Re: [Libmtp-discuss] SanDisk Sansa View connection problem

[Linus W. <lin...@gm...>]

2008/2/12, Sciamano <sci...@iw...>:

> Got it. I wonder how Amarok2, Rhythmbox and Gnomad2 have solved this issue.
> Don't they depend on libmtp too?

Yeah, we put in other dependencies to auto-detect device plugin.
Rhythmbox and Gnomad2 do something like this:

We include D-Bus and HAL libraries, define some useful
global variables:

#include <dbus/dbus-glib.h>
#include <dbus/dbus.h>
#include <dbus/dbus-glib-lowlevel.h>
#include <hal/libhal.h>
DBusConnection *dbus_system_conn;
DBusError dbus_error;
LibHalContext *hal_context;
char *global_udistr = NULL;

In some main() function we connect to the D-Bus system bus,
then we connect the HAL handlers (callbacks) to that system bus:

 // Connect to the SYSTEM bus to detect devices and stuff.
  dbus_error_init(&dbus_error);
  dbus_system_conn = dbus_bus_get(DBUS_BUS_SYSTEM, &dbus_error);
  if (dbus_system_conn == NULL) {
    g_printerr ("Failed to open connection to D-BUS SYSTEM bus\n");
  } else {
    dbus_connection_setup_with_g_main(dbus_system_conn, NULL);
    // dbus_system_proxy = dbus_g_proxy_new_for_name (dbus_system_conn,
    //                                     "org.freedesktop.Hal",
    //                                     "/",
    //                                     "org.freedesktop.Hal.Device");
    // g_object_unref (dbus_system_proxy);
    hal_context = libhal_ctx_new ();
    if (hal_context == NULL) {
      g_print("Couldn't create HAL context!\n");
    } else {
      //libhal_ctx_set_user_data (hal_context, (void *) NULL);
      libhal_ctx_set_dbus_connection (hal_context, dbus_system_conn);
      libhal_ctx_set_device_added (hal_context, hal_device_added);
      libhal_ctx_set_device_removed (hal_context, hal_device_removed);
      libhal_device_property_watch_all (hal_context, &dbus_error);
      if (!libhal_ctx_init (hal_context, &dbus_error)) {
        g_print("error: libhal_ctx_init: %s: %s\n", dbus_error.name,
dbus_error.message);
      }
    }
  }
  dbus_error_free(&dbus_error);

Then we have to have the functions that HAL shall call whenever
a device is added/removed. In gnomad2 they look like this, and
use the MTP libmtp.fdi HAL definitions to identify MTP devices.

static void
hal_device_added (LibHalContext *context, const char *udi)
{
  int vendor_id;
  int product_id;
  char *protocol;

  vendor_id = libhal_device_get_property_int (context, udi,
"usb.vendor_id", NULL);
  product_id = libhal_device_get_property_int (context, udi,
"usb.product_id", NULL);
  protocol = libhal_device_get_property_string (context, udi,
"portable_audio_player.type", NULL);

  if (protocol != NULL) {
    g_print("Detected connected device VID=0x%04x/PID=0x%04x,
protocol: %s\n", vendor_id, product_id, protocol);
    g_print("UDI: %s\n", udi);
    if (!strcmp(protocol, "mtp")) {
      // LIBMTP_Get_First_Device(void)  ETC.
    }
  }

  libhal_free_string(protocol);
}

static void
hal_device_removed (LibHalContext *context, const char *udi)
{
  int vendor_id;
  int product_id;

  vendor_id = libhal_device_get_property_int (context, udi,
"usb.vendor_id", NULL);
  product_id = libhal_device_get_property_int (context, udi,
"usb.product_id", NULL);

  if (global_udistr != NULL && !strcmp(udi, global_udistr)) {
    g_print("Device VID=0x%04x/PID=0x%04x detached.\n", vendor_id, product_id);
    g_print("UDI: %s\n", udi);
    // LIBMTP_Release_Device() ETC.
    free(global_udistr);
    global_udistr = NULL;
  }
}

The result is that libmtp is called to identify a device whenever
a device listed in out HAL .fdi file is plugged in. Cute eh?

This is all from glib C code, there are other (similar) constructs
in other languages, not that I know how these work tho.

Linus