TCPPing changed in 2.9.0 release?

[luckman212]

Is there an existing issue for this?

• I have searched the existing issues

Current Behavior

Has something changed with TCPPing (/usr/bin/tcpping) or the way it's being parsed between 2.8 and 2.9? I had a single Target using this type of Probe and it's broken since this morning when I upgraded my container.

Everything else working fine, and I can execute the command from inside the container and it seems to work...

root@smokeping:/# tcpping -x 3 some.host.com 8443
seq 0: tcp response from pool-10-20-33-83.nwrknj.fios.verizon.net (173.54.100.1) [open]  10.957 ms
seq 1: tcp response from pool-10-20-33-83.nwrknj.fios.verizon.net (173.54.100.1) [open]  8.964 ms
seq 2: tcp response from pool-10-20-33-83.nwrknj.fios.verizon.net (173.54.100.1) [open]  12.270 ms

It was working...

Environment

OS: Synology DSM 7.2.2-72806

Probes (unchanged)

+ TCPPing
binary = /usr/bin/tcpping
forks = 10
offset = random
pings = 3
port = 80

Targets

+ Custom_TCP
menu = Custom TCP Checks
title = Custom TCP Checks
probe = TCPPing

++ Some.host
menu = Some.host
title = Some.host (Unifi/8443)
host = some.host.com
port = 8443

CPU architecture

x86-64

Docker creation

docker run -d \
--name smokeping \
--hostname smokeping \
--net bridge \
--dns 1.1.1.1 \
--dns 1.0.0.1 \
-p 8010:80 \
-v /volume1/docker/smokeping/config:/config \
-v /volume1/docker/smokeping/data:/data \
-v /volume1/docker/smokeping/cache:/var/cache/smokeping \
-e TZ='America/New_York' \
-e PUID=1030 \
-e PGID=100 \
linuxserver/smokeping

Container logs

[migrations] started
[migrations] no migrations found
usermod: no changes
───────────────────────────────────────

      ██╗     ███████╗██╗ ██████╗
      ██║     ██╔════╝██║██╔═══██╗
      ██║     ███████╗██║██║   ██║
      ██║     ╚════██║██║██║   ██║
      ███████╗███████║██║╚██████╔╝
      ╚══════╝╚══════╝╚═╝ ╚═════╝

   Brought to you by linuxserver.io
───────────────────────────────────────

To support LSIO projects visit:
https://www.linuxserver.io/donate/

───────────────────────────────────────
GID/UID
───────────────────────────────────────

User UID:    1030
User GID:    100
───────────────────────────────────────
Linuxserver.io version: 2.9.0-r0-ls133
Build-date: 2025-06-03T14:47:34+00:00
───────────────────────────────────────

[custom-init] No custom files found, skipping...
Connection to localhost (127.0.0.1) 80 port [tcp/http] succeeded!
[ls.io-init] done.
### assuming you are using an tcpping copy reporting in milliseconds
### assuming you are using an fping copy reporting in milliseconds
Smokeping version 2.009000 successfully launched.
Entering multiprocess mode.
### parsing dig output...OK
Child process 297 started for probe FPing.
FPing: probing 40 targets with step 120 s and offset 60 s.
Child process 298 started for probe TCPPing.
No targets defined for probe FPing6, skipping.
TCPPing: probing 1 targets with step 120 s and offset 94 s.
Child process 299 started for probe DNS.
All probe processes started successfully.
DNS: probing 11 targets with step 120 s and offset 71 s.

[thespad]

TCPPing was bumped to a newer version e472e5e#diff-de0a6f9e5e3a1e414c4ebf2543e3fb7cacfca08d8cb2f8bffc7bda5674b7d448 but I'm not aware of any breaking changes.

[luckman212]

Well, I messed around a bit, spun up 2 identical docker containers, one with :latest and the other with :2.8.2 and ran some tests

The results are interesting for sure...

2.9.0

Has problems with apple.com:443 and github.com:443. Also, requires an echo at the end, otherwise no newline is printed. This behavior is distinctly different vs. 2.8.2, where a the output is always newline terminated. Not sure if that matters (probably not).

2.8.2

No problems to any of the hosts tested.

The problem is definitely related to the -C option. Without it, the results appear normal.

[luckman212]

Ping @deajan any idea on this?

[deajan]

Any ideas which tcpping versions are used in 2.8.0 and 2.9.0 ?

[luckman212]

Yes, sorry I should have posted that.

• 2.8.2 used version 1.8
• 2.9.0 ships with 2.6-dev

[deajan]

Ahhh yes, indeed that's quite a gap.
What os/interpreter runs in the docker images ?

[luckman212]

The 2.9 lsio image is based on Alpine Linux 3.22. The shell would be Busybox/ash.

Bash v5.2.37 is also available at /bin/bash

The 2.8.2 image is Alpine 3.20, and has Bash 5.2.26 (but again, /bin/sh is busybox/ash)

[deajan]

Would you mind running tcpping 1.8 in the 2.9 image just to outrule a shell problem ?

[luckman212]

Sure. I just replaced the /usr/bin/tcpping script in my 2.9.0 image with the 1.8 from here and restarted the container.

The -C option works fine now...

root@smokeping:/# tcpping -C -x 5 github.com 443
github.com : 9.291 11.436 12.461 11.353 11.949

[thespad]

Would you mind stepping through a few versions between 1.8 and 2.6 and see if there's an obvious point that it breaks?

It's possible we're missing a dep etc. that's causing an issue with newer versions of tcpping and would be simple to fix.

[luckman212]

Sure. I just tested the earliest tag on this repo that exists: v2.0 and it fails even with that. So whatever the change was that is breaking things here happened between 1.8 and 2.0

I tried to eyeball the diff between the 1.8 commit and the 2.0 but it's quite a lot of changes. Someone with access to more granular changes might be able to bisect this more easily.

[luckman212]

Actually I spent a few more minutes and I think I found the issue.

I figured this out with the help of the -z flag (which I was unaware of) and a bit of manual testing. Seems to just be a problem with the way the output of tcptraceroute was being parsed by awk.

I made a quick PR over on deajan/tcpping#21 since this "bug" doesn't really belong to the LSIO project specifically. Will continue over there...

As of now, my TCP probe is back online!