Re: [PATCH obexd v2 1/3] Infrastructure for MAP function selection

On Wed, Jul 20, 2011 at 2:23 PM, Luiz Augusto von Dentz
[off-list ref] wrote:

Hi,

On Wed, Jul 20, 2011 at 2:14 PM, Slawomir Bochenski [off-list ref] wrote:

quoted

On Wed, Jul 20, 2011 at 12:47 PM, Luiz Augusto von Dentz
[off-list ref] wrote:

quoted

Hi,

On Wed, Jul 20, 2011 at 1:08 PM, Slawomir Bochenski [off-list ref] wrote:

quoted

Hi,

On Wed, Jul 20, 2011 at 11:50 AM, Luiz Augusto von Dentz
[off-list ref] wrote:

quoted

Hi,

On Tue, Jul 19, 2011 at 3:06 PM, Slawomir Bochenski [off-list ref] wrote:

quoted

---
 plugins/mas.c |   78 ++++++++++++++++++++++++++++++++++++++++++++++----------
 1 files changed, 64 insertions(+), 14 deletions(-)

diff --git a/plugins/mas.c b/plugins/mas.c
index 0ef8c81..e88a0f0 100644
--- a/plugins/mas.c
+++ b/plugins/mas.c

@@ -27,6 +27,9 @@

 #include <errno.h>
 #include <glib.h>
+#include <sys/types.h>
+#include <sys/stat.h>
+#include <fcntl.h>
 #include <openobex/obex.h>

 #include "plugin.h"

@@ -86,9 +89,19 @@

  </attribute>                                                         \
 </record>"

+#define EVENT_TYPE             "x-bt/MAP-event-report"
+#define MESSAGE_TYPE           "x-bt/message"
+#define FOLDER_LISTING_TYPE    "x-obex/folder-listing"
+#define MESSAGES_LISTING_TYPE  "x-bt/MAP-msg-listing"
+#define NOTIFICATION_TYPE      "x-bt/MAP-NotificationRegistration"
+#define STATUS_TYPE            "x-bt/messageStatus"
+#define UPDATE_TYPE            "x-bt/MAP-messageUpdate"
+
 struct mas_session {
       struct mas_request *request;
       void *backend_data;
+       const char *name;
+       const char *type;
 };

 static const uint8_t MAS_TARGET[TARGET_SIZE] = {

@@ -137,17 +150,14 @@ static void mas_disconnect(struct obex_session *os, void *user_data)

 static int mas_get(struct obex_session *os, obex_object_t *obj, void *user_data)
 {
       struct mas_session *mas = user_data;
-       const char *type = obex_get_type(os);
-       const char *name = obex_get_name(os);
       int ret;

-       DBG("GET: name %s type %s mas %p",
-                       name, type, mas);
+       mas->name = obex_get_name(os);
+       mas->type = obex_get_type(os);

-       if (type == NULL)
-               return -EBADR;
+       DBG("GET: name %s type %s mas %p", mas->name, mas->type, mas);

-       ret = obex_get_stream_start(os, name);
+       ret = obex_get_stream_start(os, mas->name);
       if (ret < 0)
               goto failed;

@@ -160,16 +170,14 @@ failed:

 static int mas_put(struct obex_session *os, obex_object_t *obj, void *user_data)
 {
       struct mas_session *mas = user_data;
-       const char *type = obex_get_type(os);
-       const char *name = obex_get_name(os);
       int ret;

-       DBG("PUT: name %s type %s mas %p", name, type, mas);
+       mas->name = obex_get_name(os);
+       mas->type = obex_get_type(os);

-       if (type == NULL)
-               return -EBADR;
+       DBG("PUT: name %s type %s mas %p", mas->name, mas->type, mas);

-       ret = obex_put_stream_start(os, name);
+       ret = obex_put_stream_start(os, mas->name);
       if (ret < 0)
               goto failed;

@@ -179,6 +187,38 @@ failed:

       return ret;
 }

I would suggest adding a mimetype driver for each type so you don't
have to do the check inside the service driver and it is also
consistent with what other plugins do when they support multiple
types.

I would not. That's the whole point. Functions for the mime driver
would do exactly the same.

There we go again, so your point is that you don't want to be
consistent with other plugins and you like to see duplicated code
checking for matching types the very same way the core does for
mimetype drivers? Or do you mean the spec got this wrong and there is
no use of types because all of them do exactly the same?

This was discussed before. More than half of "types" are not really
mime types at all (there's no even body involved in request). Type is
used in MAP merely as a function selector (this _is_ called a function
in MAP). All specific handling of request is done in callback from MAP
API. The only thing that each and every mime driver would do is to
read buffer prepared by a callback. I've specifically chosen to use
the already present feature of universal target-specific mime driver
in order to avoid registering 6 mime drivers with the same handlers,
or worse - 6 times identical functions. The approach is minimalistic
and pretty. Implementation is easy to understand in terms of MAP
specification instead of putting it into some obexd specific
abstraction.

What does a function selection mean? Doesn't it means they do
something different? Why you are saying you will have 6 driver with
the same handlers if they map to different functions? I would
understand if you said you have done this way and it works just fine,

Yes, I've done it. Works just fine.

but arguing that your approach is minimalistic and pretty? What about
consistency, reliability and maintainability? The abstraction is there
no matter how hard you try to make it go away, it is the interface to
which the core daemon talks to backends, it doesn't really matter what
you put in between. Your minimalistic approach just confuses the core
daemon making it believe there is no driver for a specific mimetype
selecting its default fallback for the target.

Nothing confusing here. Just the normal driver selection.

quoted

quoted

Sorry but I
don't think this has any valid technical point, in fact the mimetype
and service separation does help plugins to not mix object and session
handling.

There can be only one active request per session. No mixing problems there.

Partially correct, basically you can only handle one request per
session, period. Note, it doesn't break OBEX or MAP, but if we already
have infrastructure in place to handle this properly why not do it?
Why not doing it right since the beginning? Shall our answer be
minimalism and prettiness?

"Right" is quite a subjective term here.

quoted

quoted

quoted

quoted

quoted

+static int start_get(struct mas_session *mas)
+{
+       /* NOTE: type is case-insensitive! */
+       if (g_ascii_strcasecmp(mas->type, FOLDER_LISTING_TYPE) == 0)
+               return -EINVAL;
+       else if (g_ascii_strcasecmp(mas->type, MESSAGES_LISTING_TYPE) == 0)
+               return -EINVAL;
+       else if (g_ascii_strcasecmp(mas->type, MESSAGE_TYPE) == 0)
+               return -EINVAL;
+       else {
+               DBG("Incorrect type for get: %s", mas->type);
+               return -EBADR;
+       }
+}
+
+static int start_put(struct mas_session *mas)
+{
+       /* NOTE: type is case-insensitive! */
+       if (g_ascii_strcasecmp(mas->type, NOTIFICATION_TYPE) == 0)
+               return -EINVAL;
+       else if (g_ascii_strcasecmp(mas->type, STATUS_TYPE) == 0)
+               return -EINVAL;
+       else if (g_ascii_strcasecmp(mas->type, MESSAGE_TYPE) == 0)
+               return -EINVAL;
+       else if (g_ascii_strcasecmp(mas->type, UPDATE_TYPE) == 0)
+               return -EINVAL;
+       else {
+               DBG("Incorrect type for put: %s", mas->type);
+               return -EBADR;
+       }
+}
+
 static int mas_setpath(struct obex_session *os, obex_object_t *obj,
               void *user_data)
 {

@@ -210,7 +250,17 @@ static void *any_open(const char *name, int oflag, mode_t mode,

       DBG("");

-       *err = 0;
+       if ((oflag & O_RDONLY) == O_RDONLY) {
+               *err = start_get(mas);
+       } else if ((oflag & O_WRONLY) == O_WRONLY) {
+               *err = start_put(mas);
+       } else {
+               DBG("Invalid open flag!");
+               *err = -EIO;
+       }
+
+       if (*err)
+               return NULL;

       return mas;
 }
--
1.7.4.1

--
Luiz Augusto von Dentz

--
Slawomir Bochenski

--
Luiz Augusto von Dentz

--
Slawomir Bochenski

--
Luiz Augusto von Dentz

-- 
Slawomir Bochenski