TELCORE-176: Sync FreeSWITCH fork with upstream v1.11.1 by damirn · Pull Request #601 · team-telnyx/freeswitch

damirn · 2026-06-01T12:14:43Z

Summary

Merges upstream FreeSWITCH tag v1.11.1 (c2c59645f6) into the Telnyx fork. This carries the full v1.11.0 sync plus v1.11.1.

Review follow-up (post-merge)

Review surfaced one real merge defect and several items that turned out to be non-issues after investigation. Summary below; per-comment detail is in the review threads.

Fixed — duplicate SIP Reason handling in `sofia_handle_sip_r_invite()` (`45bc70443a`)

The merge stacked upstream's new Reason-capture / 603+ detection block on top of the pre-existing Telnyx block without reconciling them, which caused:

sip->sip_reason rendered twice per status >= 400 response;
the first sip_header_as_string() allocation leaked (only the second was su_free'd) — bounded to the NUA handle home, but still wrong (pristine upstream v1.11.1 has the same non-free; the Telnyx fork was the tidy side);
partner-leg sip_reason propagated unconditionally for every failure, silently bypassing the Telnyx PFLAG_PASS_603_NETWORK_BLOCKED gate, which became dead code.

Reconciled into a single block: render once / free once, current-leg sip_reason unchanged, upstream's sip_603plus_reason (ATIS-1000099) detection kept on both legs, q850 / ignore_reason_header_by_sip_code mapping kept.

Partner-propagation policy — Option A (preserved): partner-leg sip_reason is propagated only for 603 "Network Blocked" under PFLAG_PASS_603_NETWORK_BLOCKED, exactly as before the merge. This sync PR intentionally does not adopt upstream's broader "propagate partner sip_reason for all failures" behavior. If we later want that broadening, it should be a separate, explicitly-reviewed change.

Investigated — no change needed

DTLS peer-cert verification refactor (upstream 8babcee3ea, switch_rtp.c:4001/4267): this is a security fix, not a regression. The old path had switch_core_cert_extract_fingerprint() overwrite both remote_fp->str and ->data with the peer cert's own fingerprint before switch_core_cert_verify(), so it compared the cert against itself and always passed. The new code compares the cert fingerprint against the SDP-advertised one. SDP stores the fingerprint verbatim as uppercase colon-hex (switch_core_media.c:5404), extract_fingerprint emits the same form, and the compare is strncasecmp — formats match. Recommend a quick real-browser WebRTC DTLS smoke test before deploy (cheap; non-blocking).
switch_core_media.h "dropped" switch_rtp_set_ice_role() declaration: intentional. TELCORE-138 (TELCORE-138: fix ICE role for non-DTLS inbound calls with ICE #591) refactored that function out of switch_rtp.c into inline rtp_ice_role dialplan-override logic (switch_core_media.c:5030-5048); declaration, definition, and caller were removed together. Not a silent drop.
mod_sofia HOLD-handler polycom→poly resolution (mod_sofia.c:2563-2573): "keep both" is correct — the TELCORE-67 sanitized hold_display and the broadened UA match are orthogonal and compose cleanly.

Verification

mod_sofia builds clean after the Reason reconciliation (incl. test/test_603plus).
TODO before merge: full make all on the standard build image; real-browser DTLS/WebRTC smoke test (see DTLS note above).

Coordination

Land-order with PR TELCORE-170: Add aggressive-nomination fallback for ICE-controlled peers #600 (TELCORE-170 ICE nomination fallback): this PR brings upstream 1585ca7aaf [core] Fix - add missing ice_mutex to protect dtls, which touches the same ICE/DTLS mutex region TELCORE-170: Add aggressive-nomination fallback for ICE-controlled peers #600 modifies. Whichever lands second needs a small conflict resolution — coordinating with @baloeng.

[mod_skypopen] Remove from tree.

[mod_gsmopen] Remove from tree.

[mod_sangoma_codec] Remove from tree

[mod_ladspa] Remove from tree

[mod_cepstral] Remove from tree

[mod_sms_flowroute] Remove from tree

[mod_dahdi_codec] Remove from tree

[mod_portaudio, mod_portaudio_stream] Remove from tree

[mod_mp4, mod_mp4v, mod_mp4v2] Remove from tree

[mod_isac] Remove from tree

[mod_abstraction] Remove from tree

[mod_stress] Remove from tree.

[mod_unicall] Remove from tree

[mod_soundtouch] Remove from tree

[mod_sonar] Remove from tree

baloeng · 2026-06-30T13:09:31Z

Rebase needed: this branch is missing TELCORE-156/TELCORE-160 trickle ICE fixes from base. Preserve them to avoid reconnect/trickle ICE regressions.

Both are on the branch now — cherry-picked from deploy-development rather than rebased:

TELCORE-156 (TELCORE-156: clear stale EOC/PAC state in switch_core_media_clear_ice() #613) 70e79acaad — clears stale EOC/PAC state in switch_core_media_clear_ice().

TELCORE-160 (TELCORE-160: fix b2bua-rtc trickle ICE NULL ufrag/pwd race #598) 5939788afc — trickle-ICE NULL ufrag/pwd race fix (touches both switch_core_media.c and switch_rtp.c).

I cherry-picked the 14 commits added to deploy-development since this sync branch forked instead of merging, because deploy-development sits on the 1.10.12 base and a merge would drag the version/base backward and revert v1.11.1 sync work. No trickle-ICE fixes are dropped.

baloeng · 2026-06-30T13:09:33Z

Rebase needed: this branch is missing TELCORE-160's ICE credential guard. Preserve it to avoid NULL ufrag/pwd STUN regression.

Present now via TELCORE-160 (#598) 5939788afc. The credential guard is in switch_rtp.c: ICE activation is refused when any of the local/remote ufrag or pwd is empty —

(zstr(login) || zstr(rlogin) || zstr(password) || zstr(rpassword)) -> "Refusing to activate %s ICE: missing credentials"

so the NULL/empty ufrag/pwd STUN path is covered. Brought in by cherry-pick, not a rebase, for the base-version reason in the switch_core_media.c thread.

baloeng · 2026-06-30T13:09:34Z

+		dtls_fingerprint_t fp = {0};
+
+		fp.type = dtls->remote_fp->type;
+


Rebase needed + WebRTC test coverage: this now strictly compares SDP fingerprint vs peer cert and may reject peers previously accepted.

Same upstream change dev-ryanc flagged above — full analysis in that thread. Short version:

The pre-1.11 path was effectively a no-op. At verify time switch_core_cert_extract_fingerprint(cert, dtls->remote_fp) overwrote both remote_fp->str and remote_fp->data with the peer cert's own fingerprint, then switch_core_cert_verify() parsed ->str back into bytes and compared against ->data — i.e. it compared the cert against itself, so peer-fingerprint verification always returned true.

The new code (upstream 8babcee3ea, Fix DTLS Peer Certificate verification) compares the extracted cert fingerprint against the SDP-advertised dtls->remote_fp->str, which is the actual check.

On "may reject peers previously accepted": the formats match, so a peer presenting the cert it advertised in SDP still passes. SDP stores the fingerprint verbatim as uppercase colon-hex (switch_core_media.c, switch_set_string, no transform), extract_fingerprint emits the same %.2X: uppercase-colon form, and the compare is strncasecmp (case-insensitive). It tightens a previously-permissive check rather than breaking conformant peers.

Agreed a real-browser WebRTC DTLS smoke test is cheap insurance — it's on the verification TODO in the PR description.

baloeng · 2026-06-30T13:09:35Z

-				switch_stun_ip_t *ip = (switch_stun_ip_t *) attr->value;
+				switch_stun_ip_t *ip;
+
+				if (bytes_left < sizeof(switch_stun_ip_t)) {


Rebase needed: PRIORITY is a 4-byte attr, not an address attr. This guard can reject valid ICE/STUN packets.

Good eye, but this is safe and it's upstream, not a rebase gap:

The PRIORITY-in-address-group is pre-existing. deploy-development already lumps SWITCH_STUN_ATTR_PRIORITY in with the address attrs and does the same switch_stun_ip_t cast. What v1.11.1 added is the bounds check — before, that cast ran with no guard at all and would over-read on a short/truncated attr.

bytes_left is whole-packet-remaining, not attr length. At this point the 4-byte attr header has already been subtracted, so the guard rejects only when fewer than 8 bytes remain in the entire message from attr->value onward (and alen > bytes_left was already validated just above).

You're right that PRIORITY is a 4-byte value, so an 8-byte (sizeof(switch_stun_ip_t)) floor is wider than PRIORITY strictly needs. But for that to reject a packet, PRIORITY would have to be the trailing attribute with <8 bytes after it. In real ICE connectivity checks PRIORITY is never last — USERNAME / ICE-CONTROLLING|CONTROLLED, then MESSAGE-INTEGRITY (24B) and FINGERPRINT (8B) always follow — so bytes_left at PRIORITY is always ≫ 8 and no valid ICE packet is rejected.

Net vs deploy-development: strictly safer (unchecked over-read → bounded parse), and identical to upstream FS 1.11. If we want it pedantically correct we could move PRIORITY into the 4-byte uint32_t case group below, but since FS doesn't consume the PRIORITY value here that's cosmetic — happy to do it as a follow-up if you'd prefer.

…ers (#600)

…ck (#602)

…sion test (#585) * TELCORE-100: add ZMQ_SNDTIMEO/RCVTIMEO to Telnyx::ZMQ::ZMQSocket The wrapper only set ZMQ_LINGER, leaving SNDTIMEO/RCVTIMEO at the libzmq default of -1 (block forever). When a recovery peer became unresponsive, the call_replicator worker thread parked indefinitely inside zmq::socket_base_t::process_commands; producers kept enqueuing CDR-payload Events until the in-memory queue reached 158 GB on a prod canary host. Apply a 5 s timeout in zeromq_create_socket() so send/recv return EAGAIN instead of blocking; internal_send_request's existing reconnect logic then fires correctly. Adds tests/unit_cpp/test_zmqsocket_timeouts: three assertions (PUSH has finite SNDTIMEO, PULL has finite RCVTIMEO, sendRequest returns false on a stuck peer within a watchdog window). * TELCORE-100: drop ZMQ_RCVTIMEO from Telnyx::ZMQ::ZMQSocket Per review on PR: the receive helpers (zeromq_receive / internal_receive_*) do not check socket.recv()'s return value and treat a timed-out empty frame as a successful receive. With ZMQ_RCVTIMEO=5000 set globally on every socket, a REP socket like zmq_request_handler would synthesize a successful empty request every 5 s and call send_reply with no matching inbound REQ, violating libzmq's REQ/REP state machine (EFSM). Drop the RCVTIMEO addition for now and keep ZMQ_SNDTIMEO, which is what actually fixes the call_replicator queue leak. Auditing and fixing the receive helpers to detect timeout/EAGAIN can ship as its own change. Test loses the PULL/RCVTIMEO assertion; SNDTIMEO + behavioural test still pass (8/0).

…ry (#587) * TELCORE-125: Fix NULL deref in switch_rtp_set_flag during call recovery. When switch_rtp_new() fails (typically bind EADDRINUSE on a recovered port), switch_core_media_activate_rtp() called switch_rtp_set_flag() unguarded for the telnyx_voicemail and ignore_rtp_during_dtmf paths, dereferencing NULL at switch_rtp.c:6373. Defensive NULL guards added to switch_rtp_set_flag and switch_rtp_clear_flag, plus switch_rtp_ready() checks at the call sites to match the surrounding pattern. Regression test in tests/unit/test_rtp_set_flag_null_crash.c isolates each scenario in a forked child. * TELCORE-125: make bind-conflict test use an OS-assigned port Address PR review: the regression test hardcoded BUSY_PORT 47823 and is run under `make check` (TESTS = noinst_PROGRAMS), so it would flake on any host where that UDP port is already occupied. Bind the hog socket to port 0 and read the assigned port back with getsockname() before handing it to switch_rtp_new(), making the bind-conflict scenario deterministic. * TELCORE-125: make bind-conflict test precondition contractual Address PR review (round 2): the bind-conflict scenario was observational — it logged switch_rtp_new()'s result but did 'Do NOT assert here', so the child could exit 0 even if switch_rtp_new() returned non-NULL, in which case it never touched the recovery NULL path it claims to cover. - Enforce the precondition: the child now exits non-zero (3) if switch_rtp_new() returns non-NULL, and (4) if it fails for a reason other than the bind conflict (err must contain 'Bind Error'). Only then is the guarded switch_rtp_set_flag() call made. - Decode the child exit code in the parent (ASSERT_CHILD_OK) so a CI failure reports why (crash vs. unmet precondition vs. harness setup), not just 'exited non-zero'. - Add a direct switch_rtp_clear_flag(NULL) scenario: the production fix guards both set and clear, so cover both. Verified locally: 3/3 pass; bind-conflict child logs 'switch_rtp_new -> (nil), err=Bind Error! 127.0.0.1:44698', confirming the NULL/bind precondition is reproduced before the guarded call. * TELCORE-125: harden child-reap in NULL-crash test against EINTR/waitpid failure run_in_child() now retries waitpid() on EINTR and reports any other waitpid()/fork() failure as a distinct harness error via its return value, rather than letting a zero-initialized status be misread as a clean child exit. ASSERT_CHILD_OK() treats that harness failure as a hard fail, closing the false-pass window in a crash-regression test. * TELCORE-125: avoid uint16_t tx_port wraparound in bind-conflict test The bind-conflict regression scenario passed busy_port + 2 as the remote tx_port. tx_port is a switch_port_t (uint16_t), so when the OS assigns busy_port == 65534, busy_port + 2 narrows to 0 at the call boundary. switch_rtp_new() then rejects with "Missing remote port" before reaching switch_rtp_set_local_address(), so the NULL is not caused by the bind conflict and the child exits 4 -- a low-probability but real flake. Compute remote_port without wrap (busy_port >= 65534 ? -2 : +2). The remote port is irrelevant to the local bind conflict; it only needs to stay nonzero.

…596) * TELCORE-144: fix is_resource_available throttle logic in mod_xml_rpc is_resource_available() had three defects that combined to break all HTTP API calls carrying a query string on hosts where throttle-on-idle-cpu was configured. Manifested as HTTP 500 from the Abyss base layer (handler_hook short-circuited to its `end:` label without overriding the default 500 status set early in the function as an Abyss-assert guard). Defects: 1. Inverted ternary in is_resource_available — the default branch returned SWITCH_FALSE, so any cmd not matching `cmd == "bgapi"` was reported as unavailable once a throttle list was configured. 2. is_throttled_api used a substring search against tokens wrapped in leading+trailing spaces (" originate "), which never matched a real api_str like "originate sofia/...". The throttle never actually throttled anything via the bgapi indirection. 3. set_throttled_api_calls never cleared the static set, so reload of xml_rpc.conf with a changed throttle-api silently appended entries instead of replacing them. assert(!zstr(api)) also compiled out in release builds, becoming UB on empty input. Fix: - set_throttled_api_calls clears before parsing, accepts NULL/empty as "no throttle", tokenizes on whitespace (spaces and tabs), inserts bare token strings. - is_resource_available short-circuits to SWITCH_TRUE when no throttle is configured (empty list or MIN_IDLE_CPU<=0). Otherwise, the cmd is checked against the set directly; if it's "bgapi" and the cmd itself isn't listed, the first whitespace-delimited token of api_str is checked. If neither matches, allow. If matched, deny only when idle is strictly below the watermark. Behaviour delta (intentional): - Direct calls to listed names (e.g. /api/originate?...) are now throttled when CPU is low. Previously bypassed the throttle code path entirely. - /api/<unlisted>?<query> now returns 200 in all cases. Previously 500. - Reloading the conf with a different throttle-api now replaces the list. Previously the union accumulated across reloads. Unit test added (test/test_rpc_helper.cpp) covering 24 cases, including explicit regression checks for each of the three defects. Builds and runs without libfreeswitch via stubbed switch_core_idle_cpu/zstr/ SWITCH_TRUE etc. Wired into the module's Makefile.am via noinst_PROGRAMS + TESTS so `make check` runs it. autoreconf required after this patch (adds a TESTS target). * TELCORE-144: reset throttle helper state at start of do_config() Review follow-up. The rpc_helper.cpp setters clear correctly, but do_config() only calls them from inside `if (!zstr(var) && !zstr(val))`. So an operator who empties throttle-api ("") or removes the throttle-api / throttle-on-idle-cpu params entirely would never trigger the setters on reload, leaving the previous throttle set and watermark active — production APIs stay throttled after an attempt to disable throttling. Reset throttle helper state (set_throttled_api_calls(NULL), set_min_idle_cpu_watermark("0")) before the param-parse loop, so present config values repopulate from a clean slate and absent/empty ones leave throttling disabled. Adds two config-boundary regression tests modelling the do_config() reset sequence: throttle configured -> reload with params removed, and -> reload with throttle-api emptied. test_rpc_helper now 29/29.

Accept a no-SDP INVITE, send our own SDP offer in a reliable 183 Session Progress (Require: 100rel), consume the answer from the PRACK, and complete with a 200 OK / ACK carrying no SDP. New profile param enable-3pcc-early-offer: true - emit offer-in-183 per call; override per session with the enable_3pcc_early_offer channel variable dialplan - defer every no-SDP INVITE to the dialplan so it can opt in per call via enable_3pcc_early_offer; calls that do not opt in fall back to offer-in-200 / answer-in-ACK Requires enable-100rel (silently disabled otherwise). The offer is emitted on the first 183 (pre_answer / instant_ringback / relayed early media); flows that never send a 183 fall back to offer-in-200 / answer-in-ACK. Media is terminated locally, so bypass/proxy-media calls fall through to existing handling. TELCORE-158

The fallback was added for MS Teams DR but Teams uses an app-layer media timer, not RFC 8445 ICE completion, so the fallback never helped Teams. Teams DR is now correctly handled by rtp_ice_role=controlling (b2bua PR #361), which sets role before media starts so this fallback never engages on those legs. PROD audit 2026-06-03 -> 2026-06-17 across US + EU b2bua: the fallback fires ~100-200x/day, and in every traced sample the peer answers within ~60-100 ms with STUN role-conflict 487, reverting us back to CONTROLLED. No fires on b2bua-rtc / b2bua-rtc-canary. The fallback never changes the call's effective role -- it's operational noise. Reverts: 4cd1201 (#602), 2c82dd1 (#600). switch_rtp.c and switch_rtp_pvt.h return to the parent of #600 (1d6120f); TELCORE-125 (#587) null-deref fix in switch_rtp_set_flag remains in place. Linear: https://linear.app/telnyx/issue/TELCORE-190

The trickle-recheck path in check_ice() activated ICE without calling gen_ice() first. When trickle candidates drained before the SDP-answer path populated local ufrag/pwd, FS armed ICE with NULL creds and produced outbound STUN USERNAME="<remote>:(null)" signed with an empty key. Peers 401'd every check -> one-way ICE -> media never established. - src/switch_core_media.c: call gen_ice() before the trickle-recheck switch_rtp_activate_ice() when local creds are missing. - src/switch_rtp.c: defensive guard in switch_rtp_activate_ice() that refuses VANILLA activation with empty login/rlogin/password/rpassword and logs at ERROR so future regressions surface loudly. Linear: https://linear.app/telnyx/issue/TELCORE-160

…() (#613)

Remove the explicit gpt-5.2-codex model pin from reviewpr.yml so reviewpr-internal uses its default model. Linear: https://linear.app/telnyx/issue/TELCORE-234/update-reviewpryml-to-fix-fail-gha-issue

…603)

…httapi (#608)

…teardown (#609)

…oreign codec teardown) (#604)

…gnature The cherry-picked TELCORE-218 change used the pre-1.11 switch_regex_perform signature (int *ovector, uint32_t olen). v1.11.1 migrated to PCRE2, where the call takes (switch_regex_t **, switch_regex_match_t **). Update the call and free the match data accordingly so mod_httapi builds against the synced tree.

deploy-development's 14 new commits were already cherry-picked onto this branch (TELCORE-100/125/144/155/156/160/170/190/192/193/218/234 + 3pcc), adapted to the v1.11.1 tree where signatures changed (e.g. TELCORE-218 switch_regex_perform -> PCRE2). Verified an -X ours merge introduces no deploy-development content beyond what this branch already contains, so this merge keeps the v1.11.1 tree verbatim and only records deploy-development as a parent to clear the PR #601 merge conflict.

…ent' into damir/telcore-176-sync-freeswitch-fork-with-upstream-v1111 # Conflicts: # tests/unit/.gitignore

The merge of deploy-development brought in new unit tests whose binaries were never added to .gitignore, so every build left them showing up as untracked. Add all 15 currently-unignored programs from noinst_PROGRAMS.

damirn · 2026-08-07T11:53:45Z

resolved merge conflicts

andywolk added 30 commits January 2, 2025 19:37

[mod_skypopen] Remove from tree.

8c5efce

Merge pull request signalwire#2685 from signalwire/skypopen

81796c0

[mod_skypopen] Remove from tree.

[mod_gsmopen] Remove from tree.

9ab4d17

Merge pull request signalwire#2686 from signalwire/gsmopen

4470f8d

[mod_gsmopen] Remove from tree.

[mod_sangoma_codec] Remove from tree

1944260

Merge pull request signalwire#2687 from signalwire/sangoma

2e06312

[mod_sangoma_codec] Remove from tree

[mod_ladspa] Remove from tree

4475a54

Merge pull request signalwire#2693 from signalwire/ladspa

87bfafb

[mod_ladspa] Remove from tree

[mod_cepstral] Remove from tree

b7ac2be

Merge pull request signalwire#2694 from signalwire/cepstral

eaf02c9

[mod_cepstral] Remove from tree

[mod_sms_flowroute] Remove from tree

e5ae035

Merge pull request signalwire#2696 from signalwire/flowroute

8a9977f

[mod_sms_flowroute] Remove from tree

[mod_dahdi_codec] Remove from tree

7d88add

Merge pull request signalwire#2698 from signalwire/dahdi

69fb55c

[mod_dahdi_codec] Remove from tree

[mod_portaudio, mod_portaudio_stream] Remove from tree

49e63f6

Merge pull request signalwire#2699 from signalwire/portaudio

4f1d9cd

[mod_portaudio, mod_portaudio_stream] Remove from tree

[mod_mp4, mod_mp4v, mod_mp4v2] Remove from tree

4235d86

Merge pull request signalwire#2700 from signalwire/mp4

c1eb76c

[mod_mp4, mod_mp4v, mod_mp4v2] Remove from tree

[mod_isac] Remove from tree

6286c51

Merge pull request signalwire#2702 from signalwire/isac

122c589

[mod_isac] Remove from tree

[mod_abstraction] Remove from tree

c1f4586

Merge pull request signalwire#2704 from signalwire/abstraction

0acb3f2

[mod_abstraction] Remove from tree

[mod_stress] Remove from tree.

866c235

Merge pull request signalwire#2692 from signalwire/stress

cdd6763

[mod_stress] Remove from tree.

[mod_unicall] Remove from tree

b033d27

Merge pull request signalwire#2705 from signalwire/unicall

1d8ff14

[mod_unicall] Remove from tree

[mod_soundtouch] Remove from tree

d912e9f

Merge pull request signalwire#2707 from signalwire/soundtouch

6a5ac7e

[mod_soundtouch] Remove from tree

[mod_sonar] Remove from tree

f248169

Merge pull request signalwire#2711 from signalwire/sonar

33f7be8

[mod_sonar] Remove from tree

minhtuan1407-telnyx self-requested a review June 19, 2026 07:02

dev-ryanc requested changes Jun 19, 2026

View reviewed changes

Comment thread src/include/switch_utils.h Outdated

baloeng force-pushed the telnyx/telephony/deploy-development branch from 4668f86 to 50f1d16 Compare June 22, 2026 20:41

TELCORE-176: remove duplicate switch_rand declaration from merge

96eedb9

damirn requested a review from dev-ryanc June 23, 2026 10:58

damirn mentioned this pull request Jun 23, 2026

TELCORE-81: upstream sync with freeswitch v1.11.0 release #586

Closed

5 tasks

baloeng reviewed Jun 30, 2026

View reviewed changes

baloeng and others added 16 commits July 1, 2026 15:15

TELCORE-170: Add aggressive-nomination fallback for ICE-controlled pe…

1d94834

…ers (#600)

Use switch_core_session_get_channel accessor in ICE nomination fallba…

e1bbe3a

…ck (#602)

TELCORE-156: clear stale EOC/PAC state in switch_core_media_clear_ice…

70e79ac

…() (#613)

TELCORE-234: remove pinned reviewpr model

1592c2f

Remove the explicit gpt-5.2-codex model pin from reviewpr.yml so reviewpr-internal uses its default model. Linear: https://linear.app/telnyx/issue/TELCORE-234/update-reviewpryml-to-fix-fail-gha-issue

TELCORE-155: respond immediately to inbound BYE after local teardown (#…

489c0be

…603)

TELCORE-218: skip HEAD pre-flight for S3 presigned media URLs in mod_…

dfb58f7

…httapi (#608)

TELCORE-193: fix assertion crash in switch_event_merge on conference …

5f6af4c

…teardown (#609)

TELCORE-192: fix crash segfault in switch_core_session_write_frame (f…

64e3d77

…oreign codec teardown) (#604)

damirn requested a review from baloeng July 1, 2026 13:32

damirn added 2 commits August 7, 2026 13:42

Merge remote-tracking branch 'origin/telnyx/telephony/deploy-developm…

bcd7b03

…ent' into damir/telcore-176-sync-freeswitch-fork-with-upstream-v1111 # Conflicts: # tests/unit/.gitignore

tests/unit: ignore built test binaries missing from .gitignore

c2e55bb

The merge of deploy-development brought in new unit tests whose binaries were never added to .gitignore, so every build left them showing up as untracked. Add all 15 currently-unignored programs from noinst_PROGRAMS.

Navigation Menu

Uh oh!

TELCORE-176: Sync FreeSWITCH fork with upstream v1.11.1 - #601

TELCORE-176: Sync FreeSWITCH fork with upstream v1.11.1#601
damirn wants to merge 397 commits into
telnyx/telephony/deploy-developmentfrom
damir/telcore-176-sync-freeswitch-fork-with-upstream-v1111

damirn commented Jun 1, 2026 •

edited

Loading

Uh oh!

Uh oh!

baloeng Jun 30, 2026

Uh oh!

damirn Jul 1, 2026

Uh oh!

baloeng Jun 30, 2026

Uh oh!

damirn Jul 1, 2026

Uh oh!

baloeng Jun 30, 2026

Uh oh!

damirn Jul 1, 2026

Uh oh!

baloeng Jun 30, 2026

Uh oh!

damirn Jul 1, 2026

Uh oh!

damirn commented Aug 7, 2026

Uh oh!

Reviewers

Assignees

Labels

Milestone

Development

Uh oh!

20 participants

		dtls_fingerprint_t fp = {0};

		fp.type = dtls->remote_fp->type;

Uh oh!

Conversation

damirn commented Jun 1, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Summary

Review follow-up (post-merge)

Fixed — duplicate SIP Reason handling in sofia_handle_sip_r_invite() (45bc70443a)

Investigated — no change needed

Verification

Coordination

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

damirn commented Aug 7, 2026

Uh oh!

Reviewers

Assignees

Labels

Milestone

Development

Uh oh!

20 participants

damirn commented Jun 1, 2026 •

edited

Loading

Fixed — duplicate SIP Reason handling in `sofia_handle_sip_r_invite()` (`45bc70443a`)