1093,6 +1106,7 @@ gen_ranged_for(gen_ctx *ctx, c4m_loop_info_t *li)
     // value. So if it's 0 .. 10, and we call subtract to calculate the
     // length, we're going to be computing 0 - 10, not the opposite.
     // So we first swap the two numbers.
+
     emit(ctx, C4M_ZSwap);
     // Subtract the 2 numbers now, but DO NOT POP; we need these.
     emit(ctx, C4M_ZSubNoPop);
@@ -1137,6 +1151,7 @@ gen_ranged_for(gen_ctx *ctx, c4m_loop_info_t *li)
     if (using_index) {
         emit(ctx, C4M_ZPopToR3);
     }
+
     // pop the step for a second.
     emit(ctx, C4M_ZPopToR1);
     // If the two items are equal, we bail from the loop.
@@ -1147,13 +1162,14 @@ gen_ranged_for(gen_ctx *ctx, c4m_loop_info_t *li)
     // Then push the step back on,
     // And, if it's being used, the $i from R3.
 
-    GEN_JNZ(gen_sym_store(ctx, li->lvar_1, false);
-            emit(ctx, C4M_ZPushFromR1);
-            emit(ctx, C4M_ZAdd);
-            emit(ctx, C4M_ZPushFromR1);
-            possible_restore_from_r3(ctx, using_index);
-            gen_one_kid(ctx, ctx->cur_node->num_kids - 1);
-            gen_j(ctx, &ji_top));
+    GEN_JNZ(
+        gen_sym_store(ctx, li->lvar_1, false);
+        emit(ctx, C4M_ZPushFromR1);
+        emit(ctx, C4M_ZAdd);
+        emit(ctx, C4M_ZPushFromR1);
+        possible_restore_from_r3(ctx, using_index);
+        gen_one_kid(ctx, ctx->cur_node->num_kids - 1);
+        gen_j(ctx, &ji_top));
     gen_apply_waiting_patches(ctx, &li->branch_info);
     emit(ctx, C4M_ZMoveSp, c4m_kw("arg", c4m_ka(-3)));
 }
@@ -1204,21 +1220,16 @@ gen_container_for(gen_ctx *ctx, c4m_loop_info_t *li)
     //
     // Also, note that the VIEW builtin doesn't need to copy objects,
     // but it needs to give us a pointer to items, where we steal the
-    // lowest 2 bits in the pointer to represent the the log base 2 of
+    // lowest 3 bits in the pointer to represent the the log base 2 of
     // the bytes in the item count.  Currently, the only allowable
-    // view item sizes are therefore: 1 byte, 2 bytes, 4 bytes, 8
-    // bytes.
+    // view item sizes are 1 byte, 2 bytes, 4 bytes, 8 bytes. Anything
+    // else currently represents a bitfield (per below).
     //
-    // Currently, we don't need more than 8bytes, since anything
+    // Currently, we don't need more than 8 bytes, since anything
     // larger than 8 bytes is passed around as pointers. However, I
     // expect to eventually add 128-bit ints, which would entail a
     // double-word load.
     //
-    // Therefore, views must be 8 byte aligned, because eventually we
-    // will steal a third bit. But that's no problem as long as view
-    // pointers start at the beginning of an allocation, since the
-    // allocator always spits out aligned objects.
-    //
     // We have a bit of a special case for bitfield types (right now,
     // just c4m_flags_t). There, the value of the "number of bytes"
     // field will be set to 128; When we see that, we will use
@@ -1229,21 +1240,25 @@ gen_container_for(gen_ctx *ctx, c4m_loop_info_t *li)
     // of first implementation, it will just always happen, even
     // though in most cases it should be no problem to bind to the
     // right approach statically.
-
     gen_tcall(ctx, C4M_BI_VIEW, NULL);
+
     // The length of the container is on top right now; the view is
     // underneath.  We want to store a copy to $len if appropriate,
     // and then pop to a register 2, to work on the pointer. We'll
     // push it back on top when we properly enter the loop.
     gen_len_var_init(ctx, li);
+
     // Move the container length out to a register for a bit.
     emit(ctx, C4M_ZPopToR2);
+
     emit(ctx, C4M_ZUnsteal);
     // The bit length is actually encoded by taking log base 2; here
     // were expand it back out before going into the loop.
     emit(ctx, C4M_ZShlI, c4m_kw("arg", c4m_ka(0x1)));
+
     // On top of this, put the iteration count, which at the start of
     // each turn through the loop, will be the second item.
+
     bool have_index_var = gen_index_var_init(ctx, li);
     bool have_kv_pair   = li->lvar_2 != NULL;
 
@@ -1276,18 +1291,17 @@ gen_container_for(gen_ctx *ctx, c4m_loop_info_t *li)
             deal_with_iteration_count(ctx, li, have_index_var);
             emit(ctx, C4M_ZPopToR1);
             emit(ctx, C4M_ZLoadFromView, c4m_kw("arg", c4m_ka(have_kv_pair)));
-            // Store the item(s) to the appropriate loop variable(s).
             store_view_item(ctx, li);
-            emit(ctx, C4M_ZPushFromR1); // Re-groom the stack; container length.
-            emit(ctx, C4M_ZPushFromR2); // Iteration count.
+            emit(ctx, C4M_ZPushFromR1); // Re-groom the stack; iter count
+            emit(ctx, C4M_ZPushFromR2); // container len
             gen_one_kid(ctx, ctx->cur_node->num_kids - 1);
             gen_j(ctx, &ji_top));
     // After the loop:
     // 1. backpatch.
     // 2. Move the stack down four items, popping the count, len, item size,
     //    and container.
     gen_apply_waiting_patches(ctx, &li->branch_info);
-    emit(ctx, C4M_ZMoveSp, c4m_kw("arg", c4m_ka(-4)));
+    emit(ctx, C4M_ZMoveSp, c4m_kw("arg", c4m_ka(-5)));
 }
 
 static inline void
@@ -2122,6 +2136,12 @@ gen_module_code(gen_ctx *ctx, c4m_vm_t *vm)
     ctx->current_stack_offset = ctx->fctx->static_size;
     ctx->max_stack_size       = ctx->fctx->static_size;
     gen_one_node(ctx);
+
+    c4m_zinstruction_t *sp_move = c4m_list_get(module->instructions,
+                                               ctx->module_patch_loc,
+                                               NULL);
+    sp_move->arg                = ctx->max_stack_size;
+
     emit(ctx, C4M_ZModuleRet);
 
     module->module_var_size = ctx->max_stack_size;

src/con4m/compiler/disasm.c

@@ -53,6 +53,8 @@ get_bool_label(c4m_zop_t op)
         return c4m_new_utf8("currently unused");
     case C4M_ZLoadFromView:
         return c4m_new_utf8("load kv pair");
+    case C4M_ZDebug:
+        return c4m_new_utf8("set debug");
     default:
         c4m_unreachable();
     }
@@ -334,6 +336,10 @@ const inst_info_t inst_info[256] = {
     [C4M_ZPrint] = {
         .name = "ZPrint",
     },
+    [C4M_ZDebug] = {
+        .name    = "ZDebug",
+        .arg_fmt = fmt_bool,
+    },
 #endif
 };