Support basic Busy Lamp Field (BLF) function as server (#4474) by alfedotov · Pull Request #4891 · pjsip/pjproject

alfedotov · 2026-03-26T13:29:40Z

No description provided.

CLAassistant · 2026-03-26T13:29:55Z

All committers have signed the CLA.

Tested with Yealink phones (models T77U, T34W, T31P, T30P)

sauwming · 2026-03-30T05:58:12Z

This doesn't seem like a complete server implementation. So perhaps you can describe the objective and scope of this PR in the PR description.

alfedotov · 2026-04-02T08:58:55Z

Correct, this is not complete server implementation with subscribe to dialog states.

I've found that modern yealink phones has no Presence option in setup for programmable buttons, but only BLF. My PR simply adds minimal support for BLF needed for correct busy/available indication.

Copilot

Pull request overview

Adds initial support for serving Busy Lamp Field (BLF)-style subscriptions by allowing dialog event subscriptions and generating application/dialog-info+xml bodies derived from existing presence status.

Changes:

Extend presence UAS handling to recognize Event: dialog and negotiate application/dialog-info+xml.
Add a new public utility API to generate dialog-info bodies from pjsip_pres_status.
Adjust dialog-info XML construction logic to support the new output shape.

Reviewed changes

Copilot reviewed 4 out of 4 changed files in this pull request and generated 9 comments.

File	Description
`pjsip/src/pjsip-simple/presence.c`	Accepts `Event: dialog`, negotiates dialog-info content type, and routes NOTIFY body creation accordingly.
`pjsip/src/pjsip-simple/presence_body.c`	Implements `pjsip_pres_create_dialog_info()` to produce a dialog-info XML body from presence status.
`pjsip/src/pjsip-simple/dialog_info.c`	Modifies dialog-info XML node construction and where dialog state is represented.
`pjsip/include/pjsip-simple/presence.h`	Exposes the new dialog-info body creation helper as a public API.

Copilot · 2026-04-06T03:25:55Z

    /* Check that Event header contains "presence" */
    event = (pjsip_event_hdr*)
            pjsip_msg_find_hdr_by_name(rdata->msg_info.msg, &STR_EVENT, NULL);
    if (!event) {
        return PJSIP_ERRNO_FROM_SIP_STATUS(PJSIP_SC_BAD_REQUEST);
    }
    if (pj_stricmp(&event->event_type, &STR_PRESENCE) != 0) {
+    }
+    else if (pj_stricmp(&event->event_type, &STR_DIALOG) != 0) {
+    }
+    else {
        return PJSIP_ERRNO_FROM_SIP_STATUS(PJSIP_SC_BAD_EVENT);
    }


The Event header validation logic is broken: the current if/else-if/else blocks are empty, so unsupported Event types will never trigger PJSIP_SC_BAD_EVENT. This should reject when event_type is neither "presence" nor "dialog" (e.g., combine conditions with && and return BAD_EVENT in that case).

Copilot · 2026-04-06T03:25:56Z

+
+    pjrpid_element rpid = status->info[0].rpid;
+    pj_bool_t basic_open = status->info[0].basic_open;
+
+    const pj_str_t* state;


This function reads status->info[0] unconditionally, which will access out of bounds when status->info_cnt==0 (e.g., when no tuple/person info is present). Please guard on info_cnt first and choose a safe default state when no info is available.

Suggested change

pjrpid_element rpid = status->info[0].rpid;

pj_bool_t basic_open = status->info[0].basic_open;

const pj_str_t* state;

pjrpid_element rpid;

pj_bool_t basic_open;

const pj_str_t* state;

pj_bzero(&rpid, sizeof(rpid));

basic_open = PJ_FALSE;

if (status->info_cnt > 0) {

rpid = status->info[0].rpid;

basic_open = status->info[0].basic_open;

}

Copilot · 2026-04-06T03:25:56Z

+static pj_str_t STR_CONFIRMED = { "confirmed", 9 };
+static pj_str_t STR_TERMINATED = { "terminated", 10 };
+static pj_str_t STR_VERSION = { "1.0", 3 };
+static pj_str_t STR_EMPTY_STRING = { NULL, 0 };


Several of these new static pj_str_t values are mutable even though they are constants, and STR_EMPTY_STRING is currently unused (will trigger -Wunused-variable under -Wall). Consider making the constants static const pj_str_t and removing STR_EMPTY_STRING (and any other unused globals) to keep builds warning-free.

Suggested change

static pj_str_t STR_CONFIRMED = { "confirmed", 9 };

static pj_str_t STR_TERMINATED = { "terminated", 10 };

static pj_str_t STR_VERSION = { "1.0", 3 };

static pj_str_t STR_EMPTY_STRING = { NULL, 0 };

static const pj_str_t STR_CONFIRMED = { "confirmed", 9 };

static const pj_str_t STR_TERMINATED = { "terminated", 10 };

static const pj_str_t STR_VERSION = { "1.0", 3 };

Copilot · 2026-04-06T03:25:56Z

+    if (basic_open) {
+        if (rpid.activity == PJRPID_ACTIVITY_BUSY) {
+            state = &STR_CONFIRMED;
+        }
+        else {
+            state = &STR_TERMINATED;
+        }
+    }
+    else {
+        //pjsip_dlg_info_dialog_set_state(pool, dlginfo, &STR_TERMINATED);
+        state = &STR_TERMINATED;
+    }
+
+    /* Create DIALOG-INFO document. */
+    dlginfo = pjsip_dlg_info_create(pool, &STR_VERSION, state, entity);
+


pjsip_dlg_info_create()/pjsip_dlg_info_dialog_info_construct() expects state to be the dialog-info document state attribute (per existing get_state()/set_state() which look for an attribute), but here state is being passed a call-state value (confirmed/terminated). This mismatch will generate a dialog-info document that doesn't conform to the current API/RFC expectations; consider passing a valid document state (e.g., full/partial) and setting the dialog's separately.

Copilot · 2026-04-06T03:25:56Z

    xml_init_node(pool, dialog_info, &DIALOG_INFO, NULL);
    attr = xml_create_attr(pool, &VERSION, version);
    pj_xml_add_attr(dialog_info, attr);

-    attr = xml_create_attr(pool, &STATE, state);
-    pj_xml_add_attr(dialog_info, attr);
-
    attr = xml_create_attr(pool, &ENTITY, entity);
    pj_xml_add_attr(dialog_info, attr);

    dialog = PJ_POOL_ALLOC_T(pool, pjsip_dlg_info_dialog);
-    pjsip_dlg_info_local_construct(pool, dialog);
+    pjsip_dlg_info_dialog_construct(pool, dialog, NULL);
    pj_xml_add_node(dialog_info, dialog);
+
+    node = PJ_POOL_ALLOC_T(pool, pj_xml_node);
+    xml_init_node(pool, node, &STATE, state);
+    pj_xml_add_node(dialog, node);
+


dialog-info construction no longer adds the dialog-info "state" attribute, but pjsip_dlg_info_dialog_info_get_state()/set_state() still operate on that attribute. As a result, the constructed document will have an empty state via the existing API and may be non-compliant; restore the STATE attribute on (and use the per- element for call state if needed).

Copilot · 2026-04-06T03:25:57Z


    dialog = PJ_POOL_ALLOC_T(pool, pjsip_dlg_info_dialog);
-    pjsip_dlg_info_local_construct(pool, dialog);
+    pjsip_dlg_info_dialog_construct(pool, dialog, NULL);


pjsip_dlg_info_dialog_info_construct() creates a with id=NULL, which means the required "id" attribute is omitted. This can produce invalid dialog-info (RFC 4235 requires dialog id) and also breaks pjsip_dlg_info_dialog_set_id() which asserts the id attr exists. Consider always creating the id attribute (or generate a default id) when constructing a dialog.

Suggested change

pjsip_dlg_info_dialog_construct(pool, dialog, NULL);

pjsip_dlg_info_dialog_construct(pool, dialog, &EMPTY_STRING);

Copilot · 2026-04-06T03:25:57Z

+    if (name) node->name = *name;
+    else node->name.ptr=NULL, node->name.slen=0;


The else node->name.ptr=NULL, node->name.slen=0; uses the comma operator and lacks braces, which is easy to misread and brittle for future edits. Prefer a braced else block with two explicit statements assigning ptr and slen.

Suggested change

if (name) node->name = *name;

else node->name.ptr=NULL, node->name.slen=0;

if (name) {

node->name = *name;

} else {

node->name.ptr = NULL;

node->name.slen = 0;

}

Copilot · 2026-04-06T03:25:57Z

+PJ_DECL(pj_status_t) pjsip_pres_create_dialog_info(pj_pool_t* pool,
+                                                   const pjsip_pres_status* status,
+                                                   const pj_str_t* entity,
+                                                   pjsip_msg_body** p_body);


The new public API declaration uses a different pointer formatting style (pj_pool_t*, pjsip_pres_status*, etc.) than the rest of this header (e.g., pjsip_pres_create_pidf() uses pj_pool_t *pool). For consistency with existing public headers, please match the surrounding spacing convention.

Suggested change

PJ_DECL(pj_status_t) pjsip_pres_create_dialog_info(pj_pool_t* pool,

const pjsip_pres_status* status,

const pj_str_t* entity,

pjsip_msg_body** p_body);

PJ_DECL(pj_status_t) pjsip_pres_create_dialog_info(pj_pool_t *pool,

const pjsip_pres_status *status,

const pj_str_t *entity,

pjsip_msg_body **p_body);

Copilot · 2026-04-06T03:25:57Z

    /* Check that request contains compatible Accept header. */
    accept = (pjsip_accept_hdr*)
             pjsip_msg_find_hdr(rdata->msg_info.msg, PJSIP_H_ACCEPT, NULL);
    if (accept) {
        unsigned i;
        for (i=0; i<accept->count; ++i) {
            if (pj_stricmp(&accept->values[i], &STR_APP_PIDF_XML)==0) {
                content_type = CONTENT_TYPE_PIDF;
                break;
            } else
            if (pj_stricmp(&accept->values[i], &STR_APP_XPIDF_XML)==0) {
                content_type = CONTENT_TYPE_XPIDF;
                break;
+            } else
+            if (pj_stricmp(&accept->values[i], &STR_APP_DIALOG_INFO_XML) == 0) {
+                content_type = CONTENT_TYPE_DIALOG_INFO;
+                break;
            }
        }


Accept header negotiation now allows application/dialog-info+xml unconditionally. This means a normal "presence" subscription with Accept: application/dialog-info+xml would be treated as acceptable and later generate a dialog-info body, which doesn't match the presence event package. Consider only accepting dialog-info when Event is "dialog" (and keep presence limited to PIDF/XPIDF).

sauwming · 2026-04-06T03:30:18Z

Thank you for the contribution, but at its current state, I am leaning towards rejecting this since the server support seems to be still basic, and I'm not sure how it can be useful for PJSIP users.

alfedotov changed the title ~~Pr 4474 dialog info blf~~ Support basic Busy Lamp Field (BLF) function as server (#4474) Mar 26, 2026

alfedotov added 2 commits March 26, 2026 16:37

Construct valid dialog-info XML

618629f

Support basic Busy Lamp Field (BLF) function as server (pjsip#4474)

08bdaf8

Tested with Yealink phones (models T77U, T34W, T31P, T30P)

alfedotov force-pushed the pr-4474-dialog-info-blf branch from 7b055fd to 08bdaf8 Compare March 26, 2026 13:37

sauwming requested a review from Copilot April 6, 2026 03:19

Copilot started reviewing on behalf of sauwming April 6, 2026 03:20 View session

Copilot AI reviewed Apr 6, 2026

View reviewed changes

	pjsip_dlg_info_dialog_construct(pool, dialog, NULL);
	pjsip_dlg_info_dialog_construct(pool, dialog, &EMPTY_STRING);

		if (name) node->name = *name;
		else node->name.ptr=NULL, node->name.slen=0;

-    if (name) node->name = *name;
-    else node->name.ptr=NULL, node->name.slen=0;
+    if (name) {
+        node->name = *name;
+    } else {
+        node->name.ptr = NULL;
+        node->name.slen = 0;
+    }

Conversation

alfedotov commented Mar 26, 2026

Uh oh!

CLAassistant commented Mar 26, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

sauwming commented Mar 30, 2026

Uh oh!

alfedotov commented Apr 2, 2026

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull request overview

Reviewed changes

Uh oh!

Copilot AI Apr 6, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Apr 6, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Apr 6, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Apr 6, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Apr 6, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Apr 6, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Apr 6, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Apr 6, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Apr 6, 2026

Choose a reason for hiding this comment

Uh oh!

sauwming commented Apr 6, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

4 participants

CLAassistant commented Mar 26, 2026 •

edited

Loading