static void validate_lex_operand(Validator *val, MVMuint32 flags) {
MVMuint16 lex_index, frame_index, i;
MVMuint32 lex_count;
MVMStaticFrame *frame = val->frame;
/* Two steps forward, two steps back to keep the error reporting happy,
and to make the endian conversion within ensure_bytes correct.
(Both are using val->cur_op, and want it to have different values.) */
ensure_bytes(val, 2);
lex_index = GET_UI16(val->cur_op, 0);
val->cur_op += 2;
ensure_bytes(val, 2);
val->cur_op -= 2;
frame_index = GET_UI16(val->cur_op, 2);
for (i = frame_index; i; i--) {
frame = frame->body.outer;
if (!frame)
fail(val, MSG(val, "lexical operand requires %" PRIu16
" more enclosing scopes"), i);
}
lex_count = frame->body.num_lexicals;
if (lex_index >= lex_count)
fail(val, MSG(val, "lexical operand index %" PRIu16
" out of range 0.. %" PRIu32), lex_index, lex_count - 1);
val->cur_op += 4;
}
void
check_free_entry(byte *sp, byte *ep,
byte tag,
Uint16 ct_no, int ct_no_n,
Uint16 t_no, int t_no_n,
UWord ptr, int ptr_n,
Uint32 ti,int ti_n)
{
byte *p = sp;
Uint16 hdr;
ASSERT(*p == tag);
p++;
hdr = GET_UI16(p);
if (tag == ERTS_MT_CRR_FREE_BDY_TAG)
check_ui(&hdr, &p, ct_no, ct_no_n, UI16_MSB_EHF_MSK, UI16_MSB_EHF_SZ);
check_ui(&hdr, &p, t_no, t_no_n, UI16_MSB_EHF_MSK, UI16_MSB_EHF_SZ);
check_ui(&hdr, &p, ptr, ptr_n, UI_MSB_EHF_MSK, UI_MSB_EHF_SZ);
check_ui(&hdr, &p, ti, ti_n, UI32_MSB_EHF_MSK, UI32_MSB_EHF_SZ);
ASSERT(hdr == 0);
ASSERT(p == ep);
}
void
check_alloc_entry(byte *sp, byte *ep,
byte tag,
Uint16 ct_no, int ct_no_n,
Uint16 t_no, int t_no_n,
UWord res, int res_n,
Uint size, int size_n,
Uint32 ti,int ti_n)
{
byte *p = sp;
Uint16 hdr;
ASSERT(*p == tag);
p++;
hdr = GET_UI16(p);
if (tag == ERTS_MT_CRR_ALLOC_BDY_TAG)
check_ui(&hdr, &p, ct_no, ct_no_n, UI16_MSB_EHF_MSK, UI16_MSB_EHF_SZ);
check_ui(&hdr, &p, t_no, t_no_n, UI16_MSB_EHF_MSK, UI16_MSB_EHF_SZ);
check_ui(&hdr, &p, res, res_n, UI_MSB_EHF_MSK, UI_MSB_EHF_SZ);
check_ui(&hdr, &p, size, size_n, UI_MSB_EHF_MSK, UI_MSB_EHF_SZ);
check_ui(&hdr, &p, ti, ti_n, UI32_MSB_EHF_MSK, UI32_MSB_EHF_SZ);
ASSERT(hdr == 0);
ASSERT(p == ep);
}
static void validate_block(Validator *val) {
int block_id = val->cur_mark[1];
switch (block_id) {
case 'a': {
MVMuint16 index;
ensure_op(val, MVM_OP_prepargs);
validate_operands(val);
index = GET_UI16(val->cur_op, -2);
val->cur_call = val->cu->body.callsites[index];
val->cur_arg = 0;
val->expected_named_arg = 0;
val->remaining_positionals = val->cur_call->num_pos;
break;
}
default:
fail(val, MSG(val, "unknown instruction block '%c'"), block_id);
}
while (val->cur_op < val->bc_end) {
int type, id;
read_op(val);
type = val->cur_mark[0];
id = val->cur_mark[1];
if (id != block_id)
fail(val, MSG(val, "expected instruction marked '%c' but got '%c'"),
block_id, id);
switch (type) {
case MARK_body:
break;
case MARK_tail:
goto terminate_block;
default:
fail_illegal_mark(val);
}
switch (block_id) {
case 'a':
validate_operands(val);
validate_arg(val);
break;
}
}
terminate_block:
switch (block_id) {
case 'a':
validate_operands(val);
ensure_no_remaining_positionals(val);
break;
}
}
void
check_time_inc_entry(byte *sp, byte *ep,
Uint32 secs, int secs_n,
Uint32 usecs, int usecs_n)
{
byte *p = sp;
Uint16 hdr;
ASSERT(*p == ERTS_MT_TIME_INC_BDY_TAG);
p++;
hdr = GET_UI16(p);
check_ui(&hdr, &p, secs, secs_n, UI32_MSB_EHF_MSK, UI32_MSB_EHF_SZ);
check_ui(&hdr, &p, usecs, usecs_n, UI32_MSB_EHF_MSK, UI32_MSB_EHF_SZ);
ASSERT(hdr == 0);
ASSERT(p == ep);
}
static void validate_literal_operand(Validator *val, MVMuint32 flags) {
MVMuint32 type = flags & MVM_operand_type_mask;
MVMuint32 size;
switch (type) {
case MVM_operand_int8:
size = 1;
break;
case MVM_operand_int16:
size = 2;
break;
case MVM_operand_int32:
size = 4;
break;
case MVM_operand_int64:
size = 8;
break;
case MVM_operand_num32:
size = 4;
break;
case MVM_operand_num64:
size = 8;
break;
case MVM_operand_callsite:
size = 2;
break;
case MVM_operand_coderef:
size = 2;
break;
case MVM_operand_str:
size = 4;
break;
case MVM_operand_ins:
size = 4;
break;
case MVM_operand_obj:
case MVM_operand_type_var:
fail(val, MSG(val, "operand type %i can't be a literal"), type);
default:
fail(val, MSG(val, "unknown operand type %i"), type);
}
ensure_bytes(val, size);
switch (type) {
case MVM_operand_callsite: {
MVMuint16 index = GET_UI16(val->cur_op, 0);
MVMuint32 count = val->cu->body.orig_callsites;
if (index >= count)
fail(val, MSG(val, "callsite index %" PRIu16
" out of range 0..%" PRIu32), index, count - 1);
break;
}
case MVM_operand_coderef: {
MVMuint16 index = GET_UI16(val->cur_op, 0);
MVMuint32 count = val->cu->body.orig_frames;
if (index >= count)
fail(val, MSG(val, "coderef index %" PRIu16
" out of range 0..%" PRIu32), index, count - 1);
break;
}
case MVM_operand_str: {
MVMuint32 index = GET_UI32(val->cur_op, 0);
MVMuint32 count = val->cu->body.orig_strings;
if (index >= count)
fail(val, MSG(val, "string index %" PRIu32
" out of range 0..%" PRIu32), index, count - 1);
break;
}
case MVM_operand_ins: {
MVMuint32 offset = GET_UI32(val->cur_op, 0);
if (offset >= val->bc_size)
fail(val, MSG(val, "branch instruction offset %" PRIu32
" out of range 0..%" PRIu32), offset, val->bc_size - 1);
val->labels[offset] |= MVM_BC_branch_target;
}
}
val->cur_op += size;
}
static void build_cfg(MVMThreadContext *tc, MVMSpeshGraph *g, MVMStaticFrame *sf,
MVMint32 *existing_deopts, MVMint32 num_existing_deopts) {
MVMSpeshBB *cur_bb, *prev_bb;
MVMSpeshIns *last_ins;
MVMint64 i;
MVMint32 bb_idx;
/* Temporary array of all MVMSpeshIns we create (one per instruction).
* Overestimate at size. Has the flat view, matching the bytecode. */
MVMSpeshIns **ins_flat = MVM_calloc(g->bytecode_size / 2, sizeof(MVMSpeshIns *));
/* Temporary array where each byte in the input bytecode gets a 32-bit
* integer. This is used for two things:
* A) When we make the MVMSpeshIns for an instruction starting at the
* byte, we put the instruction index (into ins_flat) in the slot,
* shifting it by 2 bits to the left. We will use this to do fixups.
* B) The first bit is "I have an incoming branch" - that is, start of
* a basic block. The second bit is "I can branch" - that is, end of
* a basic block. It's possible to have both bits set.
* Anything that's just a zero has no instruction starting there. */
MVMuint32 *byte_to_ins_flags = MVM_calloc(g->bytecode_size, sizeof(MVMuint32));
/* Instruction to basic block mapping. Initialized later. */
MVMSpeshBB **ins_to_bb = NULL;
/* Make first pass through the bytecode. In this pass, we make MVMSpeshIns
* nodes for each instruction and set the start/end of block bits. Also
* set handler targets as basic block starters. */
MVMCompUnit *cu = sf->body.cu;
MVMuint8 *pc = g->bytecode;
MVMuint8 *end = g->bytecode + g->bytecode_size;
MVMuint32 ins_idx = 0;
MVMuint8 next_bbs = 1; /* Next iteration (here, first) starts a BB. */
for (i = 0; i < g->num_handlers; i++)
byte_to_ins_flags[g->handlers[i].goto_offset] |= MVM_CFG_BB_START;
while (pc < end) {
/* Look up op info. */
MVMuint16 opcode = *(MVMuint16 *)pc;
MVMuint8 *args = pc + 2;
MVMuint8 arg_size = 0;
const MVMOpInfo *info = get_op_info(tc, cu, opcode);
/* Create an instruction node, add it, and record its position. */
MVMSpeshIns *ins_node = MVM_spesh_alloc(tc, g, sizeof(MVMSpeshIns));
ins_flat[ins_idx] = ins_node;
byte_to_ins_flags[pc - g->bytecode] |= ins_idx << 2;
/* Did previous instruction end a basic block? */
if (next_bbs) {
byte_to_ins_flags[pc - g->bytecode] |= MVM_CFG_BB_START;
next_bbs = 0;
}
/* Also check we're not already a BB start due to being a branch
* target, in which case we should ensure our prior is marked as
* a BB end. */
else {
if (byte_to_ins_flags[pc - g->bytecode] & MVM_CFG_BB_START) {
MVMuint32 hunt = pc - g->bytecode;
while (!byte_to_ins_flags[--hunt]);
byte_to_ins_flags[hunt] |= MVM_CFG_BB_END;
}
}
/* Store opcode */
ins_node->info = info;
/* Go over operands. */
ins_node->operands = MVM_spesh_alloc(tc, g, info->num_operands * sizeof(MVMSpeshOperand));
for (i = 0; i < info->num_operands; i++) {
MVMuint8 flags = info->operands[i];
MVMuint8 rw = flags & MVM_operand_rw_mask;
switch (rw) {
case MVM_operand_read_reg:
case MVM_operand_write_reg:
ins_node->operands[i].reg.orig = GET_UI16(args, arg_size);
arg_size += 2;
break;
case MVM_operand_read_lex:
case MVM_operand_write_lex:
ins_node->operands[i].lex.idx = GET_UI16(args, arg_size);
ins_node->operands[i].lex.outers = GET_UI16(args, arg_size + 2);
arg_size += 4;
break;
case MVM_operand_literal: {
MVMuint32 type = flags & MVM_operand_type_mask;
switch (type) {
case MVM_operand_int8:
ins_node->operands[i].lit_i8 = GET_I8(args, arg_size);
arg_size += 1;
break;
case MVM_operand_int16:
ins_node->operands[i].lit_i16 = GET_I16(args, arg_size);
arg_size += 2;
break;
case MVM_operand_int32:
ins_node->operands[i].lit_i32 = GET_I32(args, arg_size);
arg_size += 4;
break;
case MVM_operand_int64:
ins_node->operands[i].lit_i64 = MVM_BC_get_I64(args, arg_size);
arg_size += 8;
break;
case MVM_operand_num32:
ins_node->operands[i].lit_n32 = GET_N32(args, arg_size);
arg_size += 4;
break;
case MVM_operand_num64:
ins_node->operands[i].lit_n64 = MVM_BC_get_N64(args, arg_size);
arg_size += 8;
break;
case MVM_operand_callsite:
ins_node->operands[i].callsite_idx = GET_UI16(args, arg_size);
arg_size += 2;
break;
case MVM_operand_coderef:
ins_node->operands[i].coderef_idx = GET_UI16(args, arg_size);
arg_size += 2;
break;
case MVM_operand_str:
ins_node->operands[i].lit_str_idx = GET_UI32(args, arg_size);
arg_size += 4;
break;
case MVM_operand_ins: {
/* Stash instruction offset. */
MVMuint32 target = GET_UI32(args, arg_size);
ins_node->operands[i].ins_offset = target;
/* This is a branching instruction, so it's a BB end. */
byte_to_ins_flags[pc - g->bytecode] |= MVM_CFG_BB_END;
/* Its target is a BB start, and any previous instruction
* we already passed needs marking as a BB end. */
byte_to_ins_flags[target] |= MVM_CFG_BB_START;
if (target > 0 && target < pc - g->bytecode) {
while (!byte_to_ins_flags[--target]);
byte_to_ins_flags[target] |= MVM_CFG_BB_END;
}
/* Next instruction is also a BB start. */
next_bbs = 1;
arg_size += 4;
break;
}
case MVM_operand_spesh_slot:
ins_node->operands[i].lit_i16 = GET_I16(args, arg_size);
arg_size += 2;
break;
default:
MVM_exception_throw_adhoc(tc,
"Spesh: unknown operand type %d in graph building (op %s)",
(int)type, ins_node->info->name);
}
}
break;
}
}
/* We specially handle the jumplist case, which needs to mark all of
* the possible places we could jump to in the following instructions
* as starts of basic blocks. It is, in itself, the end of one. Note
* we jump to the instruction after the n jump points if none match,
* so that is marked too. */
if (opcode == MVM_OP_jumplist) {
MVMint64 n = MVM_BC_get_I64(args, 0);
for (i = 0; i <= n; i++)
byte_to_ins_flags[(pc - g->bytecode) + 12 + i * 6] |= MVM_CFG_BB_START;
byte_to_ins_flags[pc - g->bytecode] |= MVM_CFG_BB_END;
}
/* Invocations, returns, and throws are basic block ends. */
switch (opcode) {
case MVM_OP_invoke_v:
case MVM_OP_invoke_i:
case MVM_OP_invoke_n:
case MVM_OP_invoke_s:
case MVM_OP_invoke_o:
case MVM_OP_return_i:
case MVM_OP_return_n:
case MVM_OP_return_s:
case MVM_OP_return_o:
case MVM_OP_return:
case MVM_OP_throwdyn:
case MVM_OP_throwlex:
case MVM_OP_throwlexotic:
case MVM_OP_throwcatdyn:
case MVM_OP_throwcatlex:
case MVM_OP_throwcatlexotic:
case MVM_OP_die:
case MVM_OP_rethrow:
case MVM_OP_resume:
byte_to_ins_flags[pc - g->bytecode] |= MVM_CFG_BB_END;
next_bbs = 1;
break;
}
/* Final instruction is basic block end. */
if (pc + 2 + arg_size == end)
byte_to_ins_flags[pc - g->bytecode] |= MVM_CFG_BB_END;
/* Caculate next instruction's PC. */
pc += 2 + arg_size;
/* If this is a deopt point opcode... */
if (!existing_deopts && (info->deopt_point & MVM_DEOPT_MARK_ONE))
add_deopt_annotation(tc, g, ins_node, pc, MVM_SPESH_ANN_DEOPT_ONE_INS);
if (!existing_deopts && (info->deopt_point & MVM_DEOPT_MARK_ALL))
add_deopt_annotation(tc, g, ins_node, pc, MVM_SPESH_ANN_DEOPT_ALL_INS);
if (!existing_deopts && (info->deopt_point & MVM_DEOPT_MARK_OSR))
add_deopt_annotation(tc, g, ins_node, pc, MVM_SPESH_ANN_DEOPT_OSR);
/* Go to next instruction. */
ins_idx++;
}
/* Annotate instructions that are handler-significant. */
for (i = 0; i < g->num_handlers; i++) {
MVMSpeshIns *start_ins = ins_flat[byte_to_ins_flags[g->handlers[i].start_offset] >> 2];
MVMSpeshIns *end_ins = ins_flat[byte_to_ins_flags[g->handlers[i].end_offset] >> 2];
MVMSpeshIns *goto_ins = ins_flat[byte_to_ins_flags[g->handlers[i].goto_offset] >> 2];
MVMSpeshAnn *start_ann = MVM_spesh_alloc(tc, g, sizeof(MVMSpeshAnn));
MVMSpeshAnn *end_ann = MVM_spesh_alloc(tc, g, sizeof(MVMSpeshAnn));
MVMSpeshAnn *goto_ann = MVM_spesh_alloc(tc, g, sizeof(MVMSpeshAnn));
start_ann->next = start_ins->annotations;
start_ann->type = MVM_SPESH_ANN_FH_START;
start_ann->data.frame_handler_index = i;
start_ins->annotations = start_ann;
end_ann->next = end_ins->annotations;
end_ann->type = MVM_SPESH_ANN_FH_END;
end_ann->data.frame_handler_index = i;
end_ins->annotations = end_ann;
goto_ann->next = goto_ins->annotations;
goto_ann->type = MVM_SPESH_ANN_FH_GOTO;
goto_ann->data.frame_handler_index = i;
goto_ins->annotations = goto_ann;
}
/* Annotate instructions that are inline start/end points. */
for (i = 0; i < g->num_inlines; i++) {
MVMSpeshIns *start_ins = ins_flat[byte_to_ins_flags[g->inlines[i].start] >> 2];
MVMSpeshIns *end_ins = ins_flat[byte_to_ins_flags[g->inlines[i].end] >> 2];
MVMSpeshAnn *start_ann = MVM_spesh_alloc(tc, g, sizeof(MVMSpeshAnn));
MVMSpeshAnn *end_ann = MVM_spesh_alloc(tc, g, sizeof(MVMSpeshAnn));
start_ann->next = start_ins->annotations;
start_ann->type = MVM_SPESH_ANN_INLINE_START;
start_ann->data.inline_idx = i;
start_ins->annotations = start_ann;
end_ann->next = end_ins->annotations;
end_ann->type = MVM_SPESH_ANN_INLINE_END;
end_ann->data.inline_idx = i;
end_ins->annotations = end_ann;
}
/* Now for the second pass, where we assemble the basic blocks. Also we
* build a lookup table of instructions that start a basic block to that
* basic block, for the final CFG construction. We make the entry block a
* special one, containing a noop; it will have any exception handler
* targets linked from it, so they show up in the graph. */
g->entry = MVM_spesh_alloc(tc, g, sizeof(MVMSpeshBB));
g->entry->first_ins = MVM_spesh_alloc(tc, g, sizeof(MVMSpeshIns));
g->entry->first_ins->info = get_op_info(tc, cu, 0);
g->entry->last_ins = g->entry->first_ins;
g->entry->idx = 0;
cur_bb = NULL;
prev_bb = g->entry;
last_ins = NULL;
ins_to_bb = MVM_calloc(ins_idx, sizeof(MVMSpeshBB *));
ins_idx = 0;
bb_idx = 1;
for (i = 0; i < g->bytecode_size; i++) {
MVMSpeshIns *cur_ins;
/* Skip zeros; no instruction here. */
if (!byte_to_ins_flags[i])
continue;
/* Get current instruction. */
cur_ins = ins_flat[byte_to_ins_flags[i] >> 2];
/* Start of a basic block? */
if (byte_to_ins_flags[i] & MVM_CFG_BB_START) {
/* Should not already be in a basic block. */
if (cur_bb) {
MVM_spesh_graph_destroy(tc, g);
MVM_exception_throw_adhoc(tc, "Spesh: confused during basic block analysis (in block)");
}
/* Create it, and set first instruction and index. */
cur_bb = MVM_spesh_alloc(tc, g, sizeof(MVMSpeshBB));
cur_bb->first_ins = cur_ins;
cur_bb->idx = bb_idx;
bb_idx++;
/* Record instruction -> BB start mapping. */
ins_to_bb[ins_idx] = cur_bb;
/* Link it to the previous one. */
prev_bb->linear_next = cur_bb;
}
/* Should always be in a BB at this point. */
if (!cur_bb) {
MVM_spesh_graph_destroy(tc, g);
MVM_exception_throw_adhoc(tc, "Spesh: confused during basic block analysis (no block)");
}
/* Add instruction into double-linked per-block instruction list. */
if (last_ins) {
last_ins->next = cur_ins;
cur_ins->prev = last_ins;
}
last_ins = cur_ins;
/* End of a basic block? */
if (byte_to_ins_flags[i] & MVM_CFG_BB_END) {
cur_bb->last_ins = cur_ins;
prev_bb = cur_bb;
cur_bb = NULL;
last_ins = NULL;
}
ins_idx++;
}
g->num_bbs = bb_idx;
/* Finally, link the basic blocks up to form a CFG. Along the way, any of
* the instruction operands get the target BB stored. */
cur_bb = g->entry;
while (cur_bb) {
/* If it's the first block, it's a special case; successors are the
* real successor and all exception handlers. */
if (cur_bb == g->entry) {
cur_bb->num_succ = 1 + g->num_handlers;
cur_bb->succ = MVM_spesh_alloc(tc, g, cur_bb->num_succ * sizeof(MVMSpeshBB *));
cur_bb->succ[0] = cur_bb->linear_next;
for (i = 0; i < g->num_handlers; i++) {
MVMuint32 offset = g->handlers[i].goto_offset;
cur_bb->succ[i + 1] = ins_to_bb[byte_to_ins_flags[offset] >> 2];
}
}
/* Otherwise, consider the last instruction, to see how we leave the BB. */
else {
switch (cur_bb->last_ins->info->opcode) {
int unZ(char *archiveFilename, char *extractPath)
{
FILE *fp;
fchunk_t fc;
unsigned int i, destsize;
unsigned char header[256], *toc;
unsigned int numFiles, compsize, tocOffset, tocSize, numDirs;
direntry_t *dirs;
destsize = strlen(extractPath);
fp = fopen(archiveFilename, "rb");
if (!fp)
{
printf("Error opening %s!\n", archiveFilename);
return;
}
fseek(fp, 0, SEEK_END);
compsize = ftell(fp);
fseek(fp, 0, SEEK_SET);
fread(header, 256, 1, fp);
// check signature
if (GET_UI32(header) != 0x8C655D13)
{
printf("Archive corrupt: Bad signature!\n");
return 0;
}
//numFiles = GET_UI16(header + 12);
// check size
if (compsize != GET_UI32(header + 18))
{
printf("Archive corrupt: Incorrect size!\n");
return 0;
}
// read toc
tocOffset = GET_UI32(header + 41);
tocSize = compsize - tocOffset;
toc = malloc(tocSize);
fseek(fp, tocOffset, SEEK_SET);
fread(toc, 1, tocSize, fp);
tocOffset = 0;
// parse dirs
numDirs = GET_UI16(header + 49);
dirs = malloc(sizeof(*dirs) * numDirs);
for (i = 0; i < numDirs; i++)
{
unsigned int dir_numFiles = GET_UI16(toc + tocOffset);
unsigned int dir_entrySize = GET_UI16(toc + tocOffset + 2);
unsigned int dir_nameSize = GET_UI16(toc + tocOffset + 4);
dirs[i].name = malloc(destsize + 1 + dir_nameSize + 1);
strcpy(dirs[i].name, extractPath);
dirs[i].name[destsize] = '\\';
memcpy(dirs[i].name + destsize + 1, toc + tocOffset + 6, dir_nameSize);
dirs[i].name[destsize + 1 + dir_nameSize] = '\0';
dirs[i].numFiles = dir_numFiles;
tocOffset += dir_entrySize;
}
// parse and save files
fseek(fp, 255, SEEK_SET);
fc.fp = fp;
for (i = 0; i < numDirs; i++)
{
unsigned int dir_nameSize;
int j;
_mkdir(extractPath);
if (dirs[i].name[0])
_mkdir(dirs[i].name);
printf("Dir %s, %d files\n", dirs[i].name, dirs[i].numFiles);
dir_nameSize = strlen(dirs[i].name);
for (j = 0; j < dirs[i].numFiles; j++)
{
FILE *fpw;
int blastErr;
unsigned int file_decompressedSize = GET_UI32(toc + tocOffset + 3);
unsigned int file_compressedSize = GET_UI32(toc + tocOffset + 7);
unsigned int file_modifiedDate = GET_UI16(toc + tocOffset + 15);
unsigned int file_modifiedTime = GET_UI16(toc + tocOffset + 17);
unsigned int file_entrySize = GET_UI16(toc + tocOffset + 23);
unsigned int file_nameSize = toc[tocOffset + 29];
char *filename;
struct tm tmm;
struct _utimbuf ut;
filename = malloc(dir_nameSize + file_nameSize + 2);
strcpy(filename, dirs[i].name);
filename[dir_nameSize] = '\\';
strncpy(filename + dir_nameSize + 1, toc + tocOffset + 30, file_nameSize);
filename[dir_nameSize + 1 + file_nameSize] = '\0';
printf("%s, %d bytes, %.3f%%\n", filename, file_decompressedSize, (float)file_compressedSize / file_decompressedSize * 100.0f);
memset(&tmm, 0, sizeof(tmm));
tmm.tm_year = ( file_modifiedDate >> 9) + 80;
tmm.tm_mon = ((file_modifiedDate & 0x01E0) >> 5) - 1;
tmm.tm_mday = file_modifiedDate & 0x001F;
tmm.tm_hour = file_modifiedTime >> 11;
tmm.tm_min = (file_modifiedTime & 0x07E0) >> 5;
tmm.tm_sec = (file_modifiedTime & 0x001F) * 2;
#if defined(SHOWTIME)
printf("last modified %d %s %d %02d:%02d:%02d\n",
tmm.tm_year + 1900, months[tmm.tm_mon], tmm.tm_mday, tmm.tm_hour, tmm.tm_min, tmm.tm_sec);
#endif
tocOffset += file_entrySize;
fpw = fopen(filename, "wb");
fc.size = file_compressedSize;
blastErr = blast(inf, &fc, outf, fpw);
if (blastErr == 1)
printf("Error writing file!\n");
else if (blastErr != 0)
printf("Archive corrupt: Error decompressing, trying to continue");
fclose(fpw);
ut.actime = time(NULL);
ut.modtime = mktime(&tmm);
_utime(filename, &ut);
free(filename);
}
}
fclose(fp);
for (i = 0; i < numDirs; i++)
free(dirs[i].name);
free(dirs);
free(toc);
return 0;
}
/* Validate that a static frame's bytecode is executable by the interpreter */
void MVM_validate_static_frame(MVMThreadContext *tc, MVMStaticFrame *static_frame) {
MVMCompUnit *cu = static_frame->cu;
MVMuint32 bytecode_size = static_frame->bytecode_size;
MVMuint8 *bytecode_start = static_frame->bytecode;
MVMuint8 *bytecode_end = bytecode_start + bytecode_size;
/* current position in the bytestream */
MVMuint8 *cur_op = bytecode_start;
/* positions in the bytestream that are starts of ops and goto targets */
MVMuint8 *labels = malloc(bytecode_size);
MVMuint32 num_locals = static_frame->num_locals;
MVMuint32 branch_target;
MVMuint8 bank_num;
MVMuint8 op_num;
MVMOpInfo *op_info;
MVMuint32 operand_size;
MVMuint16 operand_target;
MVMuint32 instruction = 0;
MVMuint32 i;
unsigned char op_rw;
unsigned char op_type;
unsigned char op_flags;
MVMuint32 operand_type_var;
MVMint64 num_jumplist_labels = 0;
memset(labels, 0, bytecode_size);
/* printf("bytecode_size %d cur_op %d bytecode_end %d difference %d", bytecode_size, (int)cur_op, (int)bytecode_end, (int)(bytecode_end - cur_op)); */
while (cur_op < bytecode_end - 1) {
labels[cur_op - bytecode_start] |= MVM_val_op_boundary;
bank_num = *(cur_op++);
op_num = *(cur_op++);
operand_type_var = 0;
op_info = MVM_op_get_op((unsigned char)bank_num, (unsigned char)op_num);
if (!op_info) {
cleanup_all(tc, labels);
MVM_exception_throw_adhoc(tc,
"Bytecode validation error: non-existent operation bank %u op %u",
bank_num, op_num);
}
if (num_jumplist_labels != 0 && num_jumplist_labels-- != 0
&& (bank_num != MVM_OP_BANK_primitives || op_num != MVM_OP_goto)) {
cleanup_all(tc, labels);
MVM_exception_throw_adhoc(tc,
"jumplist op must be followed by an additional %d goto ops", num_jumplist_labels + 1);
}
/*printf("validating op %s, (%d) bank %d", op_info->name, op_num, bank_num);*/
for (i = 0; i < op_info->num_operands; i++) {
op_flags = op_info->operands[i];
op_rw = op_flags & MVM_operand_rw_mask;
op_type = op_flags & MVM_operand_type_mask;
if (op_rw == MVM_operand_literal) {
switch (op_type) {
case MVM_operand_int8: operand_size = 1; break;
case MVM_operand_int16: operand_size = 2; break;
case MVM_operand_int32: operand_size = 4; break;
case MVM_operand_int64:
operand_size = 8;
if (bank_num == MVM_OP_BANK_primitives && op_num == MVM_OP_jumplist) {
if (cur_op + operand_size > bytecode_end)
throw_past_end(tc, labels);
num_jumplist_labels = GET_I64(cur_op, 0);
if (num_jumplist_labels < 0 || num_jumplist_labels > 4294967295u) {
cleanup_all(tc, labels);
MVM_exception_throw_adhoc(tc,
"num_jumplist_labels %d out of range", num_jumplist_labels);
}
}
break;
case MVM_operand_num32: operand_size = 4; break;
case MVM_operand_num64: operand_size = 8; break;
case MVM_operand_callsite:
operand_size = 2;
if (cur_op + operand_size > bytecode_end)
throw_past_end(tc, labels);
operand_target = GET_UI16(cur_op, 0);
if (operand_target >= cu->num_callsites) {
cleanup_all(tc, labels);
MVM_exception_throw_adhoc(tc,
"Bytecode validation error: callsites index (%u) out of range; frame has %u callsites",
operand_target, cu->num_callsites);
}
break;
case MVM_operand_coderef:
operand_size = 2;
if (cur_op + operand_size > bytecode_end)
throw_past_end(tc, labels);
operand_target = GET_UI16(cur_op, 0);
if (operand_target >= cu->num_frames) {
cleanup_all(tc, labels);
MVM_exception_throw_adhoc(tc,
"Bytecode validation error: coderef index (%u) out of range; frame has %u coderefs",
operand_target, cu->num_frames);
}
break; /* reset to 0 */
case MVM_operand_str:
operand_size = 2;
if (cur_op + operand_size > bytecode_end)
throw_past_end(tc, labels);
operand_target = GET_UI16(cur_op, 0);
if (operand_target >= cu->num_strings) {
cleanup_all(tc, labels);
MVM_exception_throw_adhoc(tc,
"Bytecode validation error: strings index (%u) out of range (0-%u)",
operand_target, cu->num_strings - 1);
}
break;
case MVM_operand_ins:
operand_size = 4;
if (cur_op + operand_size > bytecode_end)
throw_past_end(tc, labels);
branch_target = GET_UI32(cur_op, 0);
if (branch_target >= bytecode_size) {
cleanup_all(tc, labels);
MVM_exception_throw_adhoc(tc,
"Bytecode validation error: branch instruction offset (%u) out of range; frame has %u bytes",
branch_target, bytecode_size);
}
labels[branch_target] |= MVM_val_branch_target;
break;
case MVM_operand_obj:
case MVM_operand_type_var:
cleanup_all(tc, labels);
MVM_exception_throw_adhoc(tc,
"Bytecode validation error: that operand type (%u) can't be a literal",
(MVMuint8)op_type);
break;
default: {
cleanup_all(tc, labels);
MVM_exception_throw_adhoc(tc,
"Bytecode validation error: non-existent operand type (%u)",
(MVMuint8)op_type);
}
}
if (cur_op + operand_size > bytecode_end)
throw_past_end(tc, labels);
}
else if (op_rw == MVM_operand_read_reg || op_rw == MVM_operand_write_reg) {
/* register operand */
operand_size = 2;
if (cur_op + operand_size > bytecode_end)
throw_past_end(tc, labels);
if (GET_REG(cur_op, 0) >= num_locals) {
cleanup_all(tc, labels);
MVM_exception_throw_adhoc(tc,
"Bytecode validation error: operand register index (%u) out of range; frame has %u locals; at byte %u",
GET_REG(cur_op, 0), num_locals, cur_op - bytecode_start);
}
if (op_type == MVM_operand_type_var) {
if (operand_type_var) {
/* XXX assume only one type variable */
if ((static_frame->local_types[GET_REG(cur_op, 0)] << 3) != operand_type_var) {
cleanup_all(tc, labels);
MVM_exception_throw_adhoc(tc,
"Bytecode validation error: inconsistent operand types %d and %d to op '%s' with a type variable, at instruction %d",
static_frame->local_types[GET_REG(cur_op, 0)], operand_type_var >> 3, op_info->name, instruction);
}
}
else {
operand_type_var = (static_frame->local_types[GET_REG(cur_op, 0)] << 3);
}
}
else if ((static_frame->local_types[GET_REG(cur_op, 0)] << 3) != op_type) {
static void bytecode_dump_frame_internal(MVMThreadContext *tc, MVMStaticFrame *frame, MVMSpeshCandidate *maybe_candidate, MVMuint8 *frame_cur_op, char ***frame_lexicals, char **oo, MVMuint32 *os, MVMuint32 *ol) {
/* since "references" are not a thing in C, keep a local copy of these
* and update the passed-in pointers at the end of the function */
char *o = *oo;
MVMuint32 s = *os;
MVMuint32 l = *ol;
MVMuint32 i, j, k;
/* mostly stolen from validation.c */
MVMStaticFrame *static_frame = frame;
MVMuint32 bytecode_size = maybe_candidate ? maybe_candidate->bytecode_size : static_frame->body.bytecode_size;
MVMuint8 *bytecode_start = maybe_candidate ? maybe_candidate->bytecode : static_frame->body.bytecode;
MVMuint8 *bytecode_end = bytecode_start + bytecode_size;
/* current position in the bytestream */
MVMuint8 *cur_op = bytecode_start;
/* positions in the bytestream that are starts of ops and goto targets */
MVMuint8 *labels = MVM_calloc(1, bytecode_size);
MVMuint32 *jumps = MVM_calloc(1, sizeof(MVMuint32) * bytecode_size);
char **lines = MVM_malloc(sizeof(char *) * bytecode_size);
MVMuint32 *linelocs = MVM_malloc(sizeof(MVMuint32) * bytecode_size);
MVMuint32 lineno = 0;
MVMuint32 lineloc;
MVMuint16 op_num;
const MVMOpInfo *op_info;
MVMuint32 operand_size = 0;
unsigned char op_rw;
unsigned char op_type;
unsigned char op_flags;
MVMOpInfo tmp_extop_info;
/* stash the outer output buffer */
MVMuint32 sP = s;
MVMuint32 lP = l;
char *oP = o;
char *tmpstr;
char mark_this_line = 0;
MVMCompUnit *cu = static_frame->body.cu;
while (cur_op < bytecode_end - 1) {
/* allocate a line buffer */
s = 200;
l = 0;
o = MVM_calloc(s, sizeof(char));
lineloc = cur_op - bytecode_start;
/* mark that this line starts at this point in the bytestream */
linelocs[lineno] = lineloc;
/* mark that this point in the bytestream is an op boundary */
labels[lineloc] |= MVM_val_op_boundary;
mark_this_line = 0;
if (frame_cur_op) {
if (frame_cur_op == cur_op || frame_cur_op == cur_op + 2) {
mark_this_line = 1;
}
}
if (mark_this_line) {
a("-> ");
} else {
a(" ");
}
op_num = *((MVMint16 *)cur_op);
cur_op += 2;
if (op_num < MVM_OP_EXT_BASE) {
op_info = MVM_op_get_op(op_num);
if (op_info)
a("%-18s ", op_info->name);
else
a("invalid OP ");
}
else {
MVMint16 ext_op_num = op_num - MVM_OP_EXT_BASE;
if (0 <= ext_op_num && ext_op_num < cu->body.num_extops) {
MVMExtOpRecord r = cu->body.extops[ext_op_num];
MVMuint8 j;
memset(&tmp_extop_info, 0, sizeof(MVMOpInfo));
tmp_extop_info.name = MVM_string_utf8_encode_C_string(tc, r.name);
memcpy(tmp_extop_info.operands, r.operand_descriptor, 8);
for (j = 0; j < 8; j++)
if (tmp_extop_info.operands[j])
tmp_extop_info.num_operands++;
else
break;
op_info = &tmp_extop_info;
a("%-12s ", tmp_extop_info.name);
MVM_free((void *)tmp_extop_info.name);
tmp_extop_info.name = NULL;
}
else {
a("Extension op %d out of range", (int)op_num);
}
}
if (!op_info)
continue;
for (i = 0; i < op_info->num_operands; i++) {
if (i) a(", ");
op_flags = op_info->operands[i];
op_rw = op_flags & MVM_operand_rw_mask;
op_type = op_flags & MVM_operand_type_mask;
if (op_rw == MVM_operand_literal) {
switch (op_type) {
case MVM_operand_int8:
operand_size = 1;
a("%"PRId8, GET_I8(cur_op, 0));
break;
case MVM_operand_int16:
operand_size = 2;
a("%"PRId16, GET_I16(cur_op, 0));
break;
case MVM_operand_int32:
operand_size = 4;
a("%"PRId32, GET_I32(cur_op, 0));
break;
case MVM_operand_int64:
operand_size = 8;
a("%"PRId64, MVM_BC_get_I64(cur_op, 0));
break;
case MVM_operand_uint8:
operand_size = 1;
a("%"PRIu8, GET_I8(cur_op, 0));
break;
case MVM_operand_uint16:
operand_size = 2;
a("%"PRIu16, GET_I16(cur_op, 0));
break;
case MVM_operand_uint32:
operand_size = 4;
a("%"PRIu32, GET_I32(cur_op, 0));
break;
case MVM_operand_uint64:
operand_size = 8;
a("%"PRIu64, MVM_BC_get_I64(cur_op, 0));
break;
case MVM_operand_num32:
operand_size = 4;
a("%f", GET_N32(cur_op, 0));
break;
case MVM_operand_num64:
operand_size = 8;
a("%f", MVM_BC_get_N64(cur_op, 0));
break;
case MVM_operand_callsite:
operand_size = 2;
a("Callsite_%"PRIu16, GET_UI16(cur_op, 0));
break;
case MVM_operand_coderef:
operand_size = 2;
a("Frame_%"PRIu16, GET_UI16(cur_op, 0));
break;
case MVM_operand_str:
operand_size = 4;
if (GET_UI32(cur_op, 0) < cu->body.num_strings) {
tmpstr = MVM_string_utf8_encode_C_string(
tc, MVM_cu_string(tc, cu, GET_UI32(cur_op, 0)));
/* XXX C-string-literal escape the \ and '
and line breaks and non-ascii someday */
a("'%s'", tmpstr);
MVM_free(tmpstr);
}
else
a("invalid string index: %d", GET_UI32(cur_op, 0));
break;
case MVM_operand_ins:
operand_size = 4;
/* luckily all the ins operands are at the end
of op operands, so I can wait to resolve the label
to the end. */
if (GET_UI32(cur_op, 0) < bytecode_size) {
labels[GET_UI32(cur_op, 0)] |= MVM_val_branch_target;
jumps[lineno] = GET_UI32(cur_op, 0);
}
break;
case MVM_operand_obj:
/* not sure what a literal object is */
operand_size = 4;
break;
case MVM_operand_spesh_slot:
operand_size = 2;
a("sslot(%d)", GET_UI16(cur_op, 0));
break;
default:
fprintf(stderr, "what is an operand of type %d??\n", op_type);
abort(); /* never reached, silence compiler warnings */
}
}
else if (op_rw == MVM_operand_read_reg || op_rw == MVM_operand_write_reg) {
/* register operand */
MVMuint8 frame_has_inlines = maybe_candidate && maybe_candidate->num_inlines ? 1 : 0;
MVMuint16 *local_types = frame_has_inlines ? maybe_candidate->local_types : frame->body.local_types;
MVMuint16 num_locals = frame_has_inlines ? maybe_candidate->num_locals : frame->body.num_locals;
operand_size = 2;
a("loc_%u_%s", GET_REG(cur_op, 0),
get_typename(local_types[GET_REG(cur_op, 0)]));
}
else if (op_rw == MVM_operand_read_lex || op_rw == MVM_operand_write_lex) {
/* lexical operand */
MVMuint16 idx, frames, m;
MVMStaticFrame *applicable_frame = static_frame;
operand_size = 4;
idx = GET_UI16(cur_op, 0);
frames = GET_UI16(cur_op, 2);
m = frames;
while (m > 0) {
applicable_frame = applicable_frame->body.outer;
m--;
}
/* inefficient, I know. should use a hash. */
for (m = 0; m < cu->body.num_frames; m++) {
if (get_frame(tc, cu, m) == applicable_frame) {
char *lexname = frame_lexicals ? frame_lexicals[m][idx] : "lex??";
a("lex_Frame_%u_%s_%s", m, lexname,
get_typename(applicable_frame->body.lexical_types[idx]));
}
}
}
cur_op += operand_size;
}
lines[lineno++] = o;
}
{
MVMuint32 *linelabels = MVM_calloc(lineno, sizeof(MVMuint32));
MVMuint32 byte_offset = 0;
MVMuint32 line_number = 0;
MVMuint32 label_number = 1;
MVMuint32 *annotations = MVM_calloc(lineno, sizeof(MVMuint32));
for (; byte_offset < bytecode_size; byte_offset++) {
if (labels[byte_offset] & MVM_val_branch_target) {
/* found a byte_offset where a label should be.
now crawl up through the lines to find which line starts there */
while (line_number < lineno && linelocs[line_number] != byte_offset) line_number++;
if (line_number < lineno)
linelabels[line_number] = label_number++;
}
}
o = oP;
l = lP;
s = sP;
i = 0;
/* resolve annotation line numbers */
for (j = 0; j < frame->body.num_annotations; j++) {
MVMuint32 ann_offset = GET_UI32(frame->body.annotations_data, j*12);
for (; i < lineno; i++) {
if (linelocs[i] == ann_offset) {
annotations[i] = j + 1;
break;
}
}
}
for (j = 0; j < lineno; j++) {
if (annotations[j]) {
MVMuint16 shi = GET_UI16(frame->body.annotations_data + 4, (annotations[j] - 1)*12);
tmpstr = MVM_string_utf8_encode_C_string(
tc, MVM_cu_string(tc, cu, shi < cu->body.num_strings ? shi : 0));
a(" annotation: %s:%u\n", tmpstr, GET_UI32(frame->body.annotations_data, (annotations[j] - 1)*12 + 8));
MVM_free(tmpstr);
}
if (linelabels[j])
a(" label_%u:\n", linelabels[j]);
a("%05d %s", j, lines[j]);
MVM_free(lines[j]);
if (jumps[j]) {
/* horribly inefficient for large frames. again, should use a hash */
line_number = 0;
while (line_number < lineno && linelocs[line_number] != jumps[j]) line_number++;
if (line_number < lineno)
a("label_%u(%05u)", linelabels[line_number], line_number);
else
a("label (invalid: %05u)", jumps[j]);
}
a("\n");
}
MVM_free(lines);
MVM_free(jumps);
MVM_free(linelocs);
MVM_free(linelabels);
MVM_free(labels);
MVM_free(annotations);
}
*oo = o;
*os = s;
*ol = l;
}