diff --git a/CMakeLists.txt b/CMakeLists.txt
index 68f4395..94f620e 100644
--- a/CMakeLists.txt
+++ b/CMakeLists.txt
@@ -1,8 +1,8 @@
 cmake_minimum_required(VERSION 3.9)
 set(MT "mtrace-ng")
-project(${MT})
+project(${MT} C)
 
-set(MT_VERSION_STRING "0.7")
+set(MT_VERSION_STRING "0.8")
 
 option(DISABLE_CLIENT "whether to disable client support" OFF)
 
@@ -10,7 +10,7 @@ set(default_build_type "Release")
 if (NOT CMAKE_BUILD_TYPE AND NOT CMAKE_CONFIGURATION_TYPES)
 	set(CMAKE_BUILD_TYPE "${default_build_type}" CACHE STRING "Choose type of build" FORCE)
 	set_property(CACHE CMAKE_BUILD_TYPE PROPERTY STRINGS
-		"Debug" "Relase" "LTO")
+		"Debug" "Release" "LTO")
 endif()
 
 include(CheckFunctionExists)
@@ -45,10 +45,13 @@ include_directories(
  "${PROJECT_SOURCE_DIR}/sysdeps"
 )
 
-set(CMAKE_C_FLAGS_DEBUG "-O0 -ggdb -Wall")
-set(CMAKE_C_FLAGS_RELEASE "-O3 -Wall -DNDEBUG")
+set(CMAKE_C_FLAGS_DEBUG "${CMAKE_C_FLAGS_DEBUG} -O0 -fno-omit-frame-pointer -fsanitize=address -fsanitize=undefined -D__FORITFY_SOURCE=2 -rdynamic")
+set(CMAKE_LINKER_FLAGS_DEBUG "${CMAKE_LINKER_FLAGS_DEBUG} -fno-omit-frame-pointer -fsanitize=address -fsanitize=undefined")
+
 set(CMAKE_C_FLAGS_LTO "${CMAKE_C_FLAGS_RELEASE} -flto")
-set(CMAKE_EXE_LINKER_FLAGS_LTO "-flto")
+set(CMAKE_EXE_LINKER_FLAGS_LTO "${CMAKE_LINKER_FLAGS_RELEASE} -flto")
+
+add_compile_options(-Wall -Wextra -Werror -Werror -Wno-implicit-fallthrough)
 
 check_ipo_supported(RESULT IPO)
 #if (IPO)
diff --git a/backend.h b/backend.h
index 905d122..b33cde0 100644
--- a/backend.h
+++ b/backend.h
@@ -56,8 +56,8 @@ int trace_set_options(struct task *task);
 /* make the forked process traceable */
 void trace_me(void);
 
-/* stop tracing a task.  */
-int untrace_task(struct task *task, int signum);
+/* stop tracing of a task. */
+int untrace_task(struct task *task);
 
 /* Called when mtrace-ng needs to attach to task */
 int trace_attach(struct task *task);
diff --git a/breakpoint.c b/breakpoint.c
index 910f49d..75665b0 100644
--- a/breakpoint.c
+++ b/breakpoint.c
@@ -70,28 +70,37 @@ static void enable_sw_breakpoint(struct task *task, struct breakpoint *bp)
 {
 	static unsigned char break_insn[] = BREAKPOINT_VALUE;
 
+	assert(task->stopped);
+	assert(bp->sw == 0);
+	assert(bp->hw == 0);
+
 	debug(DEBUG_PROCESS, "pid=%d, addr=%#lx", task->pid, bp->addr);
 
 	copy_from_to_proc(task, bp->addr, break_insn, bp->orig_value, BREAKPOINT_LENGTH);
 
 	bp->break_insn = !memcmp(break_insn, bp->orig_value, BREAKPOINT_LENGTH);
 	bp->was_hw = 0;
+	bp->sw = 1;
+	bp->enabled = 1;
+
+	if (bp->break_insn)
+		fprintf(stderr, "!!!break insn pid=%d, addr=%#lx\n", task->pid, bp->addr);
 }
 
-static void disable_sw_breakpoint(struct task *task, const struct breakpoint *bp)
+static void disable_sw_breakpoint(struct task *task, struct breakpoint *bp)
 {
 	debug(DEBUG_PROCESS, "pid=%d, addr=%lx", task->pid, bp->addr);
 
+	assert(task->stopped);
+	assert(bp->sw);
+	assert(bp->hw == 0);
+
 	copy_to_proc(task, bp->addr, bp->orig_value, BREAKPOINT_LENGTH);
+
+	bp->sw = 0;
+	bp->enabled = 0;
 }
 
-int breakpoint_on_hit(struct task *task, struct breakpoint *bp)
-{
-	if (bp->on_hit)
-		return (bp->on_hit)(task, bp);
-
-	return 0;
-}
 
 struct breakpoint *breakpoint_find(struct task *task, arch_addr_t addr)
 {
@@ -109,6 +118,7 @@ static void enable_hw_bp(struct task *task, struct breakpoint *bp)
 
 	assert(bp->type != BP_SW);
 	assert(bp->type < BP_HW || task->hw_bp[slot] == NULL);
+	assert(bp->sw == 0);
 
 	task->hw_bp[slot] = bp;
 
@@ -120,6 +130,8 @@ static void disable_hw_bp(struct task *task, struct breakpoint *bp)
 {
 	unsigned int slot = bp->hw_bp_slot;
 
+	assert(bp->sw == 0);
+
 	if (unlikely(!task->hw_bp[slot]))
 		return;
 
@@ -185,13 +197,16 @@ void reorder_hw_bp(struct task *task)
 		}
 	}
 
-	if (!n)
+	if (n < HW_BREAKPOINTS - 1) {
+		for(i = 0; i < n; ++i)
+			bp_list[i]->hwcnt = 0;
 		return;
+	}
 
 	qsort(bp_list, n, sizeof(*bp_list), hw_bp_sort);
 
 	for(i = 0; i < n; ++i)
-		bp_list[i]->hwcnt = (i < HW_BREAKPOINTS - 1) ? BP_REORDER_THRESHOLD >> (i + 4) : 0;
+		bp_list[i]->hwcnt = 0;
 
 	if (n > HW_BREAKPOINTS - 1) 
 		n = HW_BREAKPOINTS - 1;
@@ -205,9 +220,11 @@ void reorder_hw_bp(struct task *task)
 			assert((hw_bp_set & 1 << bp_list[i]->hw_bp_slot) == 0);
 
 			hw_bp_set |= 1 << bp_list[i]->hw_bp_slot;
+
+			continue;
 		}
-		else
-			*p++ = bp_list[i];
+
+		*p++ = bp_list[i];
 	}
 
 	if (p == bp_list)
@@ -231,6 +248,7 @@ void reorder_hw_bp(struct task *task)
 		if (leader->hw_bp[i])
 			hw2sw_bp(leader, leader->hw_bp[i]);
 
+		bp->enabled = 1;
 		bp->hw_bp_slot = i;
 		bp->hw = 1;
 		bp->was_hw = 1;
@@ -366,6 +384,7 @@ struct breakpoint *breakpoint_new_ext(struct task *task, arch_addr_t addr, struc
 	bp->ext = ext;
 	bp->refcnt = 1;
 	bp->count = 0;
+	bp->sw = 0;
 #if HW_BREAKPOINTS > 1
 	bp->hwcnt = 0;
 #endif
@@ -432,7 +451,6 @@ void breakpoint_enable(struct task *task, struct breakpoint *bp)
 #endif
 		bp->hw = 0;
 		enable_sw_breakpoint(task, bp);
-		bp->enabled = 1;
 	}
 }
 
@@ -467,7 +485,6 @@ void breakpoint_disable(struct task *task, struct breakpoint *bp)
 		}
 #endif
 		disable_sw_breakpoint(task, bp);
-		bp->enabled = 0;
 	}
 }
 
@@ -529,6 +546,8 @@ void breakpoint_delete(struct task *task, struct breakpoint *bp)
 
 static int enable_nonlocked_bp_cb(unsigned long key, const void *value, void *data)
 {
+	(void)key;
+
 	struct breakpoint *bp = (struct breakpoint *)value;
 	struct task *leader = (struct task *)data;
 
@@ -550,6 +569,8 @@ void breakpoint_enable_all_nonlocked(struct task *leader)
 
 static int disable_nonlocked_bp_cb(unsigned long key, const void *value, void *data)
 {
+	(void)key;
+
 	struct breakpoint *bp = (struct breakpoint *)value;
 	struct task *leader = (struct task *)data;
 
@@ -567,7 +588,6 @@ void breakpoint_disable_all_nonlocked(struct task *leader)
 
 	if (leader->breakpoints)
 		dict_apply_to_all(leader->breakpoints, disable_nonlocked_bp_cb, leader);
-	leader->attached = 1;
 }
 
 static int enable_bp_cb(unsigned long key, const void *value, void *data)
@@ -575,6 +595,8 @@ static int enable_bp_cb(unsigned long key, const void *value, void *data)
 	struct breakpoint *bp = (struct breakpoint *)value;
 	struct task *leader = (struct task *)data;
 
+	(void)key;
+
 	debug(DEBUG_FUNCTION, "pid=%d", leader->pid);
 
 	breakpoint_enable(leader, bp);
@@ -595,6 +617,8 @@ static int disable_bp_cb(unsigned long key, const void *value, void *data)
 	struct breakpoint *bp = (struct breakpoint *)value;
 	struct task *leader = (struct task *)data;
 
+	(void)key;
+
 	debug(DEBUG_FUNCTION, "pid=%d", leader->pid);
 
 	breakpoint_disable(leader, bp);
@@ -608,7 +632,30 @@ void breakpoint_disable_all(struct task *leader)
 
 	if (leader->breakpoints)
 		dict_apply_to_all(leader->breakpoints, disable_bp_cb, leader);
-	leader->attached = 1;
+}
+
+static int invalidate_bp_cb(unsigned long key, const void *value, void *data)
+{
+	struct breakpoint *bp = (struct breakpoint *)value;
+	struct task *leader = (struct task *)data;
+
+	(void)key;
+
+	debug(DEBUG_FUNCTION, "pid=%d", leader->pid);
+
+	bp->enabled = 0;
+	bp->sw = 0;
+	bp->hw = 0;
+
+	return 0;
+}
+
+void breakpoint_invalidate_all(struct task *leader)
+{
+	debug(DEBUG_FUNCTION, "pid=%d", leader->pid);
+
+	if (leader->breakpoints)
+		dict_apply_to_all(leader->breakpoints, invalidate_bp_cb, leader);
 }
 
 static int destroy_breakpoint_cb(unsigned long key, const void *value, void *data)
@@ -616,6 +663,8 @@ static int destroy_breakpoint_cb(unsigned long key, const void *value, void *dat
 	struct breakpoint *bp = (struct breakpoint *)value;
 	struct task *leader = (struct task *)data;
 
+	(void)key;
+
 	breakpoint_delete(leader, bp);
 	return 0;
 }
@@ -624,7 +673,9 @@ void breakpoint_clear_all(struct task *leader)
 {
 	if (leader->breakpoints) {
 		dict_apply_to_all(leader->breakpoints, &destroy_breakpoint_cb, leader);
+#if HW_BREAKPOINTS > 0
 		assert(leader->hw_bp_num == 0);
+#endif
 		dict_clear(leader->breakpoints);
 		leader->breakpoints = NULL;
 	}
@@ -643,6 +694,8 @@ static int clone_single_cb(unsigned long key, const void *value, void *data)
 	struct task *new_task = (struct task *)data;
 	size_t ext = bp->ext;
 
+	(void)key;
+
 	if (bp->deleted)
 		return 0;
 
@@ -665,6 +718,7 @@ static int clone_single_cb(unsigned long key, const void *value, void *data)
 	new_bp->enabled = bp->enabled;
 	new_bp->locked = bp->locked;
 	new_bp->hw = bp->hw;
+	new_bp->sw = bp->sw;
 	new_bp->type = bp->type;
 	new_bp->ext = ext;
 	new_bp->refcnt = 1;
@@ -695,7 +749,8 @@ static int clone_single_cb(unsigned long key, const void *value, void *data)
 	}
 	else
 #endif
-	memcpy(new_bp->orig_value, bp->orig_value, sizeof(bp->orig_value));
+	if (new_bp->sw)
+		memcpy(new_bp->orig_value, bp->orig_value, sizeof(bp->orig_value));
 
 	if (ext)
 		memcpy((void *)new_bp + ext, (void *)bp + ext, ext);
@@ -732,6 +787,8 @@ int breakpoint_put(struct breakpoint *bp)
 		if (--bp->refcnt)
 			return 0;
 
+		assert(bp->enabled == 0);
+
 		free(bp);
 	}
 	return 1;
diff --git a/breakpoint.h b/breakpoint.h
index b391c4a..311b7e4 100644
--- a/breakpoint.h
+++ b/breakpoint.h
@@ -50,6 +50,7 @@ struct breakpoint {
 	unsigned int enabled:1;
 	unsigned int locked:1;
 	unsigned int deleted:1;
+	unsigned int sw:1;
 	unsigned int hw:1;
 	unsigned int was_hw:1;
 	unsigned int break_insn:1;
@@ -69,9 +70,6 @@ struct breakpoint {
 /* setup the basic breakpoint support for a given leader */
 void breakpoint_setup(struct task *leader);
 
-/* Call on-hit handler of BP, if any is set.  */
-int breakpoint_on_hit(struct task *task, struct breakpoint *bp);
-
 /* get a new breakpoint structure. */
 struct breakpoint *breakpoint_new(struct task *task, arch_addr_t addr, struct library_symbol *libsym, int type);
 
@@ -92,6 +90,7 @@ void breakpoint_disable(struct task *task, struct breakpoint *bp);
 
 void breakpoint_enable_all(struct task *leader);
 void breakpoint_disable_all(struct task *leader);
+void breakpoint_invalidate_all(struct task *leader);
 void breakpoint_enable_all_nonlocked(struct task *leader);
 void breakpoint_disable_all_nonlocked(struct task *leader);
 void breakpoint_clear_all(struct task *leader);
@@ -105,10 +104,14 @@ void disable_scratch_hw_bp(struct task *task, struct breakpoint *bp);
 #else
 static inline void enable_scratch_hw_bp(struct task *task, struct breakpoint *bp)
 {
+	(void)task;
+	(void)bp;
 }
 
 static inline void disable_scratch_hw_bp(struct task *task, struct breakpoint *bp)
 {
+	(void)task;
+	(void)bp;
 }
 #endif
 
@@ -120,10 +123,12 @@ void breakpoint_hw_destroy(struct task *task);
 #else
 static inline void breakpoint_hw_clone(struct task *task)
 {
+	(void)task;
 }
 
 static inline void breakpoint_hw_destroy(struct task *task)
 {
+	(void)task;
 }
 #endif
 
diff --git a/client/binfile.c b/client/binfile.c
index b1536fe..8a8b780 100644
--- a/client/binfile.c
+++ b/client/binfile.c
@@ -151,7 +151,7 @@ struct rb_sym *bin_file_lookup(struct bin_file *binfile, bfd_vma addr, unsigned
 		if (!name || !*name)
 			name = "?";
 		else {
-			alloc = bfd_demangle(binfile->abfd, name, DMGL_TYPES | DMGL_VERBOSE | DMGL_ANSI | DMGL_PARAMS);
+			alloc = bfd_demangle(binfile->abfd, name, DMGL_ANSI | DMGL_PARAMS | DMGL_RET_DROP | DMGL_AUTO);
 			if (alloc)
 				name = alloc;
 		}
@@ -289,10 +289,9 @@ void bin_file_sym_put(struct rb_sym *sym)
 	if (!--sym->refcnt) {
 		free(sym->sym);
 
-		if (!binfile)
-			return;
-
-		rb_erase(&sym->node, &binfile->sym_table);
+		if (binfile)
+			rb_erase(&sym->node, &binfile->sym_table);
+		free(sym);
 	}
 	bin_file_put(binfile);
 }
diff --git a/client/client.c b/client/client.c
index 308e526..e39313e 100644
--- a/client/client.c
+++ b/client/client.c
@@ -364,8 +364,10 @@ static void client_remove_process(struct process *process)
 {
 	process = pid_rb_delete(&pid_table, process->pid);
 
-	if (process)
+	if (process) {
+		process_reset(process);
 		free(process);
+	}
 }
 
 
@@ -487,14 +489,15 @@ static int client_func(void)
 			process_about_exit(process);
 			break;
 		case MT_EXIT:
-			process_exit(process);
+			if (process_exit(process))
+				client_remove_process(process);
 			break;
 		case MT_NOFOLLOW:
-			process_reset(process);
 			client_remove_process(process);
 			break;
 		case MT_SCAN:
-			process_scan(process, payload, mt_msg.payload_len);
+			if (process_scan(process, payload, mt_msg.payload_len))
+				client_remove_process(process);
 			break;
 		case MT_ADD_MAP:
 			process_add_map(process, payload, mt_msg.payload_len);
@@ -503,7 +506,8 @@ static int client_func(void)
 			process_del_map(process, payload, mt_msg.payload_len);
 			break;
 		case MT_DETACH:
-			process_detach(process);
+			if (process_detach(process))
+				client_remove_process(process);
 			break;
 		default:
 			fatal("protocol violation 0x%08x", mt_msg.operation);
@@ -659,8 +663,7 @@ static void signal_exit(int sig)
 	signal(SIGINT, SIG_IGN);
 	signal(SIGTERM, SIG_IGN);
 
-	if (write(pipefd[1], &signum, 1) == -1)
-		;
+	write(pipefd[1], &signum, 1);
 }
 
 static int scan_process(struct process *process)
@@ -761,6 +764,8 @@ int client_start(void)
 
 void *client_thread(void *unused)
 {
+	(void)unused;
+
 	if (options.interactive) {
 		ioevent_add_input(client_fd, client_func);
 
@@ -824,6 +829,7 @@ int client_stop(void)
 {
 	if (thread) {
 		thread_join(thread);
+		free(thread);
 		thread = NULL;
 	}
 
diff --git a/client/process.c b/client/process.c
index 7cf985b..bc969d8 100644
--- a/client/process.c
+++ b/client/process.c
@@ -424,6 +424,7 @@ static void stack_put(struct rb_stack *stack_node)
 			for(i = 0; i < stack->entries; ++i)
 				bin_file_sym_put(stack->syms[i]);
 
+			free(stack->addrs);
 			free(stack->syms);
 		}
 		free(stack);
@@ -479,7 +480,7 @@ static struct rb_stack *stack_clone(struct process *process, struct rb_stack *st
 	return this;
 }
 
-static struct rb_stack *stack_add(struct process *process, unsigned int pid, void *addrs, uint32_t stack_size, enum mt_operation operation)
+static struct rb_stack *stack_add(struct process *process, void *addrs, uint32_t stack_size, enum mt_operation operation)
 {
 	struct rb_root *root = &process->stack_table;
 	struct rb_node **new = &(root->rb_node), *parent = NULL;
@@ -775,6 +776,8 @@ void process_del_map(struct process *process, void *payload, uint32_t payload_le
 	uint64_t size = process->val64(mt_map->size);
 	struct list_head *it;
 
+	(void)payload_len;
+
 	list_for_each(it, &process->map_list) {
 		struct map *map = container_of(it, struct map, list);
 
@@ -1144,6 +1147,8 @@ static void process_dump(struct process *process, int (*sortby)(const struct rb_
 
 static int skip_none(struct rb_stack *stack)
 {
+	(void)stack;
+
 	return 0;
 }
 
@@ -1207,7 +1212,7 @@ void process_dump_stacks(struct process *process, const char *outfile, int lflag
 	process_dump(process, sort_allocations, skip_none, outfile, lflag);
 }
 
-void *process_scan(struct process *process, void *leaks, uint32_t payload_len)
+int process_scan(struct process *process, void *leaks, uint32_t payload_len)
 {
 	unsigned int new = 0;
 	unsigned long n = payload_len / process->ptr_size;
@@ -1250,10 +1255,12 @@ void *process_scan(struct process *process, void *leaks, uint32_t payload_len)
 	dump_printf("leaks total: %lu\n", process->leaks);
 	dump_flush();
 
-	if (!options.interactive)
+	if (!options.interactive) {
 		process_dump_sortby(process);
+		return 1;
+	}
 
-	return leaks;
+	return 0;
 }
 
 static inline unsigned long roundup_mask(unsigned long val, unsigned long mask)
@@ -1430,7 +1437,7 @@ void process_free(struct process *process, struct mt_msg *mt_msg, void *payload)
 
 		if (stack_size) {
 			if (!is_sane(block, mt_msg->operation)) {
-				struct rb_stack *stack = stack_add(process, process->pid, stack_data, stack_size, mt_msg->operation);
+				struct rb_stack *stack = stack_add(process, stack_data, stack_size, mt_msg->operation);
 
 				stack->n_mismatched++;
 				stack->tsc = process->tsc++;
@@ -1462,7 +1469,7 @@ void process_free(struct process *process, struct mt_msg *mt_msg, void *payload)
 			}
 
 			if (stack_size) {
-				struct rb_stack *stack = stack_add(process, process->pid, stack_data, stack_size, mt_msg->operation);
+				struct rb_stack *stack = stack_add(process, stack_data, stack_size, mt_msg->operation);
 
 				stack->n_badfree++;
 				stack->tsc = process->tsc++;
@@ -1477,6 +1484,8 @@ void process_realloc_done(struct process *process, struct mt_msg *mt_msg, void *
 	unsigned int pid;
 	struct list_head *it;
 
+	(void)mt_msg;
+
 	if (!process->tracing)
 		return;
 
@@ -1508,7 +1517,7 @@ void process_realloc_done(struct process *process, struct mt_msg *mt_msg, void *
 	}
 
 	if (unlikely(options.kill)) {
-		fprintf(stderr, ">>> unexpected realloc done pid: %u\n", pid);
+		fprintf(stderr, ">>> unexpected realloc done pid: %u ptr: %#lx\n", pid, ptr);
 		abort();
 	}
 	return;
@@ -1562,7 +1571,7 @@ void process_alloc(struct process *process, struct mt_msg *mt_msg, void *payload
 		process_rb_delete_block(process, block);
 	}
 
-	struct rb_stack *stack = stack_add(process, process->pid, stack_data, stack_size, mt_msg->operation);
+	struct rb_stack *stack = stack_add(process, stack_data, stack_size, mt_msg->operation);
 
 	if (process_rb_insert_block(process, ptr, size, stack, 0, mt_msg->operation)) {
 		fprintf(stderr, "process_rb_insert_block failed\n");
@@ -1636,14 +1645,17 @@ void process_dump_sortby(struct process *process)
 	}
 }
 
-void process_exit(struct process *process)
+int process_exit(struct process *process)
 {
 	process_set_status(process, MT_PROCESS_EXIT);
 
-	if (!options.interactive)
+	if (!options.interactive) {
 		process_dump_sortby(process);
-	else
-		fprintf(stderr, "+++ process %d exited +++\n", process->pid);
+		return 1;
+	}
+
+	fprintf(stderr, "+++ process %d exited\n", process->pid);
+	return 0;
 }
 
 void process_about_exit(struct process *process)
@@ -1656,17 +1668,25 @@ void process_about_exit(struct process *process)
 	client_send_msg(process, MT_ABOUT_EXIT, NULL, 0);
 }
 
-void process_detach(struct process *process)
+int process_detach(struct process *process)
 {
+	int ret = 0;
+
 	process_set_status(process, MT_PROCESS_DETACH);
 
-	if (options.auto_scan)
+	if (options.auto_scan) {
 		process_leaks_scan(process, SCAN_ALL);
-	else
-	if (!options.interactive)
-		process_dump_sortby(process);
+	}
+	else {
+		if (!options.interactive) {
+			process_dump_sortby(process);
+			ret = 1;
+		}
+	}
 
 	client_send_msg(process, MT_DETACH, NULL, 0);
+
+	return ret;
 }
 
 void process_set_status(struct process *process, enum process_status status)
diff --git a/client/process.h b/client/process.h
index 16c1206..0bafae9 100644
--- a/client/process.h
+++ b/client/process.h
@@ -90,15 +90,15 @@ void process_set_status(struct process *process, enum process_status status);
 void process_start_input(struct process *process);
 void process_stop_input(struct process *process);
 void process_about_exit(struct process *process);
-void process_exit(struct process *process);
+int process_exit(struct process *process);
 void process_status(struct process *process);
-void *process_scan(struct process *curr, void *leaks, uint32_t payload_len);
+int process_scan(struct process *curr, void *leaks, uint32_t payload_len);
 void process_alloc(struct process *process, struct mt_msg *msg, void *payload);
 void process_free(struct process *process, struct mt_msg *msg, void *payload);
 void process_munmap(struct process *process, struct mt_msg *msg, void *payload);
 void process_add_map(struct process *process, void *payload, uint32_t payload_len);
 void process_del_map(struct process *process, void *payload, uint32_t payload_len);
-void process_detach(struct process *process);
+int process_detach(struct process *process);
 void process_realloc_done(struct process *process, struct mt_msg *mt_msg, void *payload);
 
 unsigned long process_leaks_scan(struct process *process, int mode);
diff --git a/client/readline.c b/client/readline.c
index c8f79d3..a8972af 100644
--- a/client/readline.c
+++ b/client/readline.c
@@ -90,35 +90,35 @@ static const char *outfile;
 static int quit;
 
 static struct cmd_opt dump_opts[] = {
-	{ "allocations", 2, process_dump_sort_allocations, "sort by number of open allocations" },
-	{ "average", 2, process_dump_sort_average, "sort by average allocation of bytes (usage / allocations)" },
-	{ "badfree", 1, process_dump_sort_badfree, "sort by number of badfree releases" },
-	{ "bytes-leaked", 1, process_dump_sort_bytes_leaked, "sort by number of leaked bytes" },
-	{ "leaks", 1, process_dump_sort_leaks, "sort by number of detected leaks" },
-	{ "mismatched", 1, process_dump_sort_mismatched, "sort by number of mismatched releases" },
-	{ "stacks", 1, process_dump_stacks, "dump all stack sort by number of total allocations" },
-	{ "total", 2, process_dump_sort_total, "sort by number of total allocations" },
-	{ "tsc", 2, process_dump_sort_tsc, "sort by time stamp counter" },
-	{ "usage", 1, process_dump_sort_usage, "sort by number of bytes" },
-	{ NULL, 0, NULL, "\n use > to dump the output into a file" },
+	{ "allocations", 2, process_dump_sort_allocations, "sort by number of open allocations", NULL, NULL },
+	{ "average", 2, process_dump_sort_average, "sort by average allocation of bytes (usage / allocations)", NULL, NULL },
+	{ "badfree", 1, process_dump_sort_badfree, "sort by number of badfree releases", NULL, NULL },
+	{ "bytes-leaked", 1, process_dump_sort_bytes_leaked, "sort by number of leaked bytes", NULL, NULL },
+	{ "leaks", 1, process_dump_sort_leaks, "sort by number of detected leaks", NULL, NULL },
+	{ "mismatched", 1, process_dump_sort_mismatched, "sort by number of mismatched releases", NULL, NULL },
+	{ "stacks", 1, process_dump_stacks, "dump all stack sort by number of total allocations", NULL, NULL },
+	{ "total", 2, process_dump_sort_total, "sort by number of total allocations", NULL, NULL },
+	{ "tsc", 2, process_dump_sort_tsc, "sort by time stamp counter", NULL, NULL },
+	{ "usage", 1, process_dump_sort_usage, "sort by number of bytes", NULL, NULL },
+	{ NULL, 0, NULL, "\n use > to dump the output into a file", NULL, NULL },
 };
 
 static struct cmd_opt set_opts[] = {
-	{ "searchpath", 1, do_set_searchpath, "set searchpath for binaries and libraries" },
-	{ "depth", 1, do_set_depth, "set backtrace depth" },
+	{ "searchpath", 1, do_set_searchpath, "set searchpath for binaries and libraries", NULL, NULL },
+	{ "depth", 1, do_set_depth, "set backtrace depth", NULL, NULL },
 	{ },
 };
 
 static struct cmd_opt show_opts[] = {
-	{ "info", 1, do_show_info, "show client settings" },
-	{ "searchpath", 1, do_show_searchpath, "show searchpath for binaries and libraries" },
+	{ "info", 1, do_show_info, "show client settings", NULL, NULL },
+	{ "searchpath", 1, do_show_searchpath, "show searchpath for binaries and libraries", NULL, NULL },
 	{ },
 };
 
 static struct cmd_opt scan_opts[] = {
-	{ "all", 1, (void *)SCAN_ALL, "scan all memory blocks" },
-	{ "leak", 1, (void *)SCAN_LEAK, "scan only leaked allocations" },
-	{ "new", 1, (void *)SCAN_NEW, "scan only allocations since last scan" },
+	{ "all", 1, (void *)SCAN_ALL, "scan all memory blocks", NULL, NULL },
+	{ "leak", 1, (void *)SCAN_LEAK, "scan only leaked allocations", NULL, NULL },
+	{ "new", 1, (void *)SCAN_NEW, "scan only allocations since last scan", NULL, NULL },
 	{ },
 };
 
@@ -144,21 +144,24 @@ static struct cmd_opt cmds[] = {
 		1,
 		do_proclist,
 		"list processes",
-		""
+		"",
+		NULL
 	},
 	{
 		quit_str,
 		1,
 		do_quit,
 		"exit the program",
-		""
+		"",
+		NULL
 	},
 	{
 		reset_str,
 		1,
 		do_reset,
 		"reset all current memory allocation",
-		"[<pid>]"
+		"[<pid>]",
+		NULL
 	},
 	{
 		scan_str,
@@ -189,21 +192,24 @@ static struct cmd_opt cmds[] = {
 		4,
 		do_start,
 		"start allocation tracing",
-		""
+		"",
+		NULL
 	},
 	{
 		status_str,
 		4,
 		do_status,
 		"show allocation status",
-		"[<pid>]"
+		"[<pid>]",
+		NULL
 	},
 	{
 		stop_str,
 		3,
 		do_stop,
 		"stop allocation tracing",
-		""
+		"",
+		NULL
 	},
 	{ },
 };
@@ -513,6 +519,8 @@ static int do_set_depth(struct cmd_opt *cmd, struct cmd_opt *opt, int argc, cons
 
 static int do_show_info(struct cmd_opt *cmd, struct cmd_opt *opt, int argc, const char *argv[])
 {
+	(void)argv;
+
 	if (argc > 2) {
 		fprintf(stderr, "%s: too many option argument for '%s'\n", cmd->name, opt->name);
 		return -1;
@@ -528,6 +536,8 @@ static int do_show_searchpath(struct cmd_opt *cmd, struct cmd_opt *opt, int argc
 {
 	struct opt_b_t *p = options.opt_b;
 
+	(void)argv;
+
 	if (argc > 3) {
 		fprintf(stderr, "%s: too many option argument for '%s'\n", cmd->name, opt->name);
 		return -1;
@@ -661,6 +671,9 @@ static int show_process(struct process *process)
 
 static int do_proclist(struct cmd_opt *cmd, int argc, const char *argv[])
 {
+	(void)cmd;
+	(void)argv;
+
 	if (argc > 1) {
 		fprintf(stderr, "%s: expect no arguments\n", proclist_str);
 		return -1;
@@ -674,6 +687,10 @@ static int do_proclist(struct cmd_opt *cmd, int argc, const char *argv[])
 
 static int do_quit(struct cmd_opt *cmd, int argc, const char *argv[])
 {
+	(void)cmd;
+	(void)argc;
+	(void)argv;
+
 	quit = 1;
 
 	return 0;
@@ -683,6 +700,9 @@ static int do_reset(struct cmd_opt *cmd, int argc, const char *argv[])
 {
 	struct process *process;
 
+	(void)cmd;
+	(void)argc;
+
 	process = get_process(argv[1]);
 	if (!process)
 		return -1;
@@ -702,6 +722,8 @@ static int do_scan(struct cmd_opt *cmd, int argc, const char *argv[])
 	unsigned int i;
 	int mode;
 
+	(void)argc;
+
 	if (!client_connected())
 		return -1;
 
@@ -798,6 +820,9 @@ static int do_start(struct cmd_opt *cmd, int argc, const char *argv[])
 {
 	struct process *process;
 
+	(void)cmd;
+	(void)argc;
+
 	if (!client_connected())
 		return -1;
 
@@ -819,6 +844,9 @@ static int do_status(struct cmd_opt *cmd, int argc, const char *argv[])
 {
 	struct process *process;
 
+	(void)cmd;
+	(void)argc;
+
 	process = get_process(argv[1]);
 	if (!process)
 		return -1;
@@ -832,6 +860,9 @@ static int do_stop(struct cmd_opt *cmd, int argc, const char *argv[])
 {
 	struct process *process;
 
+	(void)cmd;
+	(void)argc;
+
 	if (!client_connected())
 		return -1;
 
diff --git a/config.h.in b/config.h.in
index 1e5b393..6120ed6 100644
--- a/config.h.in
+++ b/config.h.in
@@ -1,6 +1,13 @@
 #define PACKAGE_VERSION "@MT_VERSION_STRING@"
 
+#ifndef HAVE_PROCESS_VM_READV
 #cmakedefine HAVE_PROCESS_VM_READV
-#cmakedefine HAVE_LIBSELINUX
+#endif
 
+#ifndef HAVE_LIBSELINUX
+#cmakedefine HAVE_LIBSELINUX
+#endif
+
+#ifdef DISABLE_CLIENT
 #cmakedefine DISABLE_CLIENT
+#endif
diff --git a/debug.c b/debug.c
index d89d05a..b8c34db 100644
--- a/debug.c
+++ b/debug.c
@@ -47,6 +47,9 @@ void _debug(int level, const char *file, const char *function, int line, const c
 	va_end(args);
 
 	fprintf(stderr, "DEBUG: %s():%s@%d - %s\n", function, file, line, buf);
+
+//	fflush(debug_file);
+	free(buf);
 }
 
 #endif
diff --git a/debug.h b/debug.h
index b243e53..6b95556 100644
--- a/debug.h
+++ b/debug.h
@@ -26,6 +26,8 @@
 
 #include "config.h"
 
+#define DEBUG
+
 /* debug levels:
  */
 enum {
diff --git a/dwarf.c b/dwarf.c
index 8cdd776..91401e1 100644
--- a/dwarf.c
+++ b/dwarf.c
@@ -1166,7 +1166,7 @@ static int run_cfi_program(struct dwarf_addr_space *as, struct dwarf_reg_state *
 			set_reg(rs_current, regnum, DWARF_WHERE_REG, n);
 			break;
 		case DW_CFA_remember_state:
-			rs_tmp = malloc(regs_size(as));
+			rs_tmp = malloc(sizeof(struct dwarf_reg_stack) + regs_size(as));
 			memcpy(&rs_tmp->store, rs_current, regs_size(as));
 			rs_tmp->next = rs_stack;
 			rs_stack = rs_tmp;
diff --git a/event.c b/event.c
index 3bd9131..abc8c12 100644
--- a/event.c
+++ b/event.c
@@ -53,10 +53,16 @@
 #include "timer.h"
 #include "trace.h"
 
+#define	RET_DELETED	1
+#define	RET_DEFERED	2
+
 static LIST_HEAD(event_head);
 
 void queue_event(struct task *task)
 {
+	assert(task->event.type != EVENT_NONE);
+	assert(task->stopped);
+
 	if (task) {
 		if (task->event.type != EVENT_NONE)
 			list_add_tail(&task->event.list, &event_head);
@@ -88,52 +94,47 @@ void init_event(struct task *task)
 	INIT_LIST_HEAD(&task->event.list);
 }
 
-static void show_clone(struct task *task, enum event_type type)
+static const char * get_clone_type(enum event_type type)
 {
-	const char *str;
-
 	switch(type) {
 	case EVENT_FORK:
-		str = "fork";
-		break;
+		return "fork";
 	case EVENT_VFORK:
-		str = "vfork";
-		break;
+		return "vfork";
 	case EVENT_CLONE:
-		str = "clone";
-		break;
+		return "clone";
 	default:
-		str = "?";
 		break;
 	}
-	fprintf(stderr, "+++ process pid=%d %s (newpid=%d) +++\n", task->pid, str, task->event.e_un.newpid);
+	return "?";
 }
 
-static void handle_clone(struct task *task, enum event_type type)
+static int do_clone(struct task *task, struct task *newtask)
 {
-	struct task *newtask;
-	int newpid = task->event.e_un.newpid;
-
-	debug(DEBUG_FUNCTION, "pid=%d, newpid=%d", task->pid, newpid);
+	debug(DEBUG_EVENT, "+++ process %s pid=%d, newpid=%d", get_clone_type(task->event.type), task->pid, newtask->pid);
 
 	if (unlikely(options.verbose))
-		show_clone(task, type);
+		fprintf(stderr, "+++ process %s(%d) pid=%d, newpid=%d\n", get_clone_type(task->event.type), task->event.type, task->pid, newtask->pid);
 
-	continue_task(task, 0);
+	assert(task->stopped);
+	assert(newtask->stopped);
+	assert(newtask->is_new);
 
-	newtask = pid2task(newpid);
-	if (!newtask)
-		goto fail;
+	if (newtask->event.type != EVENT_NEW)
+		fprintf(stderr, "!!!task new unexpected event for pid=%d: %d\n", newtask->pid, newtask->event.type);
+	else
+	if (newtask->event.e_un.signum)
+		fprintf(stderr, "!!!task new unexpected signal for pid=%d: %d\n", newtask->pid, newtask->event.e_un.signum);
 
 	if (newtask->leader == newtask) {
+		if (!options.follow) {
+			untrace_proc(newtask);
+			return RET_DELETED;
+		}
+
 		if (task_fork(task, newtask) < 0)
 			goto fail;
 
-		if (!options.follow) {
-			remove_proc(newtask);
-			return;
-		}
-
 		report_fork(newtask, task);
 	}
 	else {
@@ -141,77 +142,130 @@ static void handle_clone(struct task *task, enum event_type type)
 			goto fail;
 	}
 
-	continue_task(newtask, newtask->event.e_un.signum);
-
-	return;
+	newtask->is_new = 0;
+	return continue_task(newtask, 0);
 fail:
-	fprintf(stderr,
-		"Error during init of tracing process %d\n"
-		"This process won't be traced.\n",
-		newpid
-	);
+	fprintf(stderr, "Error during clone of pid=%d - This process won't be traced!\n", newtask->pid);
+	return -1;
 }
 
-static void handle_signal(struct task *task)
+static int do_clone_cb(struct task *newtask, void *data)
 {
-	if (unlikely(options.verbose > 1)) {
-		if (task->event.e_un.signum && (task->event.e_un.signum != SIGSTOP || !task->was_stopped))
-			fprintf(stderr, "+++ process pid=%d signal %d: %s +++\n", task->pid, task->event.e_un.signum, strsignal(task->event.e_un.signum));
+	int ret;
+	struct task *task = data;
+
+	debug(DEBUG_EVENT, "+++ process do clone cb pid=%d, newpid=%d", task->pid, newtask->pid);
+
+	ret = do_clone(task, newtask);
+	continue_task(task, 0);
+	return ret;
+}
+
+static int handle_child(struct task *task)
+{
+	struct task *newtask;
+	int newpid = task->event.e_un.newpid;
+
+	debug(DEBUG_EVENT, "+++ process child pid=%d, newpid=%d", task->pid, newpid);
+
+	newtask = pid2task(newpid);
+
+	assert(newtask != NULL);
+
+	if (!newtask->stopped) {
+		debug(DEBUG_EVENT, "+++ process defer child pid=%d, newpid=%d", task->pid, newpid);
+		newtask->defer_func = do_clone_cb;
+		newtask->defer_data = task;
+		return RET_DEFERED;
 	}
 
-	continue_task(task, task->event.e_un.signum);
+	do_clone(task, newtask);
+	return continue_task(task, 0);
+}
+
+static int handle_signal(struct task *task)
+{
+	debug(DEBUG_EVENT, "+++ process signal pid=%d, event signal %d", task->pid, task->event.e_un.signum);
+
+	if (unlikely(options.verbose > 1)) {
+		if (task->event.e_un.signum)
+			fprintf(stderr, "+++ process pid=%d signal %d: %s\n", task->pid, task->event.e_un.signum, strsignal(task->event.e_un.signum));
+	}
+
+	return continue_task(task, task->event.e_un.signum);
 }
 
 static void show_exit(struct task *task)
 {
 	if (unlikely(options.verbose))
-		fprintf(stderr, "+++ process pid=%d exited (status=%d) +++\n", task->pid, task->event.e_un.ret_val);
+		fprintf(stderr, "+++ process pid=%d exited (status=%d)\n", task->pid, task->event.e_un.ret_val);
 }
 
-static void handle_about_exit(struct task *task)
+static int handle_new(struct task *task)
 {
+	debug(DEBUG_EVENT, "+++ process new pid=%d, event signal %d", task->pid, task->event.e_un.signum);
+
+	assert(task->is_new);
+
+	if (task->event.e_un.signum)
+		fprintf(stderr, "!!!task unexpected signal for pid=%d: %d\n", task->pid, task->event.e_un.signum);
+	task->is_new = 0;
+
+	return continue_task(task, task->event.e_un.signum);
+}
+
+static int handle_about_exit(struct task *task)
+{
+	debug(DEBUG_EVENT, "+++ process pid=%d about exit", task->pid);
+
 	if (task->leader == task) {
 		if (!options.logfile && report_about_exit(task) != -1) {
 			task->about_exit = 1;
-			return;
+			return 0;
 		}
 	}
-	continue_task(task, 0);
+	return continue_task(task, 0);
 }
 
-static void handle_exit(struct task *task)
+static int handle_exit(struct task *task)
 {
+	debug(DEBUG_EVENT, "+++ process pid=%d exited (status=%d)", task->pid, task->event.e_un.ret_val);
+
 	show_exit(task);
 
 	if (task->leader == task) {
 		report_exit(task);
-		remove_proc(task);
+		untrace_proc(task);
 	}
 	else {
 		remove_task(task);
 	}
+	return RET_DELETED;
 }
 
-static void handle_exit_signal(struct task *task)
+static int handle_exit_signal(struct task *task)
 {
+	debug(DEBUG_EVENT, "+++ process pid=%d killed by signal %s (%d)", task->pid, strsignal(task->event.e_un.signum), task->event.e_un.signum);
+
 	if (unlikely(options.verbose))
-		fprintf(stderr, "+++ process pid=%d killed by signal %s (%d) +++\n", task->pid, strsignal(task->event.e_un.signum), task->event.e_un.signum);
+		fprintf(stderr, "+++ process pid=%d killed by signal %s (%d)\n", task->pid, strsignal(task->event.e_un.signum), task->event.e_un.signum);
 
 	if (task->leader == task) {
 		report_exit(task);
-		remove_proc(task);
+		untrace_proc(task);
 	}
 	else {
 		remove_task(task);
 	}
+	return RET_DELETED;
 }
 
-static void handle_exec(struct task *task)
+static int handle_exec(struct task *task)
 {
-	debug(DEBUG_FUNCTION, "pid=%d", task->pid);
+	debug(DEBUG_EVENT, "+++ process pid=%d exec", task->pid);
 
 	if (unlikely(options.verbose))
-		fprintf(stderr, "+++ process pid=%d exec (%s) +++\n", task->pid, library_execname(task));
+		fprintf(stderr, "+++ process pid=%d exec\n", task->pid);
 
 	if (!options.follow_exec)
 		goto nofollow;
@@ -221,18 +275,20 @@ static void handle_exec(struct task *task)
 		goto untrace;
 	}
 
-	continue_task(task, 0);
-	return;
+	return continue_task(task, 0);
 nofollow:
 	report_nofollow(task);
 untrace:
-	remove_proc(task);
+	untrace_proc(task);
+	return RET_DELETED;
 }
 
 static int handle_call_after(struct task *task, struct breakpoint *bp)
 {
 	struct timespec start;
 
+	(void)bp;
+
 	if (!task->breakpoint)
 		return 0;
 
@@ -255,16 +311,41 @@ static int handle_call_after(struct task *task, struct breakpoint *bp)
 	return 0;
 }
 
-static void handle_breakpoint(struct task *task)
+static int handle_breakpoint(struct task *task)
 {
 	struct breakpoint *bp = task->event.e_un.breakpoint;
 	unsigned int hw = bp->hw;
 
-	debug(DEBUG_FUNCTION, "pid=%d, addr=%#lx", task->pid, bp->addr);
+	debug(DEBUG_EVENT, "+++ process pid=%d breakpoint  addr=%#lx", task->pid, bp->addr);
+
+	assert(task->stopped);
 
 	if (unlikely(options.verbose > 1))
 		set_timer(&task->halt_time, hw ? &hw_bp_time : &sw_bp_time);
 
+	if (unlikely(options.verbose))
+		++bp->count;
+
+	if (unlikely(task->skip_bp)) {
+		struct breakpoint *skip_bp = task->skip_bp;
+
+		task->skip_bp = NULL;
+
+		breakpoint_put(skip_bp);
+
+		if (likely(skip_bp == bp)) {
+			skip_breakpoint(task, bp);
+			goto end;
+		}
+
+		fprintf(stderr, "%s:%d\n", __func__, __LINE__);
+	}
+
+	if (unlikely(bp->deleted)) {
+		continue_task(task, 0);
+		goto end;
+	}
+
 #if HW_BREAKPOINTS > 1
 	if (bp->type >= BP_HW) {
 		if (unlikely(++bp->hwcnt >= (BP_REORDER_THRESHOLD << hw))) {
@@ -281,28 +362,7 @@ static void handle_breakpoint(struct task *task)
 	}
 #endif
 
-	if (unlikely(options.verbose))
-		++bp->count;
-
-	if (unlikely(task->skip_bp)) {
-		struct breakpoint *skip_bp = task->skip_bp;
-
-		task->skip_bp = NULL;
-
-		breakpoint_put(skip_bp);
-
-		if (likely(skip_bp == bp)) {
-			skip_breakpoint(task, bp);
-			goto end;
-		}
-	}
-
-	if (unlikely(bp->deleted)) {
-		continue_task(task, 0);
-		goto end;
-	}
-
-	if (unlikely(breakpoint_on_hit(task, bp))) {
+	if (bp->on_hit && bp->on_hit(task, bp)) {
 		continue_task(task, 0);
 		goto end;
 	}
@@ -336,65 +396,79 @@ static void handle_breakpoint(struct task *task)
 		}
 	}
 
-	if (task->stopped)
-		skip_breakpoint(task, bp);
-
+	skip_breakpoint(task, bp);
 end:
 	breakpoint_put(bp);
+	return 0;
 }
 
-int handle_event(void)
+int handle_event(struct task *task)
 {
-	struct task *task = next_event();
+	int ret;
 
 	if (!task)
 		return 0;
 
+	debug(DEBUG_EVENT, "+++ process pid=%d event: %d", task->pid, task->event.type);
+
+	assert(task->stopped);
+
+	if (task->defer_func) {
+		ret = task->defer_func(task, task->defer_data);
+
+		task->defer_func = NULL;
+		task->defer_data = NULL;
+
+		return ret;
+	}
+
 	struct event *event = &task->event;
 	enum event_type type = event->type;
 
-	event->type = EVENT_NONE;
-	debug(DEBUG_FUNCTION, "pid=%d, type=%d", task->pid, event->type);
-
 	switch (type) {
 	case EVENT_NONE:
-		debug(DEBUG_EVENT_HANDLER, "pid=%d, event none", task->pid);
+		ret = continue_task(task, task->event.e_un.signum);
 		break;
 	case EVENT_SIGNAL:
-		debug(DEBUG_EVENT_HANDLER, "pid=%d, event signal %d", task->pid, event->e_un.signum);
-		handle_signal(task);
+		ret = handle_signal(task);
 		break;
 	case EVENT_ABOUT_EXIT:
-		debug(DEBUG_EVENT_HANDLER, "pid=%d, event exit %d", task->pid, event->e_un.ret_val);
-		handle_about_exit(task);
-		break;
+		ret = handle_about_exit(task);
+		goto out;
 	case EVENT_EXIT:
-		debug(DEBUG_EVENT_HANDLER, "pid=%d, event exit %d", task->pid, event->e_un.ret_val);
-		handle_exit(task);
+		ret = handle_exit(task);
 		break;
 	case EVENT_EXIT_SIGNAL:
-		debug(DEBUG_EVENT_HANDLER, "pid=%d, event exit signal %d", task->pid, event->e_un.signum);
-		handle_exit_signal(task);
+		ret = handle_exit_signal(task);
 		break;
 	case EVENT_FORK:
 	case EVENT_VFORK:
 	case EVENT_CLONE:
-		debug(DEBUG_EVENT_HANDLER, "pid=%d, event clone (%u)", task->pid, event->e_un.newpid);
-		handle_clone(task, type);
+		ret = handle_child(task);
 		break;
 	case EVENT_EXEC:
-		debug(DEBUG_EVENT_HANDLER, "pid=%d, event exec()", task->pid);
-		handle_exec(task);
+		ret = handle_exec(task);
 		break;
 	case EVENT_BREAKPOINT:
-		debug(DEBUG_EVENT_HANDLER, "pid=%d, event breakpoint %#lx", task->pid, event->e_un.breakpoint->addr);
-		handle_breakpoint(task);
+		ret = handle_breakpoint(task);
+		goto out;
+	case EVENT_NEW:
+		ret = handle_new(task);
 		break;
 	default:
-		fprintf(stderr, "Error! unknown event?\n");
-		return -1;
+		fprintf(stderr, "!!!pid=%d, unknown event %d\n", task->pid, type);
+		abort();
 	}
 
-	return 1;
+	if (ret == RET_DELETED)
+		return 1;
+
+	if (ret != RET_DEFERED) {
+		assert(task->event.type == EVENT_NONE);
+		assert(task->stopped == 0);
+	}
+	assert(task->is_new == 0);
+out:
+	return (ret < 0) ? ret : 0;
 }
 
diff --git a/event.h b/event.h
index f7b2ef4..51230ab 100644
--- a/event.h
+++ b/event.h
@@ -36,7 +36,8 @@ enum event_type {
 	EVENT_CLONE,
 	EVENT_VFORK,
 	EVENT_EXEC,
-	EVENT_BREAKPOINT
+	EVENT_BREAKPOINT,
+	EVENT_NEW,
 };
 
 struct event {
@@ -54,7 +55,7 @@ void init_event(struct task *task);
 void remove_event(struct task *task);
 struct task *next_event(void);
 void queue_event(struct task *task);
-int handle_event(void);
+int handle_event(struct task *task);
 
 #endif
 
diff --git a/library.c b/library.c
index 720651a..9b30cb8 100644
--- a/library.c
+++ b/library.c
@@ -171,10 +171,10 @@ void library_delete_list(struct task *leader, struct list_head *list)
 		struct library *lib = container_of(it, struct library, list);
 		struct libref *libref = lib->libref;
 
-		debug(DEBUG_FUNCTION, "%s@%#lx", libref->filename, libref->base);
+		debug(DEBUG_FUNCTION, "%s@%#lx pid=%d ", libref->filename, libref->base, leader->pid);
 
 		if (unlikely(options.verbose > 1))
-			fprintf(stderr, "+++ library del pid=%d %s@%#lx %#lx-%#lx +++\n", leader->pid, libref->filename, libref->base, libref->load_addr, libref->load_addr + libref->load_size);
+			fprintf(stderr, "+++ library del pid=%d %s@%#lx %#lx-%#lx\n", leader->pid, libref->filename, libref->base, libref->load_addr, libref->load_addr + libref->load_size);
 
 		library_delete(leader, lib);
 	}
@@ -274,7 +274,7 @@ static struct library *_library_add(struct task *leader, struct libref *libref)
 	insert_lib(leader, lib);
 
 	if (unlikely(options.verbose > 1))
-		fprintf(stderr, "+++ library add pid=%d %s@%#lx %#lx-%#lx +++\n", leader->pid, libref->filename, libref->base, libref->load_addr, libref->load_addr + libref->load_size);
+		fprintf(stderr, "+++ library add pid=%d %s@%#lx %#lx-%#lx\n", leader->pid, libref->filename, libref->base, libref->load_addr, libref->load_addr + libref->load_size);
 
 	return lib;
 }
diff --git a/list.h b/list.h
index ff64ef1..fd55551 100644
--- a/list.h
+++ b/list.h
@@ -55,41 +55,39 @@ static inline void INIT_LIST_HEAD(struct list_head *list)
  * This is only for internal list manipulation where we know
  * the prev/next entries already!
  */
-static inline void __list_add(struct list_head *new,
-			      struct list_head *prev,
-			      struct list_head *next)
+static inline void __list_add(struct list_head *elem, struct list_head *prev, struct list_head *next)
 {
-	next->prev = new;
-	new->next = next;
-	new->prev = prev;
-	prev->next = new;
+	next->prev = elem;
+	elem->next = next;
+	elem->prev = prev;
+	prev->next = elem;
 }
 
 /**
  * list_add - add a new entry
- * @new: new entry to be added
+ * @elem: new entry to be added
  * @head: list head to add it after
  *
  * Insert a new entry after the specified head.
  * This is good for implementing stacks.
  */
-static inline void list_add(struct list_head *new, struct list_head *head)
+static inline void list_add(struct list_head *elem, struct list_head *head)
 {
-	__list_add(new, head, head->next);
+	__list_add(elem, head, head->next);
 }
 
 
 /**
  * list_add_tail - add a new entry
- * @new: new entry to be added
+ * @elem: new entry to be added
  * @head: list head to add it before
  *
  * Insert a new entry before the specified head.
  * This is useful for implementing queues.
  */
-static inline void list_add_tail(struct list_head *new, struct list_head *head)
+static inline void list_add_tail(struct list_head *elem, struct list_head *head)
 {
-	__list_add(new, head->prev, head);
+	__list_add(elem, head->prev, head);
 }
 
 /*
diff --git a/main.c b/main.c
index 0543b8e..8d82abe 100644
--- a/main.c
+++ b/main.c
@@ -61,6 +61,8 @@ struct mt_timer report_in_time;
 struct mt_timer report_out_time;
 struct mt_timer skip_bp_time;
 
+pid_t mtrace_pid;
+
 static int do_exit;
 
 void mtrace_request_exit(void)
@@ -69,7 +71,7 @@ void mtrace_request_exit(void)
 		return;
 
 	if (unlikely(options.verbose))
-		fprintf(stderr, "+++ request exit +++\n");
+		fprintf(stderr, "+++ request exit\n");
 
 	do_exit = 1;
 	wait_event_wakeup();
@@ -110,6 +112,8 @@ static void mtrace_init(char **cmd_args)
 {
 	struct opt_p_t *opt_p_tmp;
 
+	mtrace_pid = getpid();
+
 	if (options.command) {
 		struct task *task = task_create(cmd_args);
 
@@ -117,7 +121,7 @@ static void mtrace_init(char **cmd_args)
 			exit(EXIT_FAILURE);
 
 		if (unlikely(options.verbose))
-			fprintf(stderr, "+++ process pid=%d created (%s) +++\n", task->pid, library_execname(task));
+			fprintf(stderr, "+++ process pid=%d created (%s)\n", task->pid, library_execname(task));
 	}
 
 	for(opt_p_tmp = options.opt_p; opt_p_tmp; opt_p_tmp = opt_p_tmp->next)
@@ -126,12 +130,17 @@ static void mtrace_init(char **cmd_args)
 
 static void mtrace_main(void)
 {
+	struct task *task;
+
 	while(!do_exit)  {
 		if (task_list_empty())
 			break;
 
-		if (handle_event() == -1)
-			break;
+		task = next_event();
+		if (task) {
+			if (handle_event(task) == -1)
+				break;
+		}
 
 		if (server_poll() == -1)
 			break;
@@ -142,8 +151,6 @@ int main(int argc, char *argv[])
 {
 	char **cmd_args = process_options(argc, argv);
 
-	init_pid_hash();
-
 	if (options.trace) {
 		if (options.logfile) {
 			if (server_logfile() == -1)
diff --git a/main.h b/main.h
index b3f3aaf..1e04304 100644
--- a/main.h
+++ b/main.h
@@ -23,6 +23,8 @@
 #ifndef _INC_MAIN_H
 #define _INC_MAIN_H
 
+#include <sys/types.h>
+
 #include "timer.h"
 
 void mtrace_request_exit(void);
@@ -36,5 +38,6 @@ struct mt_timer report_in_time;
 struct mt_timer report_out_time;
 struct mt_timer skip_bp_time;
 
+pid_t mtrace_pid;
 #endif
 
diff --git a/mtelf.c b/mtelf.c
index 5af6f7e..190d9dd 100644
--- a/mtelf.c
+++ b/mtelf.c
@@ -259,7 +259,7 @@ static int elf_lib_init(struct mt_elf *mte, struct task *task, struct libref *li
 	}
 #endif
 
-	if (mte->eh_hdr.p_filesz && mte->dyn_addr) {
+	if (mte->eh_hdr.p_filesz && mte->dyn) {
 		if (dwarf_get_unwind_table(task, libref, (struct dwarf_eh_frame_hdr *)(libref->image_addr - libref->load_offset + mte->eh_hdr.p_offset)) < 0)
 			return -1;
 	}
@@ -474,7 +474,7 @@ static int entry_breakpoint_on_hit(struct task *task, struct breakpoint *a)
 	return 1;
 }
 
-struct libref *elf_read_main_binary(struct task *task)
+struct libref *elf_read_main_binary(struct task *task, int was_attached)
 {
 	char fname[PATH_MAX];
 	int ret;
@@ -520,7 +520,7 @@ struct libref *elf_read_main_binary(struct task *task)
 
 	close_elf(&mte);
 
-	report_attach(task);
+	report_attach(task, was_attached);
 
 	library_add(task, libref);
 
@@ -529,7 +529,10 @@ struct libref *elf_read_main_binary(struct task *task)
 
 	struct mt_elf mte_ld = { };
 
-	copy_str_from_proc(task, ARCH_ADDR_T(mte.interp), fname, sizeof(fname));
+	if (copy_str_from_proc(task, ARCH_ADDR_T(mte.bias + mte.interp), fname, sizeof(fname)) == -1) {
+		fprintf(stderr, "fatal error: cannot get loader name for pid=%d\n", task->pid);
+		abort();
+	}
 
 	if (!elf_read(&mte_ld, task, fname, (GElf_Addr)base)) {
 		struct libref *libref;
@@ -547,7 +550,7 @@ struct libref *elf_read_main_binary(struct task *task)
 		if (!ret) {
 			library_add(task, libref);
 
-			if (linkmap_init(task, ARCH_ADDR_T(mte.dyn_addr))) {
+			if (linkmap_init(task, ARCH_ADDR_T(mte.bias + mte.dyn))) {
 				arch_addr_t addr = find_solib_break(&mte_ld);
 				if (!addr)
 					addr = ARCH_ADDR_T(entry);
@@ -562,7 +565,7 @@ struct libref *elf_read_main_binary(struct task *task)
 				else {
 					entry_bp->breakpoint.on_hit = entry_breakpoint_on_hit;
 					entry_bp->breakpoint.locked = 1;
-					entry_bp->dyn_addr = ARCH_ADDR_T(mte.dyn_addr);
+					entry_bp->dyn_addr = ARCH_ADDR_T(mte.bias + mte.dyn);
 
 					breakpoint_enable(task, &entry_bp->breakpoint);
 				}
diff --git a/mtelf.h b/mtelf.h
index 4030f44..7d34a8f 100644
--- a/mtelf.h
+++ b/mtelf.h
@@ -53,15 +53,10 @@ struct mt_elf {
 	GElf_Addr pltgot;
 };
 
-struct elf_image {
-	void *addr;	/* pointer to mmap'd image */
-	size_t size;	/* (file-) size of the image */
-};
-
 int elf_read_library(struct task *task, struct libref *libref, const char *filename, GElf_Addr bias);
 
 /* Create a library object representing the main binary. */
-struct libref *elf_read_main_binary(struct task *task);
+struct libref *elf_read_main_binary(struct task *task, int was_attached);
 
 #endif
 
diff --git a/options.c b/options.c
index b23b95c..97cde33 100644
--- a/options.c
+++ b/options.c
@@ -39,6 +39,7 @@
 #include <pwd.h>
 
 #include "common.h"
+#include "debug.h"
 #include "options.h"
 
 #ifndef SYSCONFDIR
@@ -171,11 +172,11 @@ static char *search_for_command(char *filename)
 
 static int add_opt_F(char *filename)
 {
-	struct opt_F_t *tmp = malloc(sizeof(*tmp));
-
 	if (access(filename, R_OK))
 		return -1;
 
+	struct opt_F_t *tmp = malloc(sizeof(*tmp));
+
 	if (!tmp) {
 		fprintf(stderr, "%s\n", strerror(errno));
 		exit(1);
diff --git a/report.c b/report.c
index ea54a5f..2afcdc0 100644
--- a/report.c
+++ b/report.c
@@ -73,8 +73,6 @@ static void report_alloc64(struct task *task, enum mt_operation op, unsigned lon
 		}
 	}
 
-	skip_breakpoint(task, task->event.e_un.breakpoint);
-
 	server_send_msg(op, task->leader->pid, alloc, sizeof(*alloc) + i * sizeof(uint64_t));
 }
 
@@ -108,8 +106,6 @@ static void report_alloc32(struct task *task, enum mt_operation op, unsigned lon
 		}
 	}
 
-	skip_breakpoint(task, task->event.e_un.breakpoint);
-
 	server_send_msg(op, task->leader->pid, alloc, sizeof(*alloc) + i * sizeof(uint32_t));
 }
 
@@ -139,6 +135,23 @@ static void _report_malloc(struct task *task, struct library_symbol *libsym)
 	_report_alloc_op(task, libsym, MT_MALLOC);
 }
 
+#if 0
+static void _report_malloc1(struct task *task, struct library_symbol *libsym)
+{
+	unsigned long ret = fetch_retval(task);
+
+	report_alloc(task, MT_MALLOC, ret, 1, options.bt_depth, libsym);
+}
+#endif
+
+static void _report_reallocarray(struct task *task, struct library_symbol *libsym)
+{
+	unsigned long size = fetch_param(task, 1) * fetch_param(task, 2);
+	unsigned long ret = fetch_retval(task);
+
+	report_alloc(task, MT_MALLOC, ret, size, options.bt_depth, libsym);
+}
+
 static void _report_new(struct task *task, struct library_symbol *libsym)
 {
 	_report_alloc_op(task, libsym, options.sanity ? MT_NEW : MT_MALLOC);
@@ -173,31 +186,34 @@ static void report_delete_array(struct task *task, struct library_symbol *libsym
 
 static void _report_realloc(struct task *task, struct library_symbol *libsym)
 {
+	unsigned long addr = fetch_param(task, 0);
+	unsigned long size = fetch_param(task, 1);
 	unsigned long ret = fetch_retval(task);
 
-	if (ret) {
-		unsigned long size = fetch_param(task, 1);
-
-		report_alloc(task, MT_REALLOC, ret, size, options.bt_depth, libsym);
+	if (!addr) {
+		if (ret)
+			report_alloc(task, MT_REALLOC, ret, size, options.bt_depth, libsym);
+		return;
 	}
 
-	if (fetch_param(task, 0)) {
-		if (task_is_64bit(task)) {
-			struct mt_alloc_payload_64 *alloc = alloca(sizeof(*alloc));
+	if (ret)
+		report_alloc(task, MT_REALLOC, ret, size, options.bt_depth, libsym);
 
-			alloc->ptr = (uint64_t)ret;
-			alloc->size = (uint64_t)task->pid;
+	if (task_is_64bit(task)) {
+		struct mt_alloc_payload_64 *alloc = alloca(sizeof(*alloc));
 
-			server_send_msg(MT_REALLOC_DONE, task->leader->pid, alloc, sizeof(*alloc));
-		}
-		else {
-			struct mt_alloc_payload_32 *alloc = alloca(sizeof(*alloc));
+		alloc->ptr = (uint64_t)ret;
+		alloc->size = (uint64_t)task->pid;
 
-			alloc->ptr = (uint32_t)ret;
-			alloc->size = (uint32_t)task->pid;
+		server_send_msg(MT_REALLOC_DONE, task->leader->pid, alloc, sizeof(*alloc));
+	}
+	else {
+		struct mt_alloc_payload_32 *alloc = alloca(sizeof(*alloc));
 
-			server_send_msg(MT_REALLOC_DONE, task->leader->pid, alloc, sizeof(*alloc));
-		}
+		alloc->ptr = (uint32_t)ret;
+		alloc->size = (uint32_t)task->pid;
+
+		server_send_msg(MT_REALLOC_DONE, task->leader->pid, alloc, sizeof(*alloc));
 	}
 }
 
@@ -205,7 +221,8 @@ static void report_realloc(struct task *task, struct library_symbol *libsym)
 {
 	unsigned long addr = fetch_param(task, 0);
 
-	report_alloc(task, MT_REALLOC_ENTER, addr, task->pid, options.sanity ? options.bt_depth : 0, libsym);
+	if (addr)
+		report_alloc(task, MT_REALLOC_ENTER, addr, task->pid, options.sanity ? options.bt_depth : 0, libsym);
 }
 
 static void _report_calloc(struct task *task, struct library_symbol *libsym)
@@ -350,6 +367,16 @@ static const struct function flist[] = {
 	{ "pvalloc",						"pvalloc",		1,	NULL,			_report_pvalloc },
 	{ "mremap",						"mremap",		0,	report_mremap,		_report_mremap },
 	{ "cfree",						"cfree",		1,	report_free,		NULL },
+	{ "reallocarray",					"reallocarray",		0,	NULL,			_report_reallocarray },
+#if 0
+	{ "strdup",						"strdup",		0,	NULL,			_report_malloc1 },
+	{ "strndup",						"strndup",		0,	NULL,			_report_malloc1 },
+	{ "__strdup",						"__strdup",		0,	NULL,			_report_malloc1 },
+	{ "__strndup",						"__strndup",		0,	NULL,			_report_malloc1 },
+	{ "asprintf",						"asprintf",		0,	NULL,			_report_malloc1 },
+	{ "vasprintf",						"vasprintf",		0,	NULL,			_report_malloc1 },
+	{ "__asprintf",						"__asprintf",		0,	NULL,			_report_malloc1 },
+#endif
 
 	{ "new(unsigned int)",					"_Znwj",		1,	NULL,			_report_new },
 	{ "new[](unsigned int)",				"_Znaj",		1,	NULL,			_report_new_array },
@@ -361,14 +388,32 @@ static const struct function flist[] = {
 	{ "new(unsigned long, std::nothrow_t const&)",		"_ZnwmRKSt9nothrow_t",	1,	NULL,			_report_new },
 	{ "new[](unsigned long, std::nothrow_t const&)",	"_ZnamRKSt9nothrow_t",	1,	NULL,			_report_new_array },
 
+	{ "new(unsigned int, std::align_val_t, std::nothrow_t const&)",		"_ZnwjSt11align_val_tRKSt9nothrow_t",	1,	NULL,	_report_new },
+	{ "new[](unsigned int, std::align_val_t, std::nothrow_t const&)",	"_ZnajSt11align_val_tRKSt9nothrow_t",	1,	NULL,	_report_new_array },
+	{ "new(unsigned int, std::align_val_t)",				"_ZnwjSt11align_val_t",			1,	NULL,	_report_new },
+	{ "new[](unsigned int, std::align_val_t)",				"_ZnajSt11align_val_t",			1,	NULL,	_report_new_array },
+	{ "new(unsigned long, std::align_val_t, std::nothrow_t const&)",	"_ZnwmSt11align_val_tRKSt9nothrow_t",	1,	NULL,	_report_new },
+	{ "new[](unsigned long, std::align_val_t, std::nothrow_t const&)",	"_ZnamSt11align_val_tRKSt9nothrow_t",	1,	NULL,	_report_new_array },
+	{ "new(unsigned long, std::align_val_t)",				"_ZnwmSt11align_val_t",			1,	NULL,	_report_new },
+	{ "new[](unsigned long, std::align_val_t)",				"_ZnamSt11align_val_t",			1,	NULL,	_report_new_array },
+
 	{ "delete(void*)",					"_ZdlPv",		1,	report_delete,		NULL },
 	{ "delete[](void*)",					"_ZdaPv",		1,	report_delete_array,	NULL },
 	{ "delete(void*, std::nothrow_t const&)",		"_ZdlPvRKSt9nothrow_t",	1,	report_delete,		NULL },
 	{ "delete[](void*, std::nothrow_t const&)",		"_ZdaPvRKSt9nothrow_t",	1,	report_delete_array,	NULL },
+	{ "delete(void*, unsigned int)",			"_ZdlPvj",		1,	report_delete,		NULL },
+	{ "delete[](void*, unsigned int)",			"_ZdaPvj",		1,	report_delete_array,	NULL },
 	{ "delete(void*, unsigned long)",			"_ZdlPvm",		1,	report_delete,		NULL },
-	{ "delete[](void*, unsigned long)",			"_ZdaPvj",		1,	report_delete_array,	NULL },
-	{ "delete(void*, unsigned long)",			"_ZdlPvj",		1,	report_delete,		NULL },
 	{ "delete[](void*, unsigned long)",			"_ZdaPvm",		1,	report_delete_array,	NULL },
+
+	{ "delete(void*, std::align_val_t)",				"_ZdlPvSt11align_val_t",		1,	report_delete,		NULL },
+	{ "delete[](void*, std::align_val_t)",				"_ZdaPvSt11align_val_t",		1,	report_delete_array,	NULL },
+	{ "delete(void*, std::align_val_t, std::nothrow_t const&)",	"_ZdlPvSt11align_val_tRKSt9nothrow_t",	1,	report_delete,		NULL },
+	{ "delete[](void*, std::align_val_t, std::nothrow_t const&)",	"_ZdaPvSt11align_val_tRKSt9nothrow_t",	1,	report_delete_array,	NULL },
+	{ "delete(void*, unsigned int, std::align_val_t)",		"_ZdlPvjSt11align_val_t",		1,	report_delete,		NULL },
+	{ "delete[](void*, unsigned int, std::align_val_t)",		"_ZdaPvjSt11align_val_t",		1,	report_delete_array,	NULL },
+	{ "delete(void*, unsigned long, std::align_val_t)",		"_ZdlPvmSt11align_val_t",		1,	report_delete,		NULL },
+	{ "delete[](void*, unsigned long, std::align_val_t)",		"_ZdaPvmSt11align_val_t",		1,	report_delete_array,	NULL },
 };
 
 const struct function *flist_matches_symbol(const char *sym_name)
@@ -466,9 +511,9 @@ int report_scan(pid_t pid, const void *data, unsigned int data_len)
 	return server_send_msg(MT_SCAN, pid, data, data_len);
 }
 
-int report_attach(struct task *task)
+int report_attach(struct task *task, int was_attached)
 {
-	struct mt_attached_payload state = { .attached = task->attached };
+	struct mt_attached_payload state = { .attached = !!was_attached };
 
 	if (!server_connected())
 		return -1;
@@ -530,7 +575,7 @@ static void report_process(struct task *leader)
 {
 	struct list_head *it;
 
-	report_attach(leader);
+	report_attach(leader, 1);
 
 	list_for_each(it, &leader->libraries_list) {
 		struct library *lib = container_of(it, struct library, list);
diff --git a/report.h b/report.h
index 03dd6f5..f4ee764 100644
--- a/report.h
+++ b/report.h
@@ -23,6 +23,8 @@
 #ifndef _INC_REPORT_H
 #define _INC_REPORT_H
 
+#include <unistd.h>
+
 #include "forward.h"
 
 struct function {
@@ -44,7 +46,7 @@ int report_add_map(struct task *task, struct library *lib);
 int report_del_map(struct task *task, struct library *lib);
 int report_info(int do_trace);
 int report_scan(pid_t pid, const void *data, unsigned int data_len);
-int report_attach(struct task *task);
+int report_attach(struct task *task, int was_attached);
 int report_fork(struct task *task, struct task *ptask);
 int report_exit(struct task *task);
 int report_about_exit(struct task *task);
diff --git a/server.c b/server.c
index 8e564e4..a1387f8 100644
--- a/server.c
+++ b/server.c
@@ -241,6 +241,8 @@ static void *server_listen_thread(void *ptr)
 {
 	int ret;
 
+	(void)ptr;
+
 	for(;;) {
 		if (!server_connected()) {
 			thread_enable_cancel();
@@ -307,6 +309,8 @@ static void *server_pair_thread(void *ptr)
 {
 	int ret;
 
+	(void)ptr;
+
 	for(;;) {
 		if (!server_connected())
 			break;
@@ -389,8 +393,11 @@ int server_stop(void)
 
 	if (listen_fd != -1) {
 		thread_cancel(thread);
-		if (thread)
+		if (thread) {
 			thread_join(thread);
+			free(thread);
+			thread = NULL;
+		}
 
 		if (is_named(options.address))
 			unlink(options.address);
diff --git a/sysdeps/linux-gnu/os.c b/sysdeps/linux-gnu/os.c
index c7d2e00..99d3184 100644
--- a/sysdeps/linux-gnu/os.c
+++ b/sysdeps/linux-gnu/os.c
@@ -63,6 +63,9 @@ struct map {
 
 static void report_fault(int signo, siginfo_t* siginf, void* arg)
 {
+	(void)siginf;
+	(void)arg;
+
 	fprintf(stderr, "fault signal %d (%s)\n", signo, strsignal(signo));
 
 #ifndef DISABLE_CLIENT
@@ -148,6 +151,8 @@ skip:
 
 static void signal_exit(int sig)
 {
+	(void)sig;
+
 	signal(SIGINT, SIG_IGN);
 	signal(SIGTERM, SIG_IGN);
 
@@ -366,28 +371,6 @@ void change_uid(void)
 	}
 }
 
-ssize_t sock_fd_write(int sock, void *buf, ssize_t buflen, int fd)
-{
-	ssize_t		size;
-	struct msghdr	msg;
-	struct iovec	iov;
-
-	iov.iov_base = buf;
-	iov.iov_len = buflen;
-
-	msg.msg_name = NULL;
-	msg.msg_namelen = 0;
-	msg.msg_iov = &iov;
-	msg.msg_iovlen = 1;
-
-	msg.msg_control = NULL;
-	msg.msg_controllen = 0;
-
-	size = sendmsg(sock, &msg, MSG_DONTWAIT);
-
-	return size;
-}
-
 int os_init(void)
 {
 	struct sigaction act;
diff --git a/sysdeps/linux-gnu/proc.c b/sysdeps/linux-gnu/proc.c
index 7b09bc7..10e9ea6 100644
--- a/sysdeps/linux-gnu/proc.c
+++ b/sysdeps/linux-gnu/proc.c
@@ -489,6 +489,8 @@ static int rdebug_bp_on_hit(struct task *task, struct breakpoint *bp)
 	struct lt_r_debug_64 rdbg;
 	struct task *leader = task->leader;
 
+	(void)bp;
+
 	debug(DEBUG_FUNCTION, "pid=%d", task->pid);
 
 	if (load_debug_struct(task, leader->os.debug_addr, &rdbg) < 0)
@@ -615,6 +617,7 @@ int os_task_init(struct task *task)
 
 void os_task_destroy(struct task *task)
 {
+	(void)task;
 }
 
 int os_task_clone(struct task *retp, struct task *task)
diff --git a/sysdeps/linux-gnu/trace.c b/sysdeps/linux-gnu/trace.c
index b8a4b71..ce97ddf 100644
--- a/sysdeps/linux-gnu/trace.c
+++ b/sysdeps/linux-gnu/trace.c
@@ -52,34 +52,46 @@
 static pthread_mutex_t mutex = PTHREAD_MUTEX_INITIALIZER;
 static volatile pid_t wakeup_pid = -1;
 
-static int inline task_kill(struct task *task, int sig)
+static inline int task_kill(struct task *task, int sig)
 {
 	errno = 0;
 
 	return syscall(__NR_tgkill, task->leader->pid, task->pid, sig);
 }
 
+static inline int wait_task(struct task *task, int *status)
+{
+	int ret;
+
+	ret = TEMP_FAILURE_RETRY(waitpid(task ? task->pid : -1, status, __WALL));
+	if (ret == -1) {
+		if (task)
+			fprintf(stderr, "!!!%s: waitpid pid=%d %s\n", __func__,  task->pid, strerror(errno));
+	}
+	return ret;
+}
+
 static int trace_setup(struct task *task, int status, int signum)
 {
-	int stop_signal;
+	int sig;
 
-	task->traced = 1;
+	task->attached = 1;
 	task->stopped = 1;
 	task->leader->threads_stopped++;
-	task->event.type = EVENT_SIGNAL;
+	task->event.type = EVENT_NEW;
 	task->event.e_un.signum = 0;
 
 	if (!WIFSTOPPED(status)) {
-		fprintf(stderr, "%s pid=%d not stopped\n", __FUNCTION__, task->pid);
+		fprintf(stderr, "!!!pid=%d not stopped\n", task->pid);
 		return -1;
 	}
 
-	stop_signal = WSTOPSIG(status);
+	sig = WSTOPSIG(status);
 
-	if (stop_signal != signum) {
-		task->event.e_un.signum = stop_signal;
+	if (sig != signum) {
+		task->event.e_un.signum = sig;
 
-		fprintf(stderr, "%s pid=%d unexpected trace signal (got:%d expected:%d)\n", __FUNCTION__, task->pid, stop_signal, signum);
+		fprintf(stderr, "!!!pid=%d unexpected trace signal (got:%d expected:%d)\n", task->pid, sig, signum);
 		return -1;
 	}
 
@@ -90,10 +102,8 @@ static int _trace_wait(struct task *task, int signum)
 {
 	int status;
 
-	if (unlikely(TEMP_FAILURE_RETRY(waitpid(task->pid, &status, __WALL)) != task->pid)) {
-		fprintf(stderr, "%s pid=%d %s\n", __FUNCTION__, task->pid, strerror(errno));
+	if (unlikely(wait_task(task, &status) == -1))
 		return -1;
-	}
 
 	if (WIFEXITED(status))
 		return -1;
@@ -103,11 +113,11 @@ static int _trace_wait(struct task *task, int signum)
 
 int trace_wait(struct task *task)
 {
+	assert(task->attached == 0);
+
 	if (_trace_wait(task, SIGTRAP))
 		return -1;
 
-	queue_event(task);
-
 	return 0;
 }
 
@@ -115,6 +125,8 @@ static int child_event(struct task *task, enum event_type ev)
 {
 	unsigned long data;
 
+	debug(DEBUG_EVENT, "child event %d pid=%d, newpid=%d", ev, task->pid, task->event.e_un.newpid);
+
 	if (unlikely(ptrace(PTRACE_GETEVENTMSG, task->pid, NULL, &data) == -1)) {
 		debug(DEBUG_EVENT, "PTRACE_GETEVENTMSG pid=%d %s", task->pid, strerror(errno));
 		return -1;
@@ -128,10 +140,7 @@ static int child_event(struct task *task, enum event_type ev)
 		if (unlikely(!child))
 			return -1;
 
-		if (_trace_wait(child, SIGSTOP)) {
-			remove_task(child);
-			return -1;
-		}
+		child->attached = 1;
 	}
 
 	task->event.e_un.newpid = pid;
@@ -142,44 +151,59 @@ static int child_event(struct task *task, enum event_type ev)
 
 static int _process_event(struct task *task, int status)
 {
-	int stop_signal;
+	int sig = WSTOPSIG(status);
+
+	task->stopped = 1;
+
+	assert(task->event.type == EVENT_NONE);
 
 	if (WIFSIGNALED(status)) {
+		debug(DEBUG_EVENT, "EXIT_SIGNAL: pid=%d, signum=%d", task->pid, task->event.e_un.signum);
+
 		task->event.type = EVENT_EXIT_SIGNAL;
 		task->event.e_un.signum = WTERMSIG(status);
-		debug(DEBUG_EVENT, "EXIT_SIGNAL: pid=%d, signum=%d", task->pid, task->event.e_un.signum);
 		return 0;
 	}
 
 	if (WIFEXITED(status)) {
+		debug(DEBUG_EVENT, "EXIT: pid=%d, status=%d", task->pid, task->event.e_un.ret_val);
+
 		task->event.type = EVENT_EXIT;
 		task->event.e_un.ret_val = WEXITSTATUS(status);
-		debug(DEBUG_EVENT, "EXIT: pid=%d, status=%d", task->pid, task->event.e_un.ret_val);
 		return 0;
 	}
 
 	if (!WIFSTOPPED(status)) {
-		/* should never happen */
-		debug(DEBUG_EVENT, "NONE: pid=%d ???", task->pid);
+		fprintf(stderr, "!!!not WIFSTOPPED pid=%d\n", task->pid);
 		return -1;
 	}
 
+	if (unlikely(task->is_new)) {
+		if (sig == SIGSTOP && !(status >> 16)) {
+			task->event.type = EVENT_NEW;
+			task->event.e_un.signum  = 0;
+
+			return 0;
+		}
+
+		if (!task->bad) {
+			fprintf(stderr, "!!!unexpected state for pid=%d, expected signal SIGSTOP (%d %d)\n", task->pid, sig, status >> 16);
+			task->bad = 1;
+		}
+	}
+
 	switch(status >> 16)  {
+	case 0:
+		break;
 	case PTRACE_EVENT_VFORK:
-		if (child_event(task, EVENT_VFORK))
-			return -1;
 		debug(DEBUG_EVENT, "VFORK: pid=%d, newpid=%d", task->pid, task->event.e_un.newpid);
-		return 0;
+		return child_event(task, EVENT_VFORK);
 	case PTRACE_EVENT_FORK:
-		if (child_event(task, EVENT_FORK))
-			return -1;
 		debug(DEBUG_EVENT, "FORK: pid=%d, newpid=%d", task->pid, task->event.e_un.newpid);
-		return 0;
+		return child_event(task, EVENT_FORK);
 	case PTRACE_EVENT_CLONE:
-		if (child_event(task, EVENT_CLONE))
-			return -1;
 		debug(DEBUG_EVENT, "CLONE: pid=%d, newpid=%d", task->pid, task->event.e_un.newpid);
-		return 0;
+		return child_event(task, EVENT_CLONE);
 	case PTRACE_EVENT_EXEC:
 		task->event.type = EVENT_EXEC;
 		debug(DEBUG_EVENT, "EXEC: pid=%d", task->pid);
@@ -188,64 +212,77 @@ static int _process_event(struct task *task, int status)
 	 {
 		unsigned long data;
 
+		debug(DEBUG_EVENT, "ABOUT_EXIT: pid=%d", task->pid);
+
 		if (unlikely(ptrace(PTRACE_GETEVENTMSG, task->pid, NULL, &data) == -1)) {
 			debug(DEBUG_EVENT, "PTRACE_GETEVENTMSG pid=%d %s", task->pid, strerror(errno));
 			return -1;
 		}
 		task->event.e_un.ret_val = WEXITSTATUS(data);
 		task->event.type = EVENT_ABOUT_EXIT;
-		debug(DEBUG_EVENT, "ABOUT_EXIT: pid=%d", task->pid);
 		return 0;
 	 }
 	default:
+	 	fprintf(stderr, "!!!PTRACE_EVENT_????? pid=%d %d\n", task->pid, status >> 16);
 		break;
 	}
 
-	stop_signal = WSTOPSIG(status);
+	if (!sig)
+		fprintf(stderr, "!!!%s: sig == 0 pid=%d\n", __func__, task->pid);
+
+	if (sig == SIGSTOP) {
+		siginfo_t siginfo;
+
+		if (unlikely(ptrace(PTRACE_GETSIGINFO, task->pid, 0, &siginfo) == -1))
+			sig = 0;
+		else {
+			if (likely(siginfo.si_pid == mtrace_pid))
+				sig = 0;
+			else
+				fprintf(stderr, "!!!%s: SIGSTOP pid=%d %d %d %d %d\n", __func__, task->pid, siginfo.si_signo, siginfo.si_errno, siginfo.si_code, siginfo.si_pid);
+		}
+	}
 
 	task->event.type = EVENT_SIGNAL;
-	task->event.e_un.signum = stop_signal;
+	task->event.e_un.signum = sig;
 
-	debug(DEBUG_EVENT, "SIGNAL: pid=%d, signum=%d", task->pid, stop_signal);
-	return stop_signal;
+	debug(DEBUG_EVENT, "SIGNAL: pid=%d, signum=%d", task->pid, sig);
+	return sig;
 }
 
-static void process_event(struct task *task, int status)
+static struct task * process_event(struct task *task, int status)
 {
-	int stop_signal;
 	struct task *leader = task->leader;
 	struct breakpoint *bp = NULL;
 	arch_addr_t ip;
+	int sig;
 
+	assert(task->stopped == 0);
 	assert(leader != NULL);
 
 	if (unlikely(options.verbose > 1))
 		start_time(&task->halt_time);
 
-	task->stopped = 1;
-
 	leader->threads_stopped++;
 
-	stop_signal = _process_event(task, status);
-
-	if (stop_signal == -1) {
-fprintf(stderr, "%s:%d\n", __FUNCTION__, __LINE__);
-		task->event.type = EVENT_NONE;
+	sig = _process_event(task, status);
+	if (sig < 0) {
 		continue_task(task, 0);
-		return;
+		return NULL;
 	}
 
-	if (stop_signal == 0)
-		return;
+	if (task->event.type == EVENT_NONE) {
+		continue_task(task, task->event.e_un.signum);
+		return NULL;
+	}
 
-	if (unlikely(stop_signal != SIGTRAP))
-		return;
+	if (unlikely(sig != SIGTRAP))
+		return task;
 
 	if (unlikely(fetch_context(task) == -1)) {
-fprintf(stderr, "%s:%d\n", __FUNCTION__, __LINE__);
 		task->event.type = EVENT_NONE;
 		continue_task(task, 0);
-		return;
+		return NULL;
 	}
 
 	ip = get_instruction_pointer(task);
@@ -270,10 +307,8 @@ fprintf(stderr, "%s:%d\n", __FUNCTION__, __LINE__);
 	{
 		bp = breakpoint_find(leader, ip - DECR_PC_AFTER_BREAK);
 		if (unlikely(!bp)) {
-fprintf(stderr, "%s:%d\n", __FUNCTION__, __LINE__);
-//			task->event.type = EVENT_NONE;
-//			continue_task(task, 0);
-			return;
+			fprintf(stderr, "!!!%s: SIGTRAP pid=%d\n", __func__, task->pid);
+			return task;
 		}
 #if HW_BREAKPOINTS > 0
 		assert(bp->type != BP_HW_SCRATCH);
@@ -293,23 +328,7 @@ fprintf(stderr, "%s:%d\n", __FUNCTION__, __LINE__);
 
 	debug(DEBUG_EVENT, "BREAKPOINT: pid=%d, addr=%#lx", task->pid, task->event.e_un.breakpoint->addr);
 
-	return;
-}
-
-static void trace_fail_warning(void)
-{
-#ifdef HAVE_LIBSELINUX
-	if (security_get_boolean_active("deny_ptrace") == 1)
-		fprintf(stderr,
-			"The SELinux boolean 'deny_ptrace' is enabled, which may prevent mtrace-ng\n"
-			"from tracing an other task. You can disable this process attach protection by\n"
-			"issuing 'setsebool deny_ptrace=0' in the superuser context.\n");
-#else
-		fprintf(stderr,
-			"Could not trace! Maybe the SELinux boolean 'deny_ptrace' is enabled, which may\n"
-			"prevent mtrace-ng from tracing an other tasks. Try to disable this process attach\n"
-			"protection by issuing 'setsebool deny_ptrace=0' in the superuser context.\n");
-#endif
+	return task;
 }
 
 void trace_me(void)
@@ -324,25 +343,25 @@ void trace_me(void)
 	}
 }
 
-static inline int fix_signal(struct task *task, int signum)
+static inline int chk_signal(struct task *task, int signum)
 {
-	if (signum == SIGSTOP && task->was_stopped) {
-		task->was_stopped = 0;
-		signum = 0;
-	}
+#if 1
+	if (signum == SIGSTOP)
+		fprintf(stderr, "!!!%s: SIGSTOP pid=%d\n", __func__, task->pid);
+
+	if (signum == SIGTRAP)
+		fprintf(stderr, "!!!%s: SIGTRAP pid=%d\n", __func__, task->pid);
+#endif
+
 	return signum;
 }
 
-int untrace_task(struct task *task, int signum)
+int untrace_task(struct task *task)
 {
-	int ret = 0;
+	int ret;
+	int sig = 0;
 
-	assert(task->leader != NULL);
-
-	debug(DEBUG_PROCESS, "pid=%d", task->pid);
-
-	if (!task->stopped)
-		return 0;
+	assert(task->stopped);
 
 	if (unlikely(ptrace(PTRACE_SETOPTIONS, task->pid, 0, (void *)0) == -1)) {
 		if (errno != ESRCH)
@@ -351,45 +370,53 @@ int untrace_task(struct task *task, int signum)
 		goto skip;
 	}
 
-	signum = fix_signal(task, signum);
+	if (task->event.type == EVENT_SIGNAL || task->event.type == EVENT_NONE)
+		sig = chk_signal(task, sig);
 
-	if (unlikely(ptrace(PTRACE_DETACH, task->pid, 0, signum) == -1)) {
-		if (task->traced) {
-			if (errno != ESRCH)
-				fprintf(stderr, "PTRACE_DETACH pid=%d %s\n", task->pid, strerror(errno));
-			ret = -1;
-			goto skip;
-		}
+	if (unlikely(ptrace(PTRACE_DETACH, task->pid, 0, sig) == -1)) {
+		if (errno != ESRCH)
+			fprintf(stderr, "PTRACE_DETACH pid=%d %s\n", task->pid, strerror(errno));
+		ret = -1;
 	}
 
 	task_kill(task, SIGCONT);
 skip:
 	task->leader->threads_stopped--;
 	task->stopped = 0;
-	task->was_stopped = 0;
-	task->traced = 0;
+	task->attached = 0;
 
 	return ret;
 }
 
+void stop_task(struct task *task)
+{
+	assert(task->attached);
+	assert(task->leader != NULL);
+
+	if (!task->stopped) {
+		int status;
+
+		task_kill(task, SIGSTOP);
+		if (wait_task(task, &status) != -1)
+			_process_event(task, status);
+	}
+}
+
 int trace_attach(struct task *task)
 {
 	debug(DEBUG_PROCESS, "pid=%d", task->pid);
 
-	assert(task->traced == 0);
+	assert(task->attached == 0);
 
 	if (unlikely(ptrace(PTRACE_ATTACH, task->pid, 0, 0) == -1)) {
 		if (errno != ESRCH)
 			fprintf(stderr, "PTRACE_ATTACH pid=%d %s\n", task->pid, strerror(errno));
-		trace_fail_warning();
 		return -1;
 	}
 
 	if (_trace_wait(task, SIGSTOP))
 		return -1;
 
-	queue_event(task);
-
 	return 0;
 }
 
@@ -409,15 +436,22 @@ int trace_set_options(struct task *task)
 
 int continue_task(struct task *task, int signum)
 {
-	debug(DEBUG_PROCESS, "pid=%d", task->pid);
+	debug(DEBUG_PROCESS, "continue task pid=%d", task->pid);
 
 	assert(task->leader != NULL);
 	assert(task->stopped);
 
+	if (signum >= 0x80)
+		fprintf(stderr, "!!!signum >= 0x80 pid=%d: %d\n", task->pid, signum);
+
 	task->leader->threads_stopped--;
 	task->stopped = 0;
+	task->event.type = EVENT_NONE;
 
-	if (unlikely(ptrace(PTRACE_CONT, task->pid, 0, fix_signal(task, signum)) == -1)) {
+	if (signum == SIGTRAP)
+		fprintf(stderr, "!!!%s: SIGTRAP pid=%d\n", __func__, task->pid);
+
+	if (unlikely(ptrace(PTRACE_CONT, task->pid, 0, chk_signal(task, signum)) == -1)) {
 		if (errno != ESRCH)
 			fprintf(stderr, "PTRACE_CONT pid=%d %s\n", task->pid, strerror(errno));
 		return -1;
@@ -427,10 +461,12 @@ int continue_task(struct task *task, int signum)
 
 static void do_stop_cb(struct task *task, void *data)
 {
+	(void)data;
+
 	if (task->stopped)
 		return;
 
-	task->was_stopped = 1;
+	debug(DEBUG_EVENT, "task stop pid=%d", task->pid);
 
 	task_kill(task, SIGSTOP);
 }
@@ -441,6 +477,8 @@ void stop_threads(struct task *task)
 
 	assert(task->leader != NULL);
 
+	debug(DEBUG_EVENT, "stop threads pid=%d", task->pid);
+
 	if (leader->threads != leader->threads_stopped) {
 		struct timespec start;
 
@@ -449,8 +487,12 @@ void stop_threads(struct task *task)
 
 		each_task(leader, &do_stop_cb, NULL);
 
-		while (leader->threads != leader->threads_stopped)
-			queue_event(wait_event());
+		while (leader->threads != leader->threads_stopped) {
+			task = wait_event();
+
+			if (task)
+				queue_event(task);
+		}
 
 		if (unlikely(options.verbose > 1))
 			set_timer(&start, &stop_time);
@@ -460,52 +502,53 @@ void stop_threads(struct task *task)
 int handle_singlestep(struct task *task, int (*singlestep)(struct task *task), struct breakpoint *bp)
 {
 	int status;
-	int stop_signal;
-	unsigned long ip;
+	int sig;
 
-	for(;;) {
-		if (unlikely(singlestep(task) == -1))
-			return -1;
+	assert(task->stopped);
+	assert(task->skip_bp == NULL);
+	assert(bp->enabled == 0);
 
-		if (unlikely(TEMP_FAILURE_RETRY(waitpid(task->pid, &status, __WALL)) != task->pid)) {
-			fprintf(stderr, "%s waitpid pid=%d %s\n", __FUNCTION__,  task->pid, strerror(errno));
-			return 0;
-		}
+	task->event.type = EVENT_NONE;
 
-		stop_signal = _process_event(task, status);
-
-		if (stop_signal == -1)
-			return 0;
-
-		ip = ptrace(PTRACE_PEEKUSER, task->pid, ip_reg_addr(), 0);
-		if (ip == (unsigned long)-1) {
-			fprintf(stderr, "%s ptrace get IP pid=%d %s\n", __FUNCTION__,  task->pid, strerror(errno));
-			return 0;
-		}
-
-		if (ip != bp->addr) {
-			if (likely(stop_signal == SIGTRAP)) {
-				if (bp->break_insn) {
-					queue_event(task);
-					return 1;
-				}
-				return 0;
-			}
-		}
-
-		if (likely(!stop_signal)) {
-			queue_event(task);
-			return 1;
-		}
-
-		if (fix_signal(task, stop_signal) > 0) {
-			queue_event(task);
-			return 1;
-		}
-
-		if (ip != bp->addr)
-			return 0;
+	if (unlikely(singlestep(task) == -1)) {
+		fprintf(stderr, "!!!%s: single step failed pid=%d\n", __func__, task->pid);
+		return -1;
 	}
+
+	if (unlikely(wait_task(task, &status) == -1))
+		return 0;
+
+	sig = _process_event(task, status);
+
+	if (sig == -1) {
+		fprintf(stderr, "!!!%s: failed _process_event pid=%d\n", __func__, task->pid);
+		return 0;
+	}
+
+	assert(task->stopped);
+	assert(task->event.type != EVENT_NONE);
+	assert(task->event.type != EVENT_BREAKPOINT);
+
+	if (task->event.type != EVENT_SIGNAL) {
+		queue_event(task);
+		return 1;
+	}
+
+	if (sig != SIGTRAP) {
+		if (sig == SIGSTOP)
+			fprintf(stderr, "!!!%s: SIGSTOP pid=%d\n", __func__, task->pid);
+		queue_event(task);
+		return 1;
+	}
+
+ 	if (bp->break_insn) {
+		queue_event(task);
+		return 0;
+	}
+
+	task->event.type = EVENT_BREAKPOINT;
+	task->event.e_un.breakpoint = bp;
+	return 0;
 }
 
 #ifndef ARCH_SINGLESTEP
@@ -513,7 +556,7 @@ static int ptrace_singlestep(struct task *task)
 {
 	if (unlikely(ptrace(PTRACE_SINGLESTEP, task->pid, 0, 0) == -1)) {
 		if (errno != ESRCH)
-			fprintf(stderr, "%s PTRACE_SINGLESTEP pid=%d %s\n", __FUNCTION__, task->pid, strerror(errno));
+			fprintf(stderr, "!!!%s: PTRACE_SINGLESTEP pid=%d %s\n", __func__, task->pid, strerror(errno));
 		return -1;
 	}
 	return 0;
@@ -531,14 +574,10 @@ struct task *wait_event(void)
 	int status;
 	int pid;
 
-	pid = waitpid(-1, &status, __WALL);
+	pid = wait_task(NULL, &status);
 	if (unlikely(pid == -1)) {
-		if (errno != EINTR) {
-			if (errno == ECHILD)
-				debug(DEBUG_EVENT, "No more traced programs");
-			else
-				fprintf(stderr, "%s waitpid %s\n", __FUNCTION__, strerror(errno));
-		}
+		if (errno == ECHILD)
+			debug(DEBUG_EVENT, "No more traced programs");
 		return NULL;
 	}
 
@@ -558,12 +597,14 @@ struct task *wait_event(void)
 
 		if (likely(task))
 			trace_setup(task, status, SIGSTOP);
-
 		return NULL;
 	}
 
-	process_event(task, status);
+	assert(!task->stopped);
 
+	task = process_event(task, status);
+	if (task)
+		assert(task->stopped);
 	return task;
 }
 
@@ -618,7 +659,7 @@ ssize_t copy_from_proc(struct task *task, arch_addr_t addr, void *dst, size_t le
 
 		if (errno != EFAULT) {
 			if (errno != ENOSYS) {
-				fprintf(stderr, "%s pid=%d process_vm_readv: %s\n", __FUNCTION__, task->pid, strerror(errno));
+				fprintf(stderr, "!!!%s: pid=%d process_vm_readv: %s\n", __func__, task->pid, strerror(errno));
 				return -1;
 			}
 
@@ -755,10 +796,10 @@ ssize_t copy_str_from_proc(struct task *task, arch_addr_t addr, char *dst, size_
 
 	errno = 0;
 
-	if (--len < 0)
+	if (!len--)
 		return -1;
 
-	while (len) {
+	while(len) {
 		a.a = ptrace(PTRACE_PEEKTEXT, task->pid, addr, 0);
 		if (unlikely(a.a == -1 && errno)) {
 			if (num_bytes && errno == EIO)
diff --git a/sysdeps/linux-gnu/x86/arch.c b/sysdeps/linux-gnu/x86/arch.c
index fcc8d85..b3dbc5d 100644
--- a/sysdeps/linux-gnu/x86/arch.c
+++ b/sysdeps/linux-gnu/x86/arch.c
@@ -106,17 +106,17 @@ static int set_breakpoint_mode(struct task *task, unsigned int n, int type, int
 	uint32_t mode;
 	uint32_t dr7, mask;
 
-	mask = (0b1111 << (16 + 4 * n)) | (0b11 << (2 * n));
+	mask = (0b1111U << (16 + 4 * n)) | (0b11U << (2 * n));
 
 	switch(type) {
 	case BP_X:
-		mode = 0b0000;
+		mode = 0b0000U;
 		break;
 	case BP_W:
-		mode = 0b0001;
+		mode = 0b0001U;
 		break;
 	case BP_RW:
-		mode = 0b0011;
+		mode = 0b0011U;
 		break;
 	default:
 		fprintf(stderr, "invalid hw breakpoint type\n");
@@ -125,16 +125,16 @@ static int set_breakpoint_mode(struct task *task, unsigned int n, int type, int
 
 	switch(len) {
 	case 1:
-		mode |= 0b0000;
+		mode |= 0b0000U;
 		break;
 	case 2:
-		mode |= 0b0100;
+		mode |= 0b0100U;
 		break;
 	case 4:
-		mode |= 0b1100;
+		mode |= 0b1100U;
 		break;
 	case 8:
-		mode |= 0b1000;
+		mode |= 0b1000U;
 		break;
 	}
 
@@ -143,19 +143,19 @@ static int set_breakpoint_mode(struct task *task, unsigned int n, int type, int
 	dr7 |= mode << (16 + 4 * n);
 
 	if (local) {
-		dr7 |= 0b01 << (2 * n);
+		dr7 |= 0b01U << (2 * n);
 		dr7 |= 1 << 8;
 	}
 	else
-	if (!(dr7 & 0b01010101))
+	if (!(dr7 & 0b01010101U))
 		dr7 &= ~(1 << 8);
 
 	if (global) {
-		dr7 |= 0b10 << (2 * n);
+		dr7 |= 0b10U << (2 * n);
 		dr7 |= 1 << 9;
 	}
 	else
-	if (!(dr7 & 0b10101010))
+	if (!(dr7 & 0b10101010U))
 		dr7 &= ~(1 << 9);
 
 	return apply_hw_bp(task, dr7);
@@ -183,14 +183,14 @@ int reset_hw_bp(struct task *task, unsigned int n)
 {
 	uint32_t dr7, mask;
 
-	mask = (0b1111 << (16 + 4 * n)) | (0b11 << (2 * n));
+	mask = (0b1111U << (16 + 4 * n)) | (0b11U << (2 * n));
 
 	dr7 = task->arch.dr7 & ~mask;
 
-	if (!(dr7 & 0b01010101))
+	if (!(dr7 & 0b01010101U))
 		dr7 &= ~(1 << 8);
 
-	if (!(dr7 & 0b10101010))
+	if (!(dr7 & 0b10101010U))
 		dr7 &= ~(1 << 9);
 
 	return apply_hw_bp(task, dr7);
@@ -210,7 +210,7 @@ int arch_task_init(struct task *task)
 {
 	unsigned int i;
 
-	for(i = 0; i < HW_BREAKPOINTS; ++i)
+	for(i = 0; i != HW_BREAKPOINTS; ++i)
 		task->arch.hw_bp[i] = 0;
 
 	return _apply_hw_bp(task, 0);
@@ -223,6 +223,9 @@ void arch_task_destroy(struct task *task)
 
 int arch_task_clone(struct task *retp, struct task *task)
 {
+	(void)retp;
+	(void)task;
+
 	return 0;
 }
 
diff --git a/sysdeps/linux-gnu/x86/dwarf-x86.c b/sysdeps/linux-gnu/x86/dwarf-x86.c
index ff645e8..8b50833 100644
--- a/sysdeps/linux-gnu/x86/dwarf-x86.c
+++ b/sysdeps/linux-gnu/x86/dwarf-x86.c
@@ -137,6 +137,7 @@ static inline int is_signal_frame(struct dwarf_cursor *c)
 
 static inline int is_plt_entry(struct dwarf_addr_space *as)
 {
+	(void)as;
 #if 0
 	struct dwarf_cursor *c = &as->cursor;
 	uint8_t data[12];
@@ -334,7 +335,7 @@ int dwarf_arch_check_call(struct dwarf_addr_space *as, arch_addr_t ip)
 			unsigned int i;
 			unsigned char *addr = libref->image_addr + ip - p->off - libref->load_addr;
 
-			for(i = 0; i < call_op[i].len; ++i) {
+			for(i = 0; i < p->len; ++i) {
 				if (unlikely((addr[i] & p->mask[i]) != p->op[i]))
 					break;
 			}
diff --git a/task.c b/task.c
index 642bda6..926d332 100644
--- a/task.c
+++ b/task.c
@@ -54,15 +54,19 @@ struct pid_hash *pid_hash[PID_HASH_SIZE];
 #ifndef OS_HAVE_PROCESS_DATA
 static inline int os_task_init(struct task *task)
 {
+	(void)task;
 	return 0;
 }
 
 static inline void os_task_destroy(struct task *task)
 {
+	(void)task;
 }
 
-static inline int os_task_clone(struct task *retp, struct task *task)
+static inline int os_task_clone(struct task *task, struct task *newtask)
 {
+	(void)task;
+	(void)newtask;
 	return 0;
 }
 #endif
@@ -70,15 +74,19 @@ static inline int os_task_clone(struct task *retp, struct task *task)
 #ifndef ARCH_HAVE_PROCESS_DATA
 static inline int arch_task_init(struct task *task)
 {
+	(void)task;
 	return 0;
 }
 
 static inline void arch_task_destroy(struct task *task)
 {
+	(void)task;
 }
 
-static inline int arch_task_clone(struct task *retp, struct task *task)
+static inline int arch_task_clone(struct task *task, struct task *newtask)
 {
+	(void)task;
+	(void)newtask;
 	return 0;
 }
 #endif
@@ -101,29 +109,53 @@ static inline void delete_pid(struct task *task)
 
 static inline void insert_pid(struct task *task)
 {
-	struct pid_hash *entry = pid_hash[PID_HASH(task->pid)];
+	unsigned int pidhash = PID_HASH(task->pid);
+	struct pid_hash *entry = pid_hash[pidhash];
 
-	if (!entry->size) {
+	if (!entry) {
 		entry = malloc(sizeof(*entry) + 8 * sizeof(entry->tasks[0]));
 		entry->num = 0;
 		entry->size = 8;
 
-		pid_hash[PID_HASH(task->pid)] = entry;
+		pid_hash[pidhash] = entry;
 	}
 	else
 	if (entry->size == entry->num) {
 		entry->size += 8;
 		entry = realloc(entry, sizeof(*entry) + entry->size * sizeof(entry->tasks[0]));
 
-		pid_hash[PID_HASH(task->pid)] = entry;
+		pid_hash[pidhash] = entry;
 	}
 
 	entry->tasks[entry->num++] = task;
 }
 
-static int leader_setup(struct task *leader)
+struct task *pid2task(pid_t pid)
 {
-	if (!elf_read_main_binary(leader))
+	struct pid_hash *entry = pid_hash[PID_HASH(pid)];
+
+	if (!entry)
+		return NULL;
+
+	struct task **p = entry->tasks;
+	unsigned int n = entry->num;
+
+	while(n) {
+		struct task *task = *p;
+
+		if (likely(task->pid == pid))
+			return task;
+
+		p++;
+		n--;
+	}
+
+	return NULL;
+}
+
+static int leader_setup(struct task *leader, int was_attached)
+{
+	if (!elf_read_main_binary(leader, was_attached))
 		return -1;
 
 	return backtrace_init(leader);
@@ -175,8 +207,6 @@ static int task_init(struct task *task)
 		list_add_tail(&task->task_list, &leader->task_list);
 	}
 
-	task->attached = 1;
-
 	breakpoint_hw_destroy(task);
 
 	return 0;
@@ -216,10 +246,18 @@ static void task_destroy(struct task *task)
 
 	task->deleted = 1;
 
+	stop_task(task);
+
+	if (task->event.type == EVENT_BREAKPOINT)
+		breakpoint_put(task->event.e_un.breakpoint);
+
 	arch_task_destroy(task);
 	os_task_destroy(task);
-	detach_task(task);
+	task_reset_bp(task);
+	remove_event(task);
+	breakpoint_hw_destroy(task);
 	delete_pid(task);
+	untrace_task(task);
 
 	if (leader != task) {
 		list_del(&task->task_list);
@@ -239,9 +277,11 @@ struct task *task_new(pid_t pid)
 	memset(task, 0, sizeof(*task));
 
 	task->pid = pid;
-	task->traced = 0;
+	task->attached = 0;
 	task->stopped = 0;
-	task->was_stopped = 0;
+	task->is_new = 1;
+	task->defer_func = NULL;
+	task->defer_data = NULL;
 
 	INIT_LIST_HEAD(&task->task_list);
 	INIT_LIST_HEAD(&task->leader_list);
@@ -278,8 +318,9 @@ int process_exec(struct task *task)
 {
 	struct task *leader = task->leader;
 
+	breakpoint_invalidate_all(leader);
+
 	each_task(leader, &remove_task_cb, leader);
-	breakpoint_disable_all(leader);
 
 	os_task_destroy(leader);
 	arch_task_destroy(leader);
@@ -290,13 +331,10 @@ int process_exec(struct task *task)
 	if (task_init(leader) < 0)
 		goto fail;
 
-	if (server_connected())
-		task->attached = 0;
-
 	assert(leader->leader == leader);
 	assert(leader->threads_stopped == 1);
 
-	if (leader_setup(leader) < 0)
+	if (leader_setup(leader, 0) < 0)
 		goto fail;
 
 	return 0;
@@ -310,8 +348,6 @@ struct task *task_create(char **argv)
 	struct task *task;
 	pid_t pid;
 
-	debug(DEBUG_FUNCTION, "`%s'", options.command);
-
 	pid = fork();
 	if (pid < 0) {
 		perror("fork");
@@ -320,6 +356,7 @@ struct task *task_create(char **argv)
 
 	if (!pid) { /* child */
 		change_uid();
+
 		trace_me();
 		execvp(options.command, argv);
 		fprintf(stderr, "Can't execute `%s': %s\n", options.command, strerror(errno));
@@ -336,10 +373,10 @@ struct task *task_create(char **argv)
 	if (trace_set_options(task) < 0)
 		goto fail2;
 
-	if (server_connected())
-		task->attached = 0;
+	if (leader_setup(task, 0) < 0)
+		goto fail1;
 
-	if (leader_setup(task) < 0)
+	if (handle_event(task))
 		goto fail1;
 
 	return task;
@@ -354,15 +391,11 @@ fail1:
 
 int task_clone(struct task *task, struct task *newtask)
 {
+	assert(newtask->attached);
 	assert(newtask->leader != newtask);
-	assert(newtask->event.type == EVENT_SIGNAL);
-	assert(newtask->event.e_un.signum == 0);
-	assert(newtask->traced);
-	assert(newtask->stopped);
 	assert(newtask->backtrace == NULL);
 
 	newtask->is_64bit = task->is_64bit;
-	newtask->attached = task->attached;
 
 	breakpoint_hw_clone(newtask);
 
@@ -374,10 +407,7 @@ int task_fork(struct task *task, struct task *newtask)
 	struct task *leader = task->leader;
 
 	assert(newtask->leader == newtask);
-	assert(newtask->event.type == EVENT_SIGNAL);
-	assert(newtask->event.e_un.signum == 0);
-	assert(newtask->traced);
-	assert(newtask->stopped);
+	assert(newtask->attached);
 	assert(newtask->backtrace == NULL);
 
 	newtask->is_64bit = task->is_64bit;
@@ -454,7 +484,9 @@ fail1:
 
 static void show_attached(struct task *task, void *data)
 {
-	fprintf(stderr, "+++ process pid=%d attached (%s) +++\n", task->pid, library_execname(task->leader));
+	(void)data;
+
+	fprintf(stderr, "+++ process pid=%d attached (%s)\n", task->pid, library_execname(task->leader));
 }
 
 
@@ -513,7 +545,7 @@ void open_pid(pid_t pid)
 		old_ntasks = ntasks;
 	}
 
-	if (leader_setup(leader) < 0)
+	if (leader_setup(leader, 1) < 0)
 		goto fail1;
 
 	list_for_each(it, &leader->task_list) {
@@ -550,13 +582,12 @@ void each_task(struct task *leader, void (*cb)(struct task *task, void *data), v
 {
 	struct list_head *it, *next;
 
-	(*cb)(leader, data);
-
 	list_for_each_safe(it, next, &leader->task_list) {
 		struct task *task = container_of(it, struct task, task_list);
 
 		(*cb)(task, data);
 	};
+	(*cb)(leader, data);
 }
 
 void remove_task(struct task *task)
@@ -572,8 +603,18 @@ void remove_proc(struct task *leader)
 
 	assert(leader->leader == leader);
 
-	breakpoint_disable_all(leader);
-	each_task(leader, &remove_task_cb, NULL);
+	each_task(leader, &remove_task_cb, leader);
+	task_destroy(leader);
+}
+
+void untrace_proc(struct task *leader)
+{
+	debug(DEBUG_FUNCTION, "pid=%d", leader->pid);
+
+	assert(leader->leader == leader);
+
+	breakpoint_invalidate_all(leader);
+	remove_proc(leader);
 }
 
 int task_list_empty(void)
@@ -587,26 +628,20 @@ void each_pid(void (*cb)(struct task *task))
 
 	for(i = 0; i < ARRAY_SIZE(pid_hash); ++i) {
 		struct pid_hash *entry = pid_hash[i];
-		unsigned int n = entry->num;
 
-		if (n) {
-			struct task **p = alloca(n * sizeof(*p));
+		if (entry) {
+			unsigned int n = entry->num;
 
-			memcpy(p, entry->tasks, n * sizeof(*p));
+			if (n) {
+				struct task **p = alloca(n * sizeof(*p));
 
-			do {
-				(*cb)(*p++);
-			} while(--n);
+				memcpy(p, entry->tasks, n * sizeof(*p));
+
+				do {
+					(*cb)(*p++);
+				} while(--n);
+			}
 		}
 	}
 }
 
-void init_pid_hash(void)
-{
-	static const struct pid_hash preset = { .size = 0, .num = 0 };
-	unsigned int i;
-
-	for(i = 0; i < ARRAY_SIZE(pid_hash); ++i)
-		pid_hash[i] = (void *)&preset;
-}
-
diff --git a/task.h b/task.h
index cf8ece4..e29e513 100644
--- a/task.h
+++ b/task.h
@@ -43,12 +43,12 @@ struct task {
 	struct event event;
 
 	unsigned int is_64bit:1;
-	unsigned int traced:1;
 	unsigned int attached:1;
 	unsigned int deleted:1;
 	unsigned int about_exit:1;
-	unsigned int was_stopped:1;
 	unsigned int stopped:1;
+	unsigned int is_new:1;
+	unsigned int bad:1;
 
 	struct breakpoint *breakpoint;
 	struct library_symbol *libsym;
@@ -101,6 +101,12 @@ struct task {
 	/* halt time for debugging purpose */
 	struct timespec halt_time;
 
+	/* defered event function */
+	int (*defer_func)(struct task *task, void *data);
+
+	/* defered event data */
+	void *defer_data;
+
 #if HW_BREAKPOINTS > 1
 	/* set in leader: list of hw breakpoints */
 	struct list_head hw_bp_list;
@@ -144,9 +150,15 @@ void each_pid(void (*cb)(struct task *task));
 /* Remove task from the list of traced processes, drop any events in the event queue, destroy it and free memory. */
 void remove_task(struct task *task);
 
+/* invalidate all breakpoints and call remove_proc */
+void untrace_proc(struct task *leader);
+
 /* Remove all threads of the process from the list of traced processes, drop any events in the event queue, destroy it and free memory. */
 void remove_proc(struct task *leader);
 
+/* halt a task */
+void stop_task(struct task *task);
+
 /* return true if no more task is traced */
 int task_list_empty(void);
 
@@ -166,27 +178,6 @@ struct pid_hash {
 #define PID_HASH_SIZE	256
 
 extern struct pid_hash *pid_hash[PID_HASH_SIZE];
-
-void init_pid_hash(void);
-
-static inline struct task *pid2task(pid_t pid)
-{
-	struct pid_hash *entry = pid_hash[PID_HASH(pid)];
-	struct task **p = entry->tasks;
-	unsigned int n = entry->num;
-
-	while(n) {
-		struct task *task = *p;
-
-		if (likely(task->pid == pid))
-			return task;
-
-		p++;
-		n--;
-	}
-
-	return NULL;
-}
-
+extern struct task *pid2task(pid_t pid);
 #endif
 
diff --git a/trace.c b/trace.c
index e244cae..4db6a70 100644
--- a/trace.c
+++ b/trace.c
@@ -46,15 +46,18 @@ int skip_breakpoint(struct task *task, struct breakpoint *bp)
 {
 	debug(DEBUG_PROCESS, "pid=%d, addr=%#lx", task->pid, bp->addr);
 
-	if (task->event.type != EVENT_NONE)
-		return 1;
+	assert(task->event.type == EVENT_BREAKPOINT);
+	assert(task->stopped);
+	assert(task->skip_bp == NULL);
 
 	if (bp->enabled && !bp->hw) {
 		int ret = 0;
 		struct timespec start;
 
-		if (task->skip_bp)
+		if (task->skip_bp) {
+			task->event.type = EVENT_NONE;
 			return 1;
+		}
 
 		if (unlikely(options.verbose > 1))
 			start_time(&start);
@@ -68,6 +71,7 @@ int skip_breakpoint(struct task *task, struct breakpoint *bp)
 
 		if (unlikely(ret)) {
 			task->skip_bp = breakpoint_get(bp);
+			assert(task->skip_bp);
 			return ret;
 		}
 	}
@@ -85,26 +89,10 @@ void fix_about_exit(struct task *task)
 	}
 }
 
-void detach_task(struct task *task)
-{
-	int sig = 0;
-
-	task_reset_bp(task);
-
-	if (task->event.type == EVENT_SIGNAL)
-		sig = task->event.e_un.signum;
-	else
-	if (task->event.type == EVENT_BREAKPOINT)
-		breakpoint_put(task->event.e_un.breakpoint);
-
-	remove_event(task);
-	breakpoint_hw_destroy(task);
-	fix_about_exit(task);
-	untrace_task(task, sig);
-}
-
 static void detach_cb(struct task *task, void *data)
 {
+	(void)data;
+
 	remove_task(task);
 }
 
@@ -115,7 +103,7 @@ void detach_proc(struct task *leader)
 	breakpoint_disable_all(leader);
 
 	if (unlikely(options.verbose > 1))
-		fprintf(stderr, "+++ process detach pid=%d +++\n", leader->pid);
+		fprintf(stderr, "+++ process detach pid=%d\n", leader->pid);
 
 	each_task(leader, &detach_cb, NULL);
 }
diff --git a/trace.h b/trace.h
index 3d18ea2..168855c 100644
--- a/trace.h
+++ b/trace.h
@@ -26,7 +26,6 @@
 #include "forward.h"
 
 void fix_about_exit(struct task *task);
-void detach_task(struct task *task);
 void detach_proc(struct task *leader);
 int skip_breakpoint(struct task *task, struct breakpoint *bp);