Sending a buddy request in ICQ fails when using libpurple

[Mark Doliner]

On Wed, Aug 5, 2009 at 6:09 PM,
michaelcbrook at msn.com<michaelcbrook at msn.com> wrote:
> On Wed, 2009-08-05 at 17:56 -0700, Mark Doliner wrote:
>> On Wed, Aug 5, 2009 at 5:29 PM,
>> michaelcbrook at msn.com<michaelcbrook at msn.com> wrote:
>> > On Wed, 2009-08-05 at 10:52 -0700, Mark Doliner wrote:
>> >> On Wed, Aug 5, 2009 at 6:22 AM,
>> >> michaelcbrook at msn.com<michaelcbrook at msn.com> wrote:
>> >> > Ok, since there's been no replies, I'll elaborate a little...
>> >> >
>> >> > I made a program using libpurple, and one of its features is the ability
>> >> > to send buddy requests to certain users in a database.  It runs strictly
>> >> > on the command line.  (And no..it is not a spam bot or any stupid thing
>> >> > like that).
>> >> >
>> >> > Now for the problem: Whenever I send a buddy request from my program to
>> >> > another ICQ account, the buddy request is never received on the other
>> >> > side.  In my program, I send the buddy request by calling
>> >> > purple_account_add_buddy().  Also as another note, when I close my
>> >> > program (signing the ICQ account off) and run it again, the buddy I just
>> >> > tried to add disappears, so it's not staying on the server-side buddy
>> >> > list.
>> >> >
>> >> > Since pidgin can add ICQ buddies successfully but my program can't (for
>> >> > some reason??), I tested out the difference between the debug logs when
>> >> > adding a buddy in pidgin and adding a buddy in my program (the buddy is
>> >> > logged in on a standard Windows ICQ client):
>> >> >
>> >> > *Adding buddy in libpurple with purple_account_add_buddy():
>> >> >
>> >> > (05:54:07) oscar: ssi: adding buddy 401787217 to group Buddies
>> >> > (05:54:07) oscar: Requesting ICQ alias for 401787217
>> >> > (05:54:07) oscar: rate change (param ID 0x0001): curavg = 4668, maxavg =
>> >> > 6000, alert at 2000, clear warning at 2500, limit at 1500, disconnect at
>> >> > 800 (window size = 80)
>> >> > (05:54:07) oscar: ssi: status is 0x000e for a 0x0008 action with name no
>> >> > item
>> >> > (05:54:07) oscar: icq response: 60 bytes, 579746952, 0x07da, 0x0028
>> >> > (05:54:07) oscar: ssi: status is 0x0000 for a 0x0009 action with name
>> >> > Buddies
>> >> > (05:54:09) util: Writing file accounts.xml to directory /dev/null
>> >> > (05:54:09) util: Error creating directory /dev/null: File exists
>> >> > (05:54:12) util: Writing file blist.xml to directory /dev/null
>> >> > (05:54:12) util: Error creating directory /dev/null: File exists
>> >> >
>> >> > The buddy request was never received on the other end...
>> >> >
>> >> > I know the natural assumption is "your program is broken" but I can
>> >> > confirm that it works with 3 out of 4 protocols...It works with AIM,
>> >> > Yahoo, and MSN, but fails with ICQ, and it still fails even in the
>> >> > simplest program.  I am really stumped and would greatly appreciate any
>> >> > help...this problem is delaying the launch of a website I'm creating.
>> >>
>> >> In Pidgin isn't there a dialog that asks the user if they want to
>> >> request authorization from the buddy, and allows the user to enter a
>> >> little message?  Does your UI handle that?  I think you would need to
>> >> implement the request_input callback in the request API.
>> >>
>> >> -Mark
>> >>
>> > Is it possible to use purple_request_input without a GUI?  If it is, I
>> > don't quite know how the callback functions would be called...
>>
>> Yes it's possible.  Your uiop function can call the appropriate
>> callback to respond to the request.
>>
>> -Mark
>>
> Ahh I see now.  I tried calling ok_cb() when the request_input uiop was
> called, but I'm getting a segmentation fault now.  Here's my code:
>
> static void
> *request_input(const char *title, const char *primary, const char
> *secondary, const char *default_value, gboolean multiline, gboolean
> masked, gchar *hint, const char *ok_text, GCallback ok_cb, const char
> *cancel_text, GCallback cancel_cb, PurpleAccount *account, const char
> *who, PurpleConversation *conv, void *user_data)
> {
> ok_cb();
> }
>
> static PurpleRequestUiOps request_uiops =
> {
>        request_input, /* request_input  */
>        NULL,          /* request_choice */
>        NULL,          /* request_action */
>        NULL,          /* request_fields */
>        NULL,          /* request_file   */
>        NULL,          /* close_request  */
>        NULL,          /* request_folder */
>        NULL,
>        NULL,
>        NULL,
>        NULL
> };

I suspect you'll probably need to set a timer to call the ok_cb() later:
    purple_timeout_add(0, request_input_cb, ui_data);

A lot of our code doesn't expect callback functions like this to be
called immediately after the request is made.  Patches to remove this
requirement are welcome.

And if this turns out not to be the problem you'll have to send us a
backtrace, as I don't think we'll be able to infer why it's crashing
without it.

-Mark