2001-01-09 Tor Lillqvist <tml@iki.fi>
+ * giowin32.c (create_reader_thread): Rework the changes needed to
+ pass mainloop-test. Now we don't need to call TerminateThread()
+ after all, which is a relief, as the docs have a BIG RED WARNING
+ SIGN about using that API. Instead, when closing a fd channel that
+ has a reader thread running, just mark it as non-running and
+ additionally mark the fd as ripe for closing. When the reader
+ thread hopefully eventually gets something (and EOF or some actual
+ data), it will note that it shouldn't be running, break out of the
+ loop, and close the fd.
+
+ The socket channel closing code should probably be changed
+ similarily, but that will have to wait until I have a test case.
+
+ (g_pipe_readable_msg, g_io_channel_win32_new_pipe,
+ g_io_channel_win32_new_pipe_with_wakeups,
+ g_io_channel_win32_pipe_request_wakeups,
+ g_io_channel_win32_pipe_readable): Remove these, have been
+ obsolete for some time.
+
* gutils.c (g_basename, g_dirname): Don't warn about deprecation
on Win32. Code written for GLib 1.2 doesn't have much choice but
to use GLib >= 1.3 on Win32.
2001-01-09 Tor Lillqvist <tml@iki.fi>
+ * giowin32.c (create_reader_thread): Rework the changes needed to
+ pass mainloop-test. Now we don't need to call TerminateThread()
+ after all, which is a relief, as the docs have a BIG RED WARNING
+ SIGN about using that API. Instead, when closing a fd channel that
+ has a reader thread running, just mark it as non-running and
+ additionally mark the fd as ripe for closing. When the reader
+ thread hopefully eventually gets something (and EOF or some actual
+ data), it will note that it shouldn't be running, break out of the
+ loop, and close the fd.
+
+ The socket channel closing code should probably be changed
+ similarily, but that will have to wait until I have a test case.
+
+ (g_pipe_readable_msg, g_io_channel_win32_new_pipe,
+ g_io_channel_win32_new_pipe_with_wakeups,
+ g_io_channel_win32_pipe_request_wakeups,
+ g_io_channel_win32_pipe_readable): Remove these, have been
+ obsolete for some time.
+
* gutils.c (g_basename, g_dirname): Don't warn about deprecation
on Win32. Code written for GLib 1.2 doesn't have much choice but
to use GLib >= 1.3 on Win32.
2001-01-09 Tor Lillqvist <tml@iki.fi>
+ * giowin32.c (create_reader_thread): Rework the changes needed to
+ pass mainloop-test. Now we don't need to call TerminateThread()
+ after all, which is a relief, as the docs have a BIG RED WARNING
+ SIGN about using that API. Instead, when closing a fd channel that
+ has a reader thread running, just mark it as non-running and
+ additionally mark the fd as ripe for closing. When the reader
+ thread hopefully eventually gets something (and EOF or some actual
+ data), it will note that it shouldn't be running, break out of the
+ loop, and close the fd.
+
+ The socket channel closing code should probably be changed
+ similarily, but that will have to wait until I have a test case.
+
+ (g_pipe_readable_msg, g_io_channel_win32_new_pipe,
+ g_io_channel_win32_new_pipe_with_wakeups,
+ g_io_channel_win32_pipe_request_wakeups,
+ g_io_channel_win32_pipe_readable): Remove these, have been
+ obsolete for some time.
+
* gutils.c (g_basename, g_dirname): Don't warn about deprecation
on Win32. Code written for GLib 1.2 doesn't have much choice but
to use GLib >= 1.3 on Win32.
2001-01-09 Tor Lillqvist <tml@iki.fi>
+ * giowin32.c (create_reader_thread): Rework the changes needed to
+ pass mainloop-test. Now we don't need to call TerminateThread()
+ after all, which is a relief, as the docs have a BIG RED WARNING
+ SIGN about using that API. Instead, when closing a fd channel that
+ has a reader thread running, just mark it as non-running and
+ additionally mark the fd as ripe for closing. When the reader
+ thread hopefully eventually gets something (and EOF or some actual
+ data), it will note that it shouldn't be running, break out of the
+ loop, and close the fd.
+
+ The socket channel closing code should probably be changed
+ similarily, but that will have to wait until I have a test case.
+
+ (g_pipe_readable_msg, g_io_channel_win32_new_pipe,
+ g_io_channel_win32_new_pipe_with_wakeups,
+ g_io_channel_win32_pipe_request_wakeups,
+ g_io_channel_win32_pipe_readable): Remove these, have been
+ obsolete for some time.
+
* gutils.c (g_basename, g_dirname): Don't warn about deprecation
on Win32. Code written for GLib 1.2 doesn't have much choice but
to use GLib >= 1.3 on Win32.
2001-01-09 Tor Lillqvist <tml@iki.fi>
+ * giowin32.c (create_reader_thread): Rework the changes needed to
+ pass mainloop-test. Now we don't need to call TerminateThread()
+ after all, which is a relief, as the docs have a BIG RED WARNING
+ SIGN about using that API. Instead, when closing a fd channel that
+ has a reader thread running, just mark it as non-running and
+ additionally mark the fd as ripe for closing. When the reader
+ thread hopefully eventually gets something (and EOF or some actual
+ data), it will note that it shouldn't be running, break out of the
+ loop, and close the fd.
+
+ The socket channel closing code should probably be changed
+ similarily, but that will have to wait until I have a test case.
+
+ (g_pipe_readable_msg, g_io_channel_win32_new_pipe,
+ g_io_channel_win32_new_pipe_with_wakeups,
+ g_io_channel_win32_pipe_request_wakeups,
+ g_io_channel_win32_pipe_readable): Remove these, have been
+ obsolete for some time.
+
* gutils.c (g_basename, g_dirname): Don't warn about deprecation
on Win32. Code written for GLib 1.2 doesn't have much choice but
to use GLib >= 1.3 on Win32.
2001-01-09 Tor Lillqvist <tml@iki.fi>
+ * giowin32.c (create_reader_thread): Rework the changes needed to
+ pass mainloop-test. Now we don't need to call TerminateThread()
+ after all, which is a relief, as the docs have a BIG RED WARNING
+ SIGN about using that API. Instead, when closing a fd channel that
+ has a reader thread running, just mark it as non-running and
+ additionally mark the fd as ripe for closing. When the reader
+ thread hopefully eventually gets something (and EOF or some actual
+ data), it will note that it shouldn't be running, break out of the
+ loop, and close the fd.
+
+ The socket channel closing code should probably be changed
+ similarily, but that will have to wait until I have a test case.
+
+ (g_pipe_readable_msg, g_io_channel_win32_new_pipe,
+ g_io_channel_win32_new_pipe_with_wakeups,
+ g_io_channel_win32_pipe_request_wakeups,
+ g_io_channel_win32_pipe_readable): Remove these, have been
+ obsolete for some time.
+
* gutils.c (g_basename, g_dirname): Don't warn about deprecation
on Win32. Code written for GLib 1.2 doesn't have much choice but
to use GLib >= 1.3 on Win32.
2001-01-09 Tor Lillqvist <tml@iki.fi>
+ * giowin32.c (create_reader_thread): Rework the changes needed to
+ pass mainloop-test. Now we don't need to call TerminateThread()
+ after all, which is a relief, as the docs have a BIG RED WARNING
+ SIGN about using that API. Instead, when closing a fd channel that
+ has a reader thread running, just mark it as non-running and
+ additionally mark the fd as ripe for closing. When the reader
+ thread hopefully eventually gets something (and EOF or some actual
+ data), it will note that it shouldn't be running, break out of the
+ loop, and close the fd.
+
+ The socket channel closing code should probably be changed
+ similarily, but that will have to wait until I have a test case.
+
+ (g_pipe_readable_msg, g_io_channel_win32_new_pipe,
+ g_io_channel_win32_new_pipe_with_wakeups,
+ g_io_channel_win32_pipe_request_wakeups,
+ g_io_channel_win32_pipe_readable): Remove these, have been
+ obsolete for some time.
+
* gutils.c (g_basename, g_dirname): Don't warn about deprecation
on Win32. Code written for GLib 1.2 doesn't have much choice but
to use GLib >= 1.3 on Win32.
2001-01-09 Tor Lillqvist <tml@iki.fi>
+ * giowin32.c (create_reader_thread): Rework the changes needed to
+ pass mainloop-test. Now we don't need to call TerminateThread()
+ after all, which is a relief, as the docs have a BIG RED WARNING
+ SIGN about using that API. Instead, when closing a fd channel that
+ has a reader thread running, just mark it as non-running and
+ additionally mark the fd as ripe for closing. When the reader
+ thread hopefully eventually gets something (and EOF or some actual
+ data), it will note that it shouldn't be running, break out of the
+ loop, and close the fd.
+
+ The socket channel closing code should probably be changed
+ similarily, but that will have to wait until I have a test case.
+
+ (g_pipe_readable_msg, g_io_channel_win32_new_pipe,
+ g_io_channel_win32_new_pipe_with_wakeups,
+ g_io_channel_win32_pipe_request_wakeups,
+ g_io_channel_win32_pipe_readable): Remove these, have been
+ obsolete for some time.
+
* gutils.c (g_basename, g_dirname): Don't warn about deprecation
on Win32. Code written for GLib 1.2 doesn't have much choice but
to use GLib >= 1.3 on Win32.
gboolean running; /* Is reader thread running. FALSE if
* EOF has been reached.
*/
+ gboolean needs_close; /* If the channel has been closed while
+ * the reader thread was still running.
+ */
guint thread_id; /* If non-NULL has a reader thread, or has
* had.*/
HANDLE thread_handle;
#endif
channel->buffer = NULL;
channel->running = FALSE;
+ channel->needs_close = FALSE;
channel->thread_id = 0;
channel->data_avail_event = NULL;
channel->space_avail_event = NULL;
if (channel->debug)
g_print ("thread %#x: got EOF, rdp=%d, wrp=%d, setting data available\n",
channel->thread_id, channel->rdp, channel->wrp);
+
+ if (channel->needs_close)
+ {
+ if (channel->debug)
+ g_print ("thread %#x: channel fd %d needs closing\n",
+ channel->thread_id, channel->fd);
+ if (channel->type == G_IO_FILE_DESC)
+ close (channel->fd);
+ else if (channel->type == G_IO_STREAM_SOCKET)
+ closesocket (channel->fd);
+ channel->fd = -1;
+ }
+
SetEvent (channel->data_avail_event);
UNLOCK (channel->mutex);
g_io_channel_unref((GIOChannel *) channel);
-#if 0
- /* All of the Microsoft docs say we should explicitly
- * end the thread...
+ /* No need to call _endthreadex(), the actual thread starter routine
+ * in MSVCRT (see crt/src/threadex.c:_threadstartex) calls
+ * _endthreadex() for us.
*/
- _endthreadex(1);
-#endif
CloseHandle (channel->thread_handle);
{
channel->reader = reader;
-#if 0
if ((channel->thread_handle =
- _beginthreadex (NULL, 0, reader_thread, channel, 0,
- &channel->thread_id)) == 0)
+ (HANDLE) _beginthreadex (NULL, 0, reader_thread, channel, 0,
+ &channel->thread_id)) == 0)
g_warning ("Error creating reader thread: %s", strerror (errno));
-#else
- if ((channel->thread_handle =
- CreateThread (NULL, 0, reader_thread, channel, 0,
- &channel->thread_id)) == 0)
- {
- gchar *msg = g_win32_error_message (GetLastError ());
- g_warning ("Error creating reader thread: %s", msg);
- g_free (msg);
- }
-#endif
WaitForSingleObject (channel->space_avail_event, INFINITE);
}
if (win32_channel->running)
{
if (win32_channel->debug)
- g_print ("thread %#x: running, terminating it\n",
- win32_channel->thread_id);
- TerminateThread (win32_channel->thread_handle, 0);
- if (win32_channel->debug)
- g_print ("thread %#x: terminated, setting id to 0, closing handle\n",
- win32_channel->thread_id);
- win32_channel->thread_id = 0;
- CloseHandle (win32_channel->thread_handle);
+ g_print ("thread %#x: running, marking fd %d for later close\n",
+ win32_channel->thread_id, win32_channel->fd);
win32_channel->running = FALSE;
+ win32_channel->needs_close = TRUE;
SetEvent (win32_channel->data_avail_event);
}
+ else
+ {
+ if (win32_channel->debug)
+ g_print ("closing fd %d\n", win32_channel->fd);
+ close (win32_channel->fd);
+ if (win32_channel->debug)
+ g_print ("closed fd %d, setting to -1\n",
+ win32_channel->fd);
+ win32_channel->fd = -1;
+ }
UNLOCK (win32_channel->mutex);
- if (win32_channel->debug)
- g_print ("closing fd %d\n", win32_channel->fd);
- close (win32_channel->fd);
- if (win32_channel->debug)
- g_print ("closed fd %d, setting to -1\n",
- win32_channel->fd);
- win32_channel->fd = -1;
}
static int
g_io_channel_init (channel);
g_io_channel_win32_init (win32_channel);
+ if (win32_channel->debug)
+ g_print ("g_io_channel_win32_new_messages: hwnd = %ud\n", hwnd);
channel->funcs = &win32_channel_msg_funcs;
win32_channel->type = G_IO_WINDOWS_MESSAGES;
win32_channel->hwnd = (HWND) hwnd;
g_io_channel_init (channel);
g_io_channel_win32_init (win32_channel);
+ if (win32_channel->debug)
+ g_print ("g_io_channel_win32_new_fd: fd = %d\n", fd);
channel->funcs = &win32_channel_fd_funcs;
win32_channel->type = G_IO_FILE_DESC;
win32_channel->fd = fd;
g_io_channel_init (channel);
g_io_channel_win32_init (win32_channel);
+ if (win32_channel->debug)
+ g_print ("g_io_channel_win32_new_stream_socket: socket = %d\n", socket);
channel->funcs = &win32_channel_sock_funcs;
win32_channel->type = G_IO_STREAM_SOCKET;
win32_channel->fd = socket;
else if (win32_channel->type == G_IO_STREAM_SOCKET)
create_reader_thread (win32_channel, sock_reader);
}
-
-/* This variable and the functions below are present just to be
- * binary compatible with old clients... But note that in GIMP, the
- * libgimp/gimp.c:gimp_extension_process() function will have to be modified
- * anyhow for this new approach.
- *
- * These will be removed after some weeks.
- */
-guint g_pipe_readable_msg = 0;
-
-GIOChannel *
-g_io_channel_win32_new_pipe (int fd)
-{
- return g_io_channel_win32_new_fd (fd);
-}
-
-GIOChannel *
-g_io_channel_win32_new_pipe_with_wakeups (int fd,
- guint peer,
- int peer_fd)
-{
- return g_io_channel_win32_new_fd (fd);
-}
-
-void
-g_io_channel_win32_pipe_request_wakeups (GIOChannel *channel,
- guint peer,
- int peer_fd)
-{
- /* Nothing needed now */
-}
-
-void
-g_io_channel_win32_pipe_readable (gint fd,
- guint offset)
-{
- /* Nothing needed now */
-}
-
g_io_channel_win32_make_pollfd
g_io_channel_win32_new_fd
g_io_channel_win32_new_messages
- g_io_channel_win32_new_pipe
- g_io_channel_win32_new_pipe_with_wakeups
g_io_channel_win32_new_stream_socket
- g_io_channel_win32_pipe_readable
- g_io_channel_win32_pipe_request_wakeups
g_io_channel_win32_poll
g_io_channel_win32_set_debug
g_io_channel_write
g_path_get_dirname
g_path_is_absolute
g_path_skip_root
- g_pipe_readable_msg
g_print
g_printerr
g_printf_string_upper_bound
gboolean running; /* Is reader thread running. FALSE if
* EOF has been reached.
*/
+ gboolean needs_close; /* If the channel has been closed while
+ * the reader thread was still running.
+ */
guint thread_id; /* If non-NULL has a reader thread, or has
* had.*/
HANDLE thread_handle;
#endif
channel->buffer = NULL;
channel->running = FALSE;
+ channel->needs_close = FALSE;
channel->thread_id = 0;
channel->data_avail_event = NULL;
channel->space_avail_event = NULL;
if (channel->debug)
g_print ("thread %#x: got EOF, rdp=%d, wrp=%d, setting data available\n",
channel->thread_id, channel->rdp, channel->wrp);
+
+ if (channel->needs_close)
+ {
+ if (channel->debug)
+ g_print ("thread %#x: channel fd %d needs closing\n",
+ channel->thread_id, channel->fd);
+ if (channel->type == G_IO_FILE_DESC)
+ close (channel->fd);
+ else if (channel->type == G_IO_STREAM_SOCKET)
+ closesocket (channel->fd);
+ channel->fd = -1;
+ }
+
SetEvent (channel->data_avail_event);
UNLOCK (channel->mutex);
g_io_channel_unref((GIOChannel *) channel);
-#if 0
- /* All of the Microsoft docs say we should explicitly
- * end the thread...
+ /* No need to call _endthreadex(), the actual thread starter routine
+ * in MSVCRT (see crt/src/threadex.c:_threadstartex) calls
+ * _endthreadex() for us.
*/
- _endthreadex(1);
-#endif
CloseHandle (channel->thread_handle);
{
channel->reader = reader;
-#if 0
if ((channel->thread_handle =
- _beginthreadex (NULL, 0, reader_thread, channel, 0,
- &channel->thread_id)) == 0)
+ (HANDLE) _beginthreadex (NULL, 0, reader_thread, channel, 0,
+ &channel->thread_id)) == 0)
g_warning ("Error creating reader thread: %s", strerror (errno));
-#else
- if ((channel->thread_handle =
- CreateThread (NULL, 0, reader_thread, channel, 0,
- &channel->thread_id)) == 0)
- {
- gchar *msg = g_win32_error_message (GetLastError ());
- g_warning ("Error creating reader thread: %s", msg);
- g_free (msg);
- }
-#endif
WaitForSingleObject (channel->space_avail_event, INFINITE);
}
if (win32_channel->running)
{
if (win32_channel->debug)
- g_print ("thread %#x: running, terminating it\n",
- win32_channel->thread_id);
- TerminateThread (win32_channel->thread_handle, 0);
- if (win32_channel->debug)
- g_print ("thread %#x: terminated, setting id to 0, closing handle\n",
- win32_channel->thread_id);
- win32_channel->thread_id = 0;
- CloseHandle (win32_channel->thread_handle);
+ g_print ("thread %#x: running, marking fd %d for later close\n",
+ win32_channel->thread_id, win32_channel->fd);
win32_channel->running = FALSE;
+ win32_channel->needs_close = TRUE;
SetEvent (win32_channel->data_avail_event);
}
+ else
+ {
+ if (win32_channel->debug)
+ g_print ("closing fd %d\n", win32_channel->fd);
+ close (win32_channel->fd);
+ if (win32_channel->debug)
+ g_print ("closed fd %d, setting to -1\n",
+ win32_channel->fd);
+ win32_channel->fd = -1;
+ }
UNLOCK (win32_channel->mutex);
- if (win32_channel->debug)
- g_print ("closing fd %d\n", win32_channel->fd);
- close (win32_channel->fd);
- if (win32_channel->debug)
- g_print ("closed fd %d, setting to -1\n",
- win32_channel->fd);
- win32_channel->fd = -1;
}
static int
g_io_channel_init (channel);
g_io_channel_win32_init (win32_channel);
+ if (win32_channel->debug)
+ g_print ("g_io_channel_win32_new_messages: hwnd = %ud\n", hwnd);
channel->funcs = &win32_channel_msg_funcs;
win32_channel->type = G_IO_WINDOWS_MESSAGES;
win32_channel->hwnd = (HWND) hwnd;
g_io_channel_init (channel);
g_io_channel_win32_init (win32_channel);
+ if (win32_channel->debug)
+ g_print ("g_io_channel_win32_new_fd: fd = %d\n", fd);
channel->funcs = &win32_channel_fd_funcs;
win32_channel->type = G_IO_FILE_DESC;
win32_channel->fd = fd;
g_io_channel_init (channel);
g_io_channel_win32_init (win32_channel);
+ if (win32_channel->debug)
+ g_print ("g_io_channel_win32_new_stream_socket: socket = %d\n", socket);
channel->funcs = &win32_channel_sock_funcs;
win32_channel->type = G_IO_STREAM_SOCKET;
win32_channel->fd = socket;
else if (win32_channel->type == G_IO_STREAM_SOCKET)
create_reader_thread (win32_channel, sock_reader);
}
-
-/* This variable and the functions below are present just to be
- * binary compatible with old clients... But note that in GIMP, the
- * libgimp/gimp.c:gimp_extension_process() function will have to be modified
- * anyhow for this new approach.
- *
- * These will be removed after some weeks.
- */
-guint g_pipe_readable_msg = 0;
-
-GIOChannel *
-g_io_channel_win32_new_pipe (int fd)
-{
- return g_io_channel_win32_new_fd (fd);
-}
-
-GIOChannel *
-g_io_channel_win32_new_pipe_with_wakeups (int fd,
- guint peer,
- int peer_fd)
-{
- return g_io_channel_win32_new_fd (fd);
-}
-
-void
-g_io_channel_win32_pipe_request_wakeups (GIOChannel *channel,
- guint peer,
- int peer_fd)
-{
- /* Nothing needed now */
-}
-
-void
-g_io_channel_win32_pipe_readable (gint fd,
- guint offset)
-{
- /* Nothing needed now */
-}
-
g_io_channel_win32_make_pollfd
g_io_channel_win32_new_fd
g_io_channel_win32_new_messages
- g_io_channel_win32_new_pipe
- g_io_channel_win32_new_pipe_with_wakeups
g_io_channel_win32_new_stream_socket
- g_io_channel_win32_pipe_readable
- g_io_channel_win32_pipe_request_wakeups
g_io_channel_win32_poll
g_io_channel_win32_set_debug
g_io_channel_write
g_path_get_dirname
g_path_is_absolute
g_path_skip_root
- g_pipe_readable_msg
g_print
g_printerr
g_printf_string_upper_bound
projects / platform / upstream / glib.git / commitdiff