rktio: repairs

All tests now pass for MacOS.

pkgs/racket-doc/scribblings/inside/ports.scrbl

@@ -785,7 +785,15 @@ If @var{read_too} is non-zero, the function produces multiple values
 Creates Racket input and output ports for a TCP socket @var{s}. The
  @var{name} argument supplies the name for the ports. If @var{close}
  is non-zero, then the ports assume responsibility for closing the
- socket. The resulting ports are written to @var{inp} and @var{outp}.}
+ socket. The resulting ports are written to @var{inp} and @var{outp}.
+
+Whether @var{close} is zero or not, closing the resulting ports
+ unregisters the file descriptor with @cpp{scheme_fd_to_semaphore}.
+ So, passing zero for @var{close} and also using the file descriptor
+ with other ports or with @cpp{scheme_fd_to_semaphore} will not work right.
+
+@history["6.9.0.6" @elem{Changed ports to always unregister with @cpp{scheme_fd_to_semaphore},
+                         since it's not safe to skip that step.}]}
 
 
 @function[(Scheme_Object* scheme_fd_to_semaphore

pkgs/racket-test-core/tests/racket/optimize.rktl

@@ -4975,8 +4975,12 @@
   (set! s? f) ; break the JIT's optimistic assumption
   
   (define (go)
+    (define init-size
+      (let ([vec (make-vector 6)])
+        (vector-set-performance-stats! vec (current-thread))
+        (vector-ref vec 3)))
     (define size (f 500000)) ; make sure that this still leads to a tail loop
-    (size . < . 80000)))
+    ((- size init-size) . < . 20000)))
 
 (test #t (dynamic-require ''check-tail-call-by-jit-for-struct-predicate 'go))
 

pkgs/racket-test-core/tests/racket/thread.rktl

@@ -728,6 +728,7 @@
 		(loop))))
 
 	  (close-output-port sw)
+	  (test cr sync cr)
 	  (test cr sync s t l sr cr)
 	  (test cr sync s t l sr cr)
 

pkgs/racket-test-core/tests/racket/udp.rktl

@@ -317,7 +317,7 @@
 		((b) (make-bytes 8 0)))
     (test (void) udp-multicast-set-interface! s2 "localhost")
     (test (void) udp-send-to s2 "233.252.0.0" lp #"hi")
-    (sleep 0.05)
+    (sleep 0.05) ; (sync (udp-receive-ready-evt s))
     (let-values (((packet-length ra1 rp1) (udp-receive!* s b)))
       (test 2 values packet-length)
       (test #"hi\0\0\0\0\0\0" values b))
@@ -334,8 +334,7 @@
   (test (void) udp-close s)
   ;; It's closed
   (err/rt-test (udp-multicast-loopback? s) exn:fail:network?)
-  (err/rt-test (udp-multicast-set-loopback! s #t) exn:fail:network?)
+  (err/rt-test (udp-multicast-set-loopback! s #t) exn:fail:network?))
-  )
 
 ;; ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
 

racket/src/racket/src/env.c

@@ -551,7 +551,6 @@ static Scheme_Env *place_instance_init(void *stack_base, int initial_main_os_thr
   scheme_init_sema_places();
   scheme_init_gmp_places();
   scheme_init_fd_semaphores();
-  scheme_alloc_global_fdset();
 #ifndef DONT_USE_FOREIGN
   scheme_init_foreign_places();
 #endif
@@ -676,7 +675,6 @@ void scheme_place_instance_destroy(int force)
   GC_destruct_child_gc();
 #endif
   scheme_free_all_code();
-  scheme_free_global_fdset();
   rktio_destroy(scheme_rktio);
 }
 

racket/src/racket/src/mzmark_network.inc

@@ -1,6 +1,5 @@
 /* >>>> Generated by mkmark.rkt from mzmarksrc.c <<<< */
 
-#ifdef USE_TCP
 static int mark_listener_SIZE(void *p, struct NewGC *gc) {
 #ifndef GC_NO_SIZE_NEEDED_FROM_PROCS
   listener_t *l = (listener_t *)p;
@@ -91,7 +90,6 @@ static int mark_tcp_FIXUP(void *p, struct NewGC *gc) {
 #define mark_tcp_IS_CONST_SIZE 1
 
 
-# ifdef UDP_IS_SUPPORTED
 static int mark_udp_SIZE(void *p, struct NewGC *gc) {
 #ifndef GC_NO_SIZE_NEEDED_FROM_PROCS
   gcBYTES_TO_WORDS(sizeof(Scheme_UDP));
@@ -179,6 +177,4 @@ static int mark_udp_evt_FIXUP(void *p, struct NewGC *gc) {
 #define mark_udp_evt_IS_ATOMIC 0
 #define mark_udp_evt_IS_CONST_SIZE 1
 
-# endif
-#endif
 

racket/src/racket/src/mzmarksrc.c

@@ -1818,7 +1818,6 @@ END print;
 
 START network;
 
-#ifdef USE_TCP
 mark_listener {
   listener_t *l = (listener_t *)p;
 
@@ -1841,7 +1840,6 @@ mark_tcp {
   gcBYTES_TO_WORDS(sizeof(Scheme_Tcp));
 }
 
-# ifdef UDP_IS_SUPPORTED
 mark_udp {
  mark:
   Scheme_UDP *udp = (Scheme_UDP *)p;
@@ -1863,8 +1861,6 @@ mark_udp_evt {
  size:
   gcBYTES_TO_WORDS(sizeof(Scheme_UDP_Evt));
 }
-# endif
-#endif
 
 END network;
 

racket/src/racket/src/network.c

@@ -188,6 +188,14 @@ static int check_fd_sema(rktio_fd_t *s, int mode, Scheme_Schedule_Info *sinfo, S
   sema = scheme_rktio_fd_to_semaphore(s, mode);
   
   if (sema) {
+    /* It would make sense to force a poll via
+       scheme_check_fd_semaphores() here, although we'd only want to
+       to that once per scheduler cycle. That would more reliably poll
+       at the OS level, since we otherwise wait on the scheduler to
+       check semaphores. It's not clear that the OS supports more
+       precise reasoning about readiness, though, and Racket has
+       traditonally not done that, so we're still skipping
+       it. */
     if (!scheme_wait_sema(sema, 1)) {
       if (sinfo && !sinfo->no_redirect)
         scheme_set_sync_target(sinfo, sema, orig, NULL, 0, 0, NULL);
@@ -424,7 +432,7 @@ static intptr_t tcp_get_string(Scheme_Input_Port *port,
 		     "tcp-read: error reading\n"
                      "  system error: %R");
     return 0;
-  } else if (data->b.bufmax == RKTIO_READ_ERROR) {
+  } else if (data->b.bufmax == RKTIO_READ_EOF) {
     data->b.bufmax = 0;
     data->b.hiteof = 1;
     return EOF;
@@ -809,7 +817,7 @@ static void wait_until_lookup(Connect_Progress_Data *pd)
   }
 }
 
-static rktio_addrinfo_t *do_resolve_address(const char *who, char *address, int id, int family, int show_id_on_error)
+static rktio_addrinfo_t *do_resolve_address(const char *who, char *address, int id, int family, int passive, int show_id_on_error)
 {
   Connect_Progress_Data *pd;
   rktio_addrinfo_lookup_t *lookup;
@@ -817,7 +825,7 @@ static rktio_addrinfo_t *do_resolve_address(const char *who, char *address, int
     
   pd = make_connect_progress_data();    
     
-  lookup = rktio_start_addrinfo_lookup(scheme_rktio, address, id, family, 0, 0);
+  lookup = rktio_start_addrinfo_lookup(scheme_rktio, address, id, family, passive, 0);
   if (!lookup) {
     addr = NULL;
   } else {
@@ -1758,7 +1766,7 @@ static Scheme_Object *make_udp(int argc, Scheme_Object *argv[])
     if (!id)
       id = 1025;
 
-    addr = do_resolve_address("upd-open-socket", address, id, RKTIO_FAMILY_ANY, show_id_on_error);
+    addr = do_resolve_address("upd-open-socket", address, id, RKTIO_FAMILY_ANY, 0, show_id_on_error);
   } else
     addr = NULL;
 
@@ -1894,10 +1902,11 @@ static Scheme_Object *udp_bind_or_connect(const char *name, int argc, Scheme_Obj
                          "  system error: %R",
                          name);
       }
+      udp->connected = 0;
       return scheme_void;
     }
 
-    addr = do_resolve_address(name, address, port, RKTIO_FAMILY_ANY, 1);
+    addr = do_resolve_address(name, address, port, RKTIO_FAMILY_ANY, do_bind, 1);
 
     if (!do_bind) {
       /* CONNECT CASE */
@@ -1918,6 +1927,8 @@ static Scheme_Object *udp_bind_or_connect(const char *name, int argc, Scheme_Obj
 
         return NULL;
       }
+
+      udp->connected = 1;
     } else {
       /* BIND CASE */
       int ok;
@@ -1938,6 +1949,8 @@ static Scheme_Object *udp_bind_or_connect(const char *name, int argc, Scheme_Obj
                          port);
         return NULL;
       }
+
+      udp->bound = 1;
     }
   }
 
@@ -2101,7 +2114,7 @@ static Scheme_Object *udp_send_it(const char *name, int argc, Scheme_Object *arg
 
     scheme_security_check_network(name, address, id, 1);
 
-    dest_addr = do_resolve_address(name, address, id, RKTIO_FAMILY_ANY, 1);
+    dest_addr = do_resolve_address(name, address, id, RKTIO_FAMILY_ANY, 0, 1);
   } else {
     dest_addr = NULL;
   }
@@ -2322,23 +2335,17 @@ static Scheme_Object *udp_receive_enable_break(int argc, Scheme_Object *argv[])
 
 static Scheme_Object *make_udp_evt(const char *name, int argc, Scheme_Object **argv, int for_read)
 {
-#ifdef UDP_IS_SUPPORTED
   Scheme_UDP_Evt *uw;
-#endif
 
   if (!SCHEME_UDPP(argv[0]))
     scheme_wrong_contract(name, "udp?", 0, argc, argv);
 
-#ifdef UDP_IS_SUPPORTED
   uw = MALLOC_ONE_TAGGED(Scheme_UDP_Evt);
   uw->so.type = scheme_udp_evt_type;
   uw->udp = (Scheme_UDP *)argv[0];
   uw->for_read = for_read;
 
   return (Scheme_Object *)uw;
-#else
-  return scheme_void;
-#endif
 }
 
 static Scheme_Object *udp_read_ready_evt(int argc, Scheme_Object *argv[])
@@ -2552,7 +2559,7 @@ udp_multicast_set_interface(int argc, Scheme_Object *argv[])
 
   if (SCHEME_CHAR_STRINGP(argv[1])) {
     bs = scheme_char_string_to_byte_string(argv[1]);
-    addr = do_resolve_address("udp-multicast-set-interface!", SCHEME_BYTE_STR_VAL(bs), -1, udp_default_family(), 0);
+    addr = do_resolve_address("udp-multicast-set-interface!", SCHEME_BYTE_STR_VAL(bs), -1, udp_default_family(), 0, 0);
   } else
     addr = NULL;
 
@@ -2581,7 +2588,6 @@ do_udp_multicast_join_or_leave_group(char const *name, int optname, Scheme_UDP *
 
   pd = make_connect_progress_data();
 
-  
   bs = scheme_char_string_to_byte_string(multiaddrname);
   address = SCHEME_BYTE_STR_VAL(bs);
 
@@ -2640,7 +2646,6 @@ do_udp_multicast_join_or_leave_group(char const *name, int optname, Scheme_UDP *
   rktio_addrinfo_free(scheme_rktio, multi_addr);
   if (intf_addr) rktio_addrinfo_free(scheme_rktio, intf_addr);
 
-
   if (!r) 
     scheme_raise_exn(MZEXN_FAIL_NETWORK,
 		     "%s: setsockopt failed\n"
@@ -2704,14 +2709,10 @@ START_XFORM_SKIP;
 
 static void register_traversers(void)
 {
-#ifdef USE_TCP
   GC_REG_TRAV(scheme_listener_type, mark_listener);
   GC_REG_TRAV(scheme_rt_tcp, mark_tcp);
-# ifdef UDP_IS_SUPPORTED
   GC_REG_TRAV(scheme_udp_type, mark_udp);
   GC_REG_TRAV(scheme_udp_evt_type, mark_udp_evt);
-# endif
-#endif
 }
 
 END_XFORM_SKIP;

racket/src/racket/src/port.c

@@ -484,14 +484,6 @@ void scheme_set_stdio_makers(Scheme_Stdio_Maker_Proc in,
 /*                                fd arrays                               */
 /*========================================================================*/
 
-void scheme_alloc_global_fdset() {
-  scheme_semaphore_fd_set = rktio_ltps_open(scheme_rktio);
-}
-
-void scheme_free_global_fdset(void) {
-  rktio_ltps_close(scheme_rktio, scheme_semaphore_fd_set);
-}
-
 void scheme_add_fd_handle(void *h, void *fds, int repost)
 {
 #if defined(WIN32_FD_HANDLES)
@@ -3596,7 +3588,7 @@ scheme_do_open_input_file(char *name, int offset, int argc, Scheme_Object *argv[
                                            | (text_mode ? RKTIO_OPEN_TEXT : 0)));
 
   if (!fd) {
-    filename_exn(name, "cannot open input file", filename, RKTIO_ERROR_DOES_NOT_EXIST);
+    filename_exn(name, "cannot open input file", filename, (for_module ? RKTIO_ERROR_DOES_NOT_EXIST : 0));
     return NULL;
   }
   
@@ -4355,9 +4347,11 @@ void scheme_filesystem_change_evt_cancel(Scheme_Object *evt, void *ignored_data)
 {
   Scheme_Filesystem_Change_Evt *fc = (Scheme_Filesystem_Change_Evt *)evt;
 
+  if (fc->rfc) {
     rktio_fs_change_forget(scheme_rktio, fc->rfc);
     fc->rfc = NULL;
   }
+}
 
 static int filesystem_change_evt_ready(Scheme_Object *evt, Scheme_Schedule_Info *sinfo)
 {
@@ -5881,7 +5875,7 @@ static Scheme_Object *subprocess(int c, Scheme_Object *args[])
   rktio_fd_t *stdout_fd = NULL;
   rktio_fd_t *stdin_fd = NULL;
   rktio_fd_t *stderr_fd = NULL;
-  int need_forget_out = 0, need_forget_in = 0, need_forget_err = 0, new_process_group = 0;
+  int need_forget_out = 0, need_forget_in = 0, need_forget_err = 0;
   rktio_envvars_t *envvars;
   rktio_process_result_t *result;
   Scheme_Config *config;
@@ -6023,9 +6017,10 @@ static Scheme_Object *subprocess(int c, Scheme_Object *args[])
   config = scheme_current_config();
   
   cust_mode = scheme_get_param(config, MZCONFIG_SUBPROC_GROUP_ENABLED);
-  new_process_group = SCHEME_TRUEP(cust_mode);
+  if (SCHEME_TRUEP(cust_mode))
-  cust_mode = scheme_get_param(config, MZCONFIG_SUBPROC_CUSTODIAN_MODE);
+    flags |= RKTIO_PROCESS_NEW_GROUP;
 
+  cust_mode = scheme_get_param(config, MZCONFIG_SUBPROC_CUSTODIAN_MODE);
   if (SCHEME_SYMBOLP(cust_mode)
       && !strcmp(SCHEME_SYM_VAL(cust_mode), "kill")
       && (rktio_process_allowed_flags(scheme_rktio) & RKTIO_PROCESS_WINDOWS_CHAIN_TERMINATION))

racket/src/racket/src/schpriv.h

@@ -435,7 +435,6 @@ void scheme_register_network_evts();
 
 void scheme_free_dynamic_extensions(void);
 void scheme_free_all_code(void);
-void scheme_free_global_fdset(void);
 
 XFORM_NONGCING int scheme_is_multithreaded(int now);
 
@@ -4405,7 +4404,6 @@ int scheme_is_user_port(Scheme_Object *port);
 int scheme_byte_ready_or_user_port_ready(Scheme_Object *p, Scheme_Schedule_Info *sinfo);
 
 int scheme_pipe_char_count(Scheme_Object *p);
-void scheme_alloc_global_fdset();
 Scheme_Object *scheme_port_name(Scheme_Object *p);
 intptr_t scheme_port_closed_p (Scheme_Object *port);
 

racket/src/rktio/rktio_ltps.c

@@ -73,6 +73,7 @@ rktio_ltps_handle_t *make_ltps_handle()
   rktio_ltps_handle_t *s;
   s = malloc(sizeof(rktio_ltps_handle_t));
   s->data = NULL;
+  s->next = NULL;
   return s;
 }
 

racket/src/rktio/rktio_network.c

@@ -927,7 +927,7 @@ int rktio_socket_shutdown(rktio_t *rktio, rktio_fd_t *rfd, int mode)
 {
   rktio_socket_t s = rktio_fd_system_fd(rktio, rfd);
   
-  if (!shutdown(s, ((mode == RKTIO_SHUTDOWN_READ) ? SHUT_RD : SHUT_RDWR))) {
+  if (shutdown(s, ((mode == RKTIO_SHUTDOWN_READ) ? SHUT_RD : SHUT_WR))) {
     get_socket_error();
     return 0;
   }
@@ -1129,10 +1129,6 @@ rktio_connect_t *rktio_start_connect(rktio_t *rktio, rktio_addrinfo_t *dest, rkt
 {
   rktio_connect_t *conn;
 
-#ifdef USE_TCP
-  TCP_INIT("tcp-connect");
-#endif
-
   conn = malloc(sizeof(rktio_connect_t));
   conn->dest = dest;
   conn->src = src;
@@ -1162,7 +1158,9 @@ static rktio_connect_t *try_connect(rktio_t *rktio, rktio_connect_t *conn)
       fcntl(s, F_SETFL, RKTIO_NONBLOCKING);
       RKTIO_WHEN_SET_SOCKBUF_SIZE(setsockopt(s, SOL_SOCKET, SO_SNDBUF, (char *)&size, sizeof(int)));
 #endif
+
       status = connect(s, RKTIO_AS_ADDRINFO(addr)->ai_addr, RKTIO_AS_ADDRINFO(addr)->ai_addrlen);
+
 #ifdef RKTIO_SYSTEM_UNIX
       if (status)
         status = errno;
@@ -1192,15 +1190,18 @@ static rktio_connect_t *try_connect(rktio_t *rktio, rktio_connect_t *conn)
 
 int rktio_poll_connect_ready(rktio_t *rktio, rktio_connect_t *conn)
 {
-  if (!conn->inprogress) {
+  if (conn->inprogress)
     return rktio_socket_poll_write_ready(rktio, conn->trying_fd);
-  } else
+  else
     return RKTIO_POLL_READY;
 }
 
 void rktio_poll_add_connect(rktio_t *rktio, rktio_connect_t *conn, rktio_poll_set_t *fds)
 {
+  if (conn->inprogress)
     rktio_poll_add(rktio, conn->trying_fd, fds, RKTIO_POLL_WRITE);
+  else
+    rktio_poll_set_add_nosleep(rktio, fds);
 }
 
 static void conn_free(rktio_connect_t *conn)
@@ -1709,24 +1710,7 @@ rktio_fd_t *rktio_udp_open(rktio_t *rktio, rktio_addrinfo_t *addr, int family)
     return NULL;
   }
 
-#ifdef RKTIO_SYSTEM_WINDOWS
+  return rktio_system_fd(rktio, s, RKTIO_OPEN_SOCKET | RKTIO_OPEN_UDP | RKTIO_OPEN_INIT);
-  {
-    unsigned long ioarg = 1;
-    BOOL bc = 1;
-    ioctlsocket(s, FIONBIO, &ioarg);
-    setsockopt(s, SOL_SOCKET, SO_BROADCAST, (char *)(&bc), sizeof(BOOL));
-  }
-#else
-  fcntl(s, F_SETFL, RKTIO_NONBLOCKING);
-# ifdef SO_BROADCAST
-  {
-    int bc = 1;
-    setsockopt(s, SOL_SOCKET, SO_BROADCAST, &bc, sizeof(bc));
-  }
-# endif
-#endif
-
-  return rktio_system_fd(rktio, s, RKTIO_OPEN_SOCKET | RKTIO_OPEN_UDP);
 }
 
 #ifdef UDP_DISCONNECT_EADRNOTAVAIL_OK
@@ -1935,6 +1919,7 @@ char *rktio_udp_multicast_interface(rktio_t *rktio, rktio_fd_t *rfd)
   int status;
   
   status = getsockopt(s, IPPROTO_IP, IP_MULTICAST_IF, (void *)&intf, &intf_len);
+
   if (status) {
     get_socket_error();
     return NULL;
@@ -1960,6 +1945,7 @@ int rktio_udp_set_multicast_interface(rktio_t *rktio, rktio_fd_t *rfd, rktio_add
   }
 
   status = setsockopt(s, IPPROTO_IP, IP_MULTICAST_IF, (void *)&intf, intf_len);
+
   if (status) {
     get_socket_error();
     return 0;
@@ -1988,10 +1974,8 @@ int rktio_udp_change_multicast_group(rktio_t *rktio, rktio_fd_t *rfd,
 
   if (action == RKTIO_ADD_MEMBERSHIP)
     optname = IP_ADD_MEMBERSHIP;
-  else if (action == RKTIO_DROP_MEMBERSHIP)
-    optname = IP_DROP_MEMBERSHIP;
   else
-    optname = 0;
+    optname = IP_DROP_MEMBERSHIP;
 
   status = setsockopt(s, IPPROTO_IP, optname, (void *) &mreq, mreq_len);
   

racket/src/rktio/rktio_pipe.c

@@ -87,7 +87,7 @@ rktio_fd_t **rktio_make_pipe(rktio_t *rktio, int flags)
   intptr_t a[2];
   rktio_fd_t **rfds;
   
-  if (!rktio_make_os_pipe(rktio, a, flags))
+  if (rktio_make_os_pipe(rktio, a, flags))
     return NULL;
 
   rfds = malloc(sizeof(rktio_fd_t*) * 2);