test: support `get_bind_addrs` and `feature_bind

l0rinc commented at 9:42 PM on January 11, 2026: contributor

Problem

Some functional tests are shown as skipped when running on macOS & BSD because test_framework/netutil.py only implemented the Linux-specific logic for checking which TCP sockets a node is listening on.

Fix

Add macOS and BSD implementations in test/functional/test_framework/netutil.py so tests can query:

which TCP sockets a node is listening on (get_bind_addrs(), via lsof)
a non-loopback interface address (all_interfaces(), via ifconfig)

Then enable the previously Linux-only tests by switching to a shared POSIX platform guard.

Commands

<details> <summary><code>get_bind_addrs()</code> (<code>lsof</code> + regex)</summary>

Command used

lsof -nP -a -p <pid> -iTCP -sTCP:LISTEN -Ftn

Flags

-D: device cache warnings
-n: no hostname resolution
-P: no service/port-name resolution
-a: AND all conditions
-p <pid>: filter by process ID
-iTCP: TCP sockets only
-sTCP:LISTEN: listening sockets only
-Ftn: machine-readable output (fields: type t, name n)

Regex parser

t(IPv[46])\nn(\*|\[.+?]|[^:]+):(\d+)

Captured groups

group 1: IPv4 / IPv6 (used to disambiguate *)
group 2: host (*, [::1], 127.0.0.1, ...)
group 3: port
</details>

<details> <summary><code>all_interfaces()</code> (<code>ifconfig</code> + regex)</summary>

Command used

ifconfig -au

Regex parsing

Interface blocks:

(?m)^(?P<iface>\S+):(?P<block>[^\n]*(?:\n[ \t]+[^\n]*)*)

IPv4 extraction within each block:

inet (\S+)

</details>

Notes

The only remaining platform skips on macOS are the USDT/BPF tracing tests (interface_usdt_*.py).

DrahtBot added the label Tests on Jan 11, 2026

DrahtBot commented at 9:42 PM on January 11, 2026: contributor

The following sections might be updated with supplementary metadata relevant to reviewers and maintainers.

Code Coverage & Benchmarks

For details see: https://corecheck.dev/bitcoin/bitcoin/pulls/34256.

Reviews

See the guideline for information on the review process.

Type	Reviewers
Stale ACK	bensig, willcl-ark, Sjors, ismaelsadeeq

If your review is incorrectly listed, please copy-paste <code></code> into the comment that the bot should ignore.

in test/functional/test_framework/netutil.py:91 in 9b3a39cf60

  92 | -    bind_addrs = []
  93 | -    for conn in netstat('tcp') + netstat('tcp6'):
  94 | -        if conn[3] == STATE_LISTEN and conn[4] in inodes:
  95 | -            bind_addrs.append(conn[1])
  96 | -    return bind_addrs
  97 | +    if sys.platform.startswith("linux"):

Sjors commented at 4:14 AM on January 12, 2026:

In 9b3a39cf60dbad453874d3474dc21ecb960ffcfc test: support get_bind_addrs on macOS nit: you can get rid of startswith, since our minimum Python version is well after 3.3 which removed "linux2" and "linux3": https://docs.python.org/3.3/whatsnew/3.3.html#porting-python-code

Replace sys.platform == ‘linux2’ with sys.platform.startswith(‘linux’), or directly sys.platform == ‘linux’ if you don’t need to support older Python versions.

l0rinc commented at 7:47 AM on January 12, 2026:

Thanks, done, added you as coauthor

Sjors commented at 1:05 PM on February 23, 2026:

e70de0a406c072474b17d7c8ccff57fc952fae4b: it still says startswith

Sjors commented at 1:13 PM on February 23, 2026:

Nvm, you added a separate commit for this.

in test/functional/test_framework/netutil.py:104 in 9b3a39cf60 outdated

 105 | +        import re
 106 | +        import subprocess
 107 | +        output = subprocess.check_output(["lsof", "-nP", "-a", "-p", str(pid), "-iTCP", "-sTCP:LISTEN", "-Ftn"], text=True)
 108 | +        return [
 109 | +            (addr_to_hex(("::" if sock_type == "IPv6" else "0.0.0.0") if host == "*" else host.strip("[]")), int(port))
 110 | +            for sock_type, host, port in re.findall(r"t(IPv[46])\nn(\*|\[.+?]|[^:]+):(\d+)", output)

Sjors commented at 4:22 AM on January 12, 2026:

(delete wrong commit reference)

Review note, this is what the command outputs (without -Ftv):

% lsof -nP -a -p 92306 -iTCP -sTCP:LISTEN
COMMAND    PID  USER   FD   TYPE             DEVICE SIZE/OFF NODE NAME
bitcoind 92306 sjors   10u  IPv6 0x41f09f8c114a7707      0t0  TCP [::1]:18443 (LISTEN)
bitcoind 92306 sjors   11u  IPv4 0x9b8ff0d88c43ff6f      0t0  TCP 127.0.0.1:18443 (LISTEN)
bitcoind 92306 sjors   19u  IPv4 0xd8de1cae67469884      0t0  TCP *:18444 (LISTEN)
bitcoind 92306 sjors   21u  IPv4 0x33676f2f9e2adc58      0t0  TCP 127.0.0.1:18445 (LISTEN)
bitcoind 92306 sjors   22u  IPv6 0x776a345d97d44893      0t0  TCP *:18444 (LISTEN)

And with -Ftv:

p92306
f10
tIPv6
n[::1]:18443
f11
tIPv4
n127.0.0.1:18443
f19
tIPv4
n*:18444
f21
tIPv4
n127.0.0.1:18445
f22
tIPv6
n*:18444

I didn't study the regex in much detail, I guess it works...

l0rinc commented at 7:47 AM on January 12, 2026:

Please see the collapsed sections in the PR description for more details. Let me know if you think more explanation is needed anywhere.

willcl-ark commented at 9:48 AM on February 5, 2026:

Instead of the double ternary inside a list comp here, this could perhaps be written in slightly more human-readable python, as:

results = []
for sock_type, host, port in re.findall(r"t(IPv[46])\nn(\*|\[.+?]|[^:]+):(\d+)", output):
    if host == "*":
        addr = "::" if sock_type == "IPv6" else "0.0.0.0"
    else:
        addr = host.strip("[]")
    results.append((addr_to_hex(addr), int(port)))
return results

I also did not study this regex precisely, but have observed that it works on my macbook.

l0rinc commented at 10:15 AM on February 5, 2026:

If I need to touch it again, I'll consider avoiding the nesting, thanks for the hint. Was going for compactness here, but maybe I overshot.

l0rinc commented at 11:48 AM on February 26, 2026:

I haven't applied this in the end, but if you feel strongly about it, let me know

l0rinc commented at 3:25 PM on March 12, 2026:

I kept the nesting for now, let me know if you feel strongly about it

Sjors commented at 4:24 AM on January 12, 2026: member

Concept ACK, still need to look at the third and fourth commit.

Tested that the two functional tests now run on macOS. I also introduced some regressions to see if they're caught.

Nit: commit messages contain literal \n instead of newline.

l0rinc force-pushed on Jan 12, 2026

bensig commented at 9:54 PM on January 12, 2026: contributor

ACK a843fde881820a70b72f0be0ff6e9602bb7ae1dc

DrahtBot requested review from Sjors on Jan 12, 2026

in test/functional/test_framework/netutil.py:111 in 2296a19c96

 105 | @@ -106,33 +106,44 @@ def get_bind_addrs(pid):
 106 |      else:
 107 |          raise NotImplementedError(f"get_bind_addrs is not supported on {sys.platform}")
 108 |  
 109 | -# from: https://code.activestate.com/recipes/439093/
 110 |  def all_interfaces():
 111 |      '''
 112 |      Return all interfaces that are up

Sjors commented at 8:58 AM on January 13, 2026:

Let's clarify that this only returns IPv4 interfaces (only obvious to people who recognise AF_INET in the Linux code).

l0rinc commented at 12:24 PM on January 16, 2026:

Done in: https://github.com/bitcoin/bitcoin/compare/a843fde881820a70b72f0be0ff6e9602bb7ae1dc..67002b0b2cb4e2520c8f48cd048e3522d9c76c32

in test/functional/test_framework/netutil.py:143 in 2296a19c96 outdated

 161 | +        import subprocess
 162 | +        output = subprocess.check_output(["ifconfig", "-au"], text=True)
 163 | +        return [
 164 | +            (m["iface"].encode(), ip)
 165 | +            for m in re.finditer(r"(?m)^(?P<iface>\S+):(?P<block>[^\n]*(?:\n[ \t]+[^\n]*)*)", output)
 166 | +            for ip in re.findall(r"inet (\S+)", m["block"])

Sjors commented at 3:16 PM on January 13, 2026:

In 2296a19c96294eca9f376249a1745cd077fb1f6a test: support all_interfaces on macOS: I don't mind keeping the regex approach, but something like this is easier to read:

# Get list of "up" interfaces
ifaces = subprocess.check_output(["ifconfig", "-lu"], text=True).split()
return [
    (iface.encode(), res.strip()) for iface in ifaces
    if (res := subprocess.run(["ipconfig", "getifaddr", iface], text=True).stdout) is not None
]

Sjors commented at 8:05 PM on January 13, 2026:

Actually the above short version doesn't work, it just returns an empty result, which the test seems to be happy with.

Back to the more verbose version:

results = []
for iface in ifaces:
    try:
        res = subprocess.check_output(
            ["ipconfig", "getifaddr", iface],
            text=True
        ).strip()
    except subprocess.CalledProcessError:
        continue
    if res:
        results.append((iface.encode(), res))

return results

However it's still not identical to your version, because it doesn't the return the loopback IP. That's because ipconfig getifaddr lo0 returns nothing. Might be a matter of tweaking the getifaddr command, so the man page doesn't suggest any obvious way.

l0rinc commented at 12:24 PM on January 16, 2026:

Added a check for empty return, added you as coauthor: https://github.com/bitcoin/bitcoin/compare/a843fde881820a70b72f0be0ff6e9602bb7ae1dc..67002b0b2cb4e2520c8f48cd048e3522d9c76c32

DrahtBot requested review from Sjors on Jan 13, 2026

l0rinc force-pushed on Jan 16, 2026

ismaelsadeeq commented at 1:53 PM on February 4, 2026: member

Tested ACK 67002b0b2cb4e2520c8f48cd048e3522d9c76c32 on macOS M2 Pro Tahoe 26.1.

Confirmed that the previously skipped networking tests now run successfully:

feature_bind_extra.py
rpc_bind.py --ipv4
rpc_bind.py --ipv6
rpc_bind.py --nonloopback

Before: 21 skipped tests (including the 4 above) After: 16 skipped tests (only USDT/tracing + compatibility tests remain skipped)

Only did testing in this round.

willcl-ark approved

willcl-ark commented at 9:54 AM on February 5, 2026: member

tACK 13a791c85bbcbaa0c2bf7312a7b295dfd0a9f927

This looks ok to me. Left a comment on breaking up the large list comp with two ternaries, in case you touch again.

I was trying to think of cleaner approaches, but couldn't think of any outside of adding psutil as a test dependency, which I think we can avoid while we only need this single function.

All tests pass on an M3 Pro running MacOS Tahoe 26.2 with lsof 4.91

DrahtBot requested review from willcl-ark on Feb 5, 2026

willcl-ark commented at 10:09 AM on February 5, 2026: member

I have noticed that above I ACKed the commits I had rebased onto master.

ACK 67002b0b2cb4e2520c8f48cd048e3522d9c76c32

(after checking git range-diff 13a791c85bbcbaa0c2bf7312a7b295dfd0a9f927...67002b0b2cb4e2520c8f48cd048e3522d9c76c32)

in test/functional/test_framework/test_framework.py:910 in 67002b0b2c

 906 | @@ -907,6 +907,11 @@ def skip_if_platform_not_linux(self):
 907 |          if platform.system() != "Linux":
 908 |              raise SkipTest("not on a Linux system")
 909 |  
 910 | +    def skip_if_platform_not_linux_or_mac(self):

fanquake commented at 10:16 AM on February 5, 2026:

Why can't this use skip_if_platform_not_posix?

l0rinc commented at 12:42 PM on February 5, 2026:

Valid question, I remember the CI giving me some problems, let me recheck https://github.com/l0rinc/bitcoin/pull/115

fanquake commented at 1:51 PM on February 5, 2026:

Looks like it's working fine: https://github.com/l0rinc/bitcoin/actions/runs/21711801125/job/62617072040?pr=115.

l0rinc commented at 2:45 PM on February 5, 2026:

It does work for Mac, I already tried that locally, I remembered that it enabled something that I didn't want on CI - let's see if the PR passes and I'll adjust it here if it does, thanks for the hint.

l0rinc commented at 11:49 PM on February 5, 2026:

Changed it to use the posix helper, not sure what went wrong originally, but seems to be working now, thanks: https://github.com/bitcoin/bitcoin/compare/67002b0b2cb4e2520c8f48cd048e3522d9c76c32..10e1e38c75312c5d5219c1afd1232df252aaaa9a

l0rinc force-pushed on Feb 5, 2026

l0rinc requested review from ismaelsadeeq on Feb 6, 2026

l0rinc requested review from fanquake on Feb 6, 2026

ismaelsadeeq commented at 5:14 PM on February 9, 2026: member

re-tACK 10e1e38c75312c5d5219c1afd1232df252aaaa9a

Previously skipped tests are now running as reported in my comment here #34256#pullrequestreview-3751197721 after the recent push

Sjors commented at 1:14 PM on February 23, 2026: member

ACK 10e1e38c75312c5d5219c1afd1232df252aaaa9a

l0rinc commented at 1:38 PM on February 27, 2026: contributor

rfm?

fanquake commented at 10:28 AM on March 6, 2026: member

@willcl-ark want to take another look here?

in test/functional/rpc_bind.py:20 in 9b6aa99710 outdated

  16 | @@ -17,8 +17,7 @@ def set_test_params(self):
  17 |          self.supports_cli = False
  18 |  
  19 |      def skip_test_if_missing_module(self):
  20 | -        # due to OS-specific network stats queries, this test works only on Linux
  21 | -        self.skip_if_platform_not_linux()
  22 | +        self.skip_if_platform_not_posix()

willcl-ark commented at 1:28 PM on March 11, 2026:

nit: the commit message still states

allow rpc_bind to run there as well by switching to skip_if_platform_not_linux_or_mac().

l0rinc commented at 2:34 PM on March 11, 2026:

Indeed, will @ismaelsadeeq and @Sjors reack if I adjust it?

in test/functional/test_framework/netutil.py:146 in 10e1e38c75 outdated

 172 | +            (m["iface"].encode(), ip)
 173 | +            for m in re.finditer(r"(?m)^(?P<iface>\S+):(?P<block>[^\n]*(?:\n[ \t]+[^\n]*)*)", output)
 174 | +            for ip in re.findall(r"inet (\S+)", m["block"])
 175 | +        ]
 176 | +    else:
 177 | +        raise NotImplementedError(f"all_interfaces is not supported on {sys.platform}")

willcl-ark commented at 1:56 PM on March 11, 2026:

How will this work on *BSD? We dropped the gate from if: linux to if: posix, but then only handle linux and darwin here. So will BSD now run these tests and always hit this exception?

l0rinc commented at 2:33 PM on March 11, 2026:

Not sure, but this only affects the tests - is anyone running them on BSD? If it ever becomes a priority, we can adjust the tests.

willcl-ark commented at 2:35 PM on March 11, 2026:

They are being run in a few places I know of, e.g. here https://github.com/hebasto/bitcoin-core-nightly/actions/runs/22936550559/workflow#L97

maflcko commented at 2:48 PM on March 11, 2026:

Yeah, also the docs recommend to run the tests:

To run the test suite (recommended), you will need to have Python 3 installed:

See doc/build-freebsd.md

l0rinc commented at 3:10 PM on March 11, 2026:

So that was the reason for #34256 (review). Did we remove a BSD CI task since then? I remember a similar failure, but it seems I forgot the reason. I'll revert that change, thanks.

fanquake commented at 3:12 PM on March 11, 2026:

Did we remove a BSD CI task since then?

No.

fanquake commented at 3:13 PM on March 11, 2026:

So will BSD now run these tests and always hit this exception?

Why can't this just skip, for an unimplemented platform? Rather than re-adding more redundant code?

l0rinc commented at 3:21 PM on March 11, 2026:

Yes, that's what the origin skip_if_platform_not_linux_or_mac did - I just completely forgot about BSD

fanquake commented at 3:22 PM on March 11, 2026:

I mean skip in the test, not introduce a new function, to differentiate a subset of posix (for a single functional test).

l0rinc commented at 12:48 PM on March 13, 2026:

Thanks for the feedback. The PR is simpler now, and the tests are passing on the 3 BSD versions as well, without any significant change to the original implementation.

in test/functional/test_framework/netutil.py:143 in 10e1e38c75 outdated

 169 | +        import subprocess
 170 | +        output = subprocess.check_output(["ifconfig", "-au"], text=True)
 171 | +        return [
 172 | +            (m["iface"].encode(), ip)
 173 | +            for m in re.finditer(r"(?m)^(?P<iface>\S+):(?P<block>[^\n]*(?:\n[ \t]+[^\n]*)*)", output)
 174 | +            for ip in re.findall(r"inet (\S+)", m["block"])

willcl-ark commented at 1:57 PM on March 11, 2026:

Agree it's a (slight) shame about all the regex needed, but don't know a better way to do it either without a large rewrite to ctypes or pulling in psutil (as Sjors already noted).

DrahtBot requested review from willcl-ark on Mar 11, 2026

l0rinc force-pushed on Mar 12, 2026

l0rinc commented at 3:19 PM on March 12, 2026: contributor

Managed to enable BSD support for both tests, squashed most multi-step commits, updated commit messages, see: https://github.com/bitcoin/bitcoin/compare/10e1e38c75312c5d5219c1afd1232df252aaaa9a..b57910b21e63237b69bb06e1e8e9cc9631814922#diff-f19c9e4c8e34419e2d51415cf93c489820045753de415b6d5fc0339b4ef58bc3

l0rinc referenced this in commit 8132885af0 on Mar 12, 2026

l0rinc renamed this:
~~test: support `get_bind_addrs` and `feature_bind_extra` on macOS~~
test: support `get_bind_addrs` and `feature_bind_extra` on macOS & BSD
on Mar 13, 2026

ismaelsadeeq commented at 3:49 PM on March 30, 2026: member

Partial (Because I did not test on BSD platform) re-Tested ACK b57910b21e on macOS M2 Pro (Tahoe 26.1).

Confirmed that the networking-related tests enabled by this PR continue to run successfully on macOS:

These tests feature_bind_extra.py rpc_bind.py --ipv4 rpc_bind.py --ipv6 rpc_bind.py --nonloopback

were skipped on master f6e6fad0d9

These tests execute and pass on this PR b57910b21e63237b69bb06e1e8e9cc9631814922

l0rinc commented at 8:10 AM on April 6, 2026: contributor

@willcl-ark, @Sjors, @fanquake, I have addressed the concerns, please let me know if anything else is needed.

Edit:

Because I did not test on BSD platform

Rebased to enable the CI's new BSD job

test: enable `feature_bind_extra` on macOS and BSD

`feature_bind_extra` checks `-bind` and `-whitebind` by comparing the node's listening sockets with the expected addresses.
Add `get_bind_addrs` support for macOS, FreeBSD, NetBSD, and OpenBSD using `lsof`, and switch the test to the POSIX platform guard so it runs there too.
On FreeBSD, pass `-Di` to avoid device-cache warnings on stderr that the functional test runner treats as failures.

Co-authored-by: willcl-ark <will8clark@gmail.com>
Co-authored-by: fanquake <fanquake@gmail.com>

7236a05503

test: enable `rpc_bind` on macOS and BSD

`rpc_bind` uses `all_interfaces` to find a non-loopback IPv4 address and `get_bind_addrs` to verify the node's listening sockets.
Add `all_interfaces` support for macOS, FreeBSD, NetBSD, and OpenBSD using `ifconfig -au`, switch the test to the POSIX platform guard so it runs there too, and fail early if no IPv4 interfaces are returned.

Co-authored-by: Sjors Provoost <sjors@sprovoost.nl>

1950da94fc

l0rinc force-pushed on Apr 6, 2026

test: support `get_bind_addrs` and `feature_bind_extra` on macOS & BSD #34256

Problem

Fix

Commands

Notes

Code Coverage & Benchmarks

Reviews