net: Use mockable time for ping/pong, add tests #18638

pull MarcoFalke wants to merge 2 commits into bitcoin:master from MarcoFalke:2004-netMockPing changing 10 files +173 −36
  1. MarcoFalke commented at 7:11 PM on April 14, 2020: member

    Switch CNode::m_ping_start and CNetMessage::m_time to mockable time, so that tests can be added.

    Mockable time is also type-safe, since it uses std::chrono

  2. MarcoFalke renamed this:
    net: Make ping mockable, add tests
    net: Use mocktime for ping/pong, add tests
    on Apr 14, 2020
  3. MarcoFalke renamed this:
    net: Use mocktime for ping/pong, add tests
    net: Use mockable time for ping/pong, add tests
    on Apr 14, 2020
  4. MarcoFalke force-pushed on Apr 14, 2020
  5. MarcoFalke force-pushed on Apr 14, 2020
  6. DrahtBot commented at 8:24 PM on April 14, 2020: member

    <!--e57a25ab6845829454e8d69fc972939a-->

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

    <!--174a7506f384e20aa4161008e828411d-->

    Conflicts

    Reviewers, this pull request conflicts with the following ones:

    • #19472 ([net processing] Reduce cs_main scope in MaybeDiscourageAndDisconnect() by jnewbery)
    • #19107 (p2p: Move all header verification into the network layer, extend logging by troygiorshev)

    If you consider this pull request important, please also help to review the conflicting pull requests. Ideally, start with the one that should be merged first.<!--2502f1a698b3751726fa55edcda76cd3-->

    Coverage

    Coverage Change (pull 18638, 0f96c4b9c8eed32c3fdf1203a2f9a2be59a5f4a2) Reference (master, 75fcfdaf8b4b68ef1d1fcd606fb36c9265e15918)
    Lines +0.0020 % 90.3582 %
    Functions +0.1431 % 85.0787 %
    Branches +0.0089 % 51.8202 %

    <sup>Updated at: 2020-04-15T00:09:25.322904.</sup>

  7. DrahtBot added the label P2P on Apr 14, 2020
  8. DrahtBot added the label Tests on Apr 14, 2020
  9. DrahtBot added the label Utils/log/libs on Apr 14, 2020
  10. MarcoFalke removed the label Tests on Apr 14, 2020
  11. MarcoFalke removed the label Utils/log/libs on Apr 14, 2020
  12. MarcoFalke force-pushed on Apr 14, 2020
  13. MarcoFalke force-pushed on Apr 14, 2020
  14. MarcoFalke force-pushed on Apr 14, 2020
  16. DrahtBot added the label Needs rebase on Apr 24, 2020
  17. MarcoFalke force-pushed on Apr 25, 2020
  18. DrahtBot removed the label Needs rebase on Apr 25, 2020
  19. MarcoFalke force-pushed on Apr 25, 2020
  20. laanwj commented at 1:22 PM on May 4, 2020: member

    Concept ACK.

    Won't this potentially cause issues (e.g. ping timeouts) in P2P tests that already use mockable time for other things, but are not expecting it to have an effect on pings?

    (seems not, given that all the tests still pass)

  21. MarcoFalke commented at 1:42 PM on May 4, 2020: member

    The -peertimeout is still measured in wall clock time, and because most of the test run through faster than what the -peertimeout is, this should not lead to issues in general. There are two tests test/functional/feature_maxuploadtarget.py and test/functional/feature_bip68_sequence.py which mine a lot of blocks (more than 100) under mocktime. This can take time when the tests are run under valgrind, so I had to bump the -peertimeout for those tests.

    Other tests should not be affected. If they are, it should be trivial to fix them up.

  22. practicalswift commented at 2:03 PM on May 4, 2020: contributor

    Concept ACK: mockable is better than non-mockable :)

  23. DrahtBot added the label Needs rebase on May 12, 2020
  24. MarcoFalke force-pushed on May 12, 2020
  25. DrahtBot removed the label Needs rebase on May 12, 2020
  26. MarcoFalke commented at 10:16 PM on May 12, 2020: member

    Rebased

  27. laanwj added this to the "Blockers" column in a project
  28. in test/functional/p2p_ping.py:102 in fac63aba8d outdated 
      97 | +        ping_delay = 29
  98 | +        self.mock_forward(ping_delay)
  99 | +        wait_until(lambda: 'ping' in no_pong_node.last_message)
 100 | +        no_pong_node.send_and_ping(msg_pong(no_pong_node.last_message.pop('ping').nonce))
 101 | +        self.check_peer_info(pingtime=ping_delay, minping=ping_delay, pingwait=None)
 102 | +
    glozow commented at 10:00 PM on May 18, 2020:

    From what I understand, minping is the best/shortest round trip time measured so far. Thus, would it be a good idea to repeat this chunk with a shorter ping_delay in order to check that minping gets updated correctly? I tried with ping_delay = 25 locally and it was updated from 29 to 25 as expected.

    glozow commented at 10:04 PM on May 18, 2020:

    Also I'm just wondering, why is ping_delay 29? I'm not sure if it's arbitrary or a chosen magic number.

    MarcoFalke commented at 10:37 PM on May 19, 2020:

    Good point! Will extend the test as suggested by you.

  29. in test/functional/p2p_ping.py:65 in fac63aba8d outdated 
      60 | +        self.mock_forward(3)
  61 | +        assert no_pong_node.last_message.pop('ping').nonce != 0
  62 | +        self.check_peer_info(pingtime=None, minping=None, pingwait=3)
  63 | +
  64 | +        self.log.info('Reply without nonce cancels ping')
  65 | +        with self.nodes[0].assert_debug_log(['pong peer=0: Short payload']):
    glozow commented at 10:20 PM on May 18, 2020:

    Hm, I'm surprised it isn't Nonce zero? I thought no nonce is equivalent to nonce=0 for msg_pong. The test failed when I changed the error message, though.

    MarcoFalke commented at 10:37 PM on May 19, 2020:

    Be aware that a msg_pong_corrupt is sent (not a msg_pong)

    glozow commented at 12:01 AM on May 20, 2020:

    Ah, that's it 👍

  30. in src/test/fuzz/process_message.cpp:77 in fac63aba8d outdated 
      73 | @@ -74,7 +74,7 @@ void test_one_input(const std::vector<uint8_t>& buffer)
  74 |      p2p_node.SetSendVersion(PROTOCOL_VERSION);
  75 |      g_setup->m_node.peer_logic->InitializeNode(&p2p_node);
  76 |      try {
  77 | -        (void)ProcessMessage(&p2p_node, random_message_type, random_bytes_data_stream, GetTimeMillis(), Params(), *g_setup->m_node.mempool, g_setup->m_node.connman.get(), g_setup->m_node.banman.get(), std::atomic<bool>{false});
    glozow commented at 10:37 PM on May 18, 2020:

    I think microseconds is correct (I think millis would be a bug if this wasn't in a unit test), but since it looks like you're switching from millis to micros, could this be an issue at all? The test passes for me.

    MarcoFalke commented at 10:35 PM on May 19, 2020:

    Yes, this looks like a bug in the test to me as well. Luckily by making the time type safe, this is less likely to happen in the future.

    Regardless, instead of depending on the current time, which is not deterministic and might cause issues while fuzzing, the fuzz test should ask the fuzz engine for a time value.

    However, this seems out of scope for the changes in this pull request.

  31. glozow commented at 11:13 PM on May 18, 2020: member

    Was able to build and all the tests passed locally. I don't know enough to solidly review the cpp code, but I tried to be more thorough with the functional tests. Everything makes sense to me, I just left a few questions.

  32. MarcoFalke force-pushed on May 19, 2020
  33. MarcoFalke commented at 11:28 PM on May 19, 2020: member

    Addressed feedback by @gzhao408

  34. jonatack commented at 6:49 PM on May 22, 2020: member

    Concept ACK. Reviewing/testing.

  35. DrahtBot added the label Needs rebase on May 23, 2020
  36. MarcoFalke force-pushed on May 23, 2020
  37. DrahtBot removed the label Needs rebase on May 23, 2020
  38. in src/net.cpp:689 in fa70471952 outdated 
     685 | @@ -686,7 +686,8 @@ const uint256& V1TransportDeserializer::GetMessageHash() const
 686 |      return data_hash;
 687 |  }
 688 |  
 689 | -CNetMessage V1TransportDeserializer::GetMessage(const CMessageHeader::MessageStartChars& message_start, int64_t time) {
 690 | +CNetMessage V1TransportDeserializer::GetMessage(const CMessageHeader::MessageStartChars& message_start, const std::chrono::microseconds& time)
    jonatack commented at 8:52 AM on May 24, 2020:

    std::chrono times are cheaply copied ints; would it be simpler/safer/faster to pass by value rather than reference to const?

    If yes, idem in net.h::L645+698, net_processing.cpp::L2085, and fuzz/process_message.cpp::L32

    Reference: https://isocpp.github.io/CppCoreGuidelines/CppCoreGuidelines#Rf-in

    MarcoFalke commented at 12:25 PM on May 24, 2020:

    I always follow rule 1 (const&), because it is always cheap and prevents the overhead of having to think how many words a type has.

    void f1(const string& s);  // OK: pass by reference to const; always cheap
    MarcoFalke commented at 12:25 PM on May 24, 2020:

    Changed, though.

  39. in src/net.cpp:1158 in fa70471952 outdated 
    1141 | @@ -1141,9 +1142,9 @@ void CConnman::InactivityCheck(CNode *pnode)
1142 |              LogPrintf("socket receive timeout: %is\n", nTime - pnode->nLastRecv);
1143 |              pnode->fDisconnect = true;
1144 |          }
1145 | -        else if (pnode->nPingNonceSent && pnode->nPingUsecStart + TIMEOUT_INTERVAL * 1000000 < GetTimeMicros())
1146 | +        else if (pnode->nPingNonceSent && pnode->m_ping_start.load() + std::chrono::seconds{TIMEOUT_INTERVAL} < GetTime<std::chrono::microseconds>())
    jonatack commented at 9:21 AM on May 24, 2020:

    Not changed by this PR, but noting that this seems to be the only place in the codebase where nPingNonceSent is tested for truthiness with implicit conversion instead of against an integer value.

    MarcoFalke commented at 12:27 PM on May 24, 2020:

    Indeed unrelated.

  40. in src/net.cpp:574 in fa70471952 outdated 
     560 |  
 561 |      // Raw ping time is in microseconds, but show it to user as whole seconds (Bitcoin users should be well used to small numbers with many decimal places by now :)
 562 |      stats.m_ping_usec = nPingUsecTime;
 563 |      stats.m_min_ping_usec  = nMinPingUsecTime;
 564 | -    stats.m_ping_wait_usec = nPingUsecWait;
 565 | +    stats.m_ping_wait_usec = count_microseconds(ping_wait);
    jonatack commented at 9:26 AM on May 24, 2020:

    Do you think it would be better to eventually convert these primitives to duration to no longer need helpers like count_microseconds? (or am I confused)

    MarcoFalke commented at 11:24 AM on May 24, 2020:

    I admit that the documentation isn't the best https://doxygen.bitcoincore.org/time_8h.html#a50aae909d5edf426afb66cd4d30c8d00

    However, those helpers will be needed to convert durations to raw integers for:

    • The RPC interface
    • The GUI
    • The logs
    • ... so pretty much anywhere the value "leaves" Bitcoin Core
    MarcoFalke commented at 12:27 PM on May 24, 2020:

    Extended documentation

    jonatack commented at 3:00 PM on May 24, 2020:

    Thanks!

  41. jonatack commented at 9:44 AM on May 24, 2020: member

    ACK fa704719529 good to reduce use of primitives for counting time in favor of the standard library and mockability, though it's fiddly and easy to introduce regressions while doing so (thanks for adding the tests). Builds with gcc and clang compile cleanly and the node seems to run normally. Built and ran the fuzzers without issue.

    <details><summary><code>fuzz/process_message</code> (ran for more than a day)</summary><p>

    $ src/test/fuzz/process_message ../qa-assets/fuzz_seed_corpus/

INFO: Seed: 2449945750
INFO: Loaded 1 modules   (498754 inline 8-bit counters): 498754 [0x561ad95753c8, 0x561ad95ef00a), 
INFO: Loaded 1 PC tables (498754 PCs): 498754 [0x561ad95ef010,0x561ad9d8b430), 
INFO:    52251 files found in ../qa-assets/fuzz_seed_corpus/
INFO: -max_len is not provided; libFuzzer will not generate inputs larger than 1048576 bytes
INFO: seed corpus: files: 52251 min: 1b max: 3984182b total: 462001668b rss: 227Mb
[#512](/bitcoin-bitcoin/512/)	pulse  cov: 6973 ft: 7092 corp: 10/13b exec/s: 256 rss: 516Mb
[#1024](/bitcoin-bitcoin/1024/)	pulse  cov: 7037 ft: 7168 corp: 12/19b exec/s: 256 rss: 516Mb
[#2048](/bitcoin-bitcoin/2048/)	pulse  cov: 7317 ft: 7541 corp: 18/45b exec/s: 292 rss: 516Mb

.../...

[#3094033](/bitcoin-bitcoin/3094033/)	REDUCE cov: 34621 ft: 110638 corp: 2221/41Mb exec/s: 23 rss: 1025Mb L: 92/1048576 MS: 1 EraseBytes-
[#3102635](/bitcoin-bitcoin/3102635/)	NEW    cov: 34621 ft: 110890 corp: 2222/41Mb exec/s: 22 rss: 1025Mb L: 111/1048576 MS: 2 CopyPart-ChangeByte-
[#3102867](/bitcoin-bitcoin/3102867/)	NEW    cov: 34621 ft: 110899 corp: 2223/41Mb exec/s: 22 rss: 1025Mb L: 107/1048576 MS: 2 CopyPart-CMP- DE: "-bans"-

    </p></details>

    <details><summary><code>fuzz/p2p_transport_deserializer</code></summary><p>

    $ src/test/fuzz/p2p_transport_deserializer ../qa-assets/fuzz_seed_corpus/

INFO: Seed: 159618322
INFO: Loaded 1 modules   (498301 inline 8-bit counters): 498301 [0x55c1d40d0d48, 0x55c1d414a7c5), 
INFO: Loaded 1 PC tables (498301 PCs): 498301 [0x55c1d414a7c8,0x55c1d48e4f98), 
INFO:    52840 files found in ../qa-assets/fuzz_seed_corpus/
INFO: -max_len is not provided; libFuzzer will not generate inputs larger than 1048576 bytes
INFO: seed corpus: files: 52840 min: 1b max: 3984182b total: 491369132b rss: 167Mb
[#8192](/bitcoin-bitcoin/8192/)	pulse  cov: 452 ft: 453 corp: 1/1b exec/s: 4096 rss: 254Mb
[#16384](/bitcoin-bitcoin/16384/)	pulse  cov: 1131 ft: 2680 corp: 34/1054b exec/s: 4096 rss: 389Mb
[#32768](/bitcoin-bitcoin/32768/)	pulse  cov: 1137 ft: 5091 corp: 77/7383b exec/s: 2978 rss: 536Mb

.../...

[#1036084](/bitcoin-bitcoin/1036084/)	REDUCE cov: 1137 ft: 7819 corp: 149/844Kb exec/s: 293 rss: 824Mb L: 6113/666283 MS: 4 ChangeByte-EraseBytes-CMP-CopyPart- DE: "\x0d\x00\x00\x00\x00\x00\x00\x00"-
[#1048576](/bitcoin-bitcoin/1048576/)	pulse  cov: 1137 ft: 7819 corp: 149/844Kb exec/s: 294 rss: 824Mb
[#1050168](/bitcoin-bitcoin/1050168/)	REDUCE cov: 1137 ft: 7819 corp: 149/843Kb exec/s: 294 rss: 824Mb L: 25076/666283 MS: 4 ShuffleBytes-InsertByte-ShuffleBytes-EraseBytes-

    </p></details>

  42. in test/functional/p2p_ping.py:40 in fa70471952 outdated 
      35 | +        pass
  36 | +
  37 | +
  38 | +class PingPongTest(BitcoinTestFramework):
  39 | +    def set_test_params(self):
  40 | +        self.setup_clean_chain = True
    jonatack commented at 10:49 AM on May 24, 2020:

    Is a clean chain needed for p2p testing here? The tests seem to be fine with this set to the default of false.

    MarcoFalke commented at 11:21 AM on May 24, 2020:

    Sending a ping/pong shouldn't require any blocks, so I set this to false for now. This can be removed when the test is extended and needs blocks.

    jonatack commented at 3:00 PM on May 24, 2020:

    Did you still want to set this to false? :angel:

    MarcoFalke commented at 3:41 PM on May 24, 2020:

    Seems overly expensive to copy the blockchain when it is not needed, no?

    jonatack commented at 3:58 PM on May 24, 2020:

    This might be a wrong conception I've had for a long time. I thought it was cheaper to use the cached data directories rather than set up a clean chain, and that the test framework defaults to self.setup_clean_chain = False for this reason. I reckoned the default was the cheaper option and the more expensive one was opt-in, so I always used False where possible.

    jonatack commented at 4:05 PM on May 24, 2020:

    It might be a good idea to add a sentence on performance in the "General test-writing advice" section in functional/README.md where it states "Set the self.setup_clean_chain variable in set_test_params() to control whether or not to use the cached data directories."

    MarcoFalke commented at 4:25 PM on May 24, 2020:

    The reason this default to a chain is that most tests require a chain, so so the least verbose option is the default.

    Also, defaulting to a chain prevents test writers from accidentally mining a chain when it is not needed. Mining is orders of magnitude slower than copying.

  43. MarcoFalke force-pushed on May 24, 2020
  44. MarcoFalke commented at 12:28 PM on May 24, 2020: member

    Added documentation to address @jonatack's feedback

  45. jonatack commented at 3:01 PM on May 24, 2020: member

    Code review re-ACK fabb382 changes since last review per git diff fa70471 fabb382 are passing time by value to GetMessage and ProcessMessage and improved documentation.

  46. laanwj removed this from the "Blockers" column in a project
  47. DrahtBot added the label Needs rebase on Jun 4, 2020
  48. MarcoFalke force-pushed on Jun 8, 2020
  49. MarcoFalke force-pushed on Jun 8, 2020
  50. in test/functional/p2p_ping.py:20 in fa94d6fef7 outdated 
      15 | +    wait_until,
  16 | +)
  17 | +from test_framework.test_framework import BitcoinTestFramework
  18 | +from test_framework.util import assert_equal
  19 | +
  20 | +PING_INTERVAL = 2 * 60
    jonatack commented at 2:08 PM on June 8, 2020:

    nit: if you retouch, could add a comment expressing the time units or along the lines of:

    PING_INTERVAL = 2 * 60  # 2 minutes (corresponds to net_processing::PING_INTERVAL)
  51. jonatack commented at 2:28 PM on June 8, 2020: member

    re-ACK fa94d6fef799a900c15dce9638eace5cc10087ae changes per git range-diff b3ec1fe fabb382 fa94d6f are updates for a recent merge converting pointers to references in net processing code. Re-did clang and fuzz debug builds, re-ran the new test p2p_ping and the process_message and p2p_transport_deserializer fuzz tests locally.

  52. DrahtBot removed the label Needs rebase on Jun 8, 2020
  53. DrahtBot added the label Needs rebase on Jun 19, 2020
  54. util: Add count_microseconds helper faab4aaf2f
  55. net: Use mockable time for ping/pong, add tests fa3365430c
  56. MarcoFalke force-pushed on Jun 19, 2020
  57. MarcoFalke commented at 11:51 AM on June 19, 2020: member

    Rebased. Should be easy to re-ACK with git range-diff bitcoin-core/master A B

  58. DrahtBot removed the label Needs rebase on Jun 19, 2020
  59. jonatack commented at 2:59 PM on June 21, 2020: member

    Code review re-ACK fa33654 re-read code, verified rebase per git range-diff 4b5c919 fa94d6f fa33654, previous tested ACKs still valid

  60. in test/functional/p2p_ping.py:60 in fa3365430c 
      55 | +        self.mock_time = int(time.time())
  56 | +        self.mock_forward(0)
  57 | +
  58 | +        self.log.info('Check that ping is sent after connection is established')
  59 | +        no_pong_node = self.nodes[0].add_p2p_connection(NodeNoPong())
  60 | +        self.mock_forward(3)
    fridokus commented at 11:48 AM on June 22, 2020:

    This is peertimeout right? If yes, we use peertimeout at 3 places any maybe it's more clear to define it at the outermost scope as well like PING_INTERVAL? Or maybe there's no need :)

    MarcoFalke commented at 2:04 PM on July 10, 2020:

    peertimeout is wall clock time, not mock time, I think. So this can be any value, as long as it matches pingwait=3 below, I'd say

  61. fridokus commented at 11:48 AM on June 22, 2020: contributor

    ACK faab4aaf2fa1153c6d76efc8113fa01b06943ece, ran functional tests and read the python scripts.

  62. in test/functional/p2p_ping.py:61 in fa3365430c 
      56 | +        self.mock_forward(0)
  57 | +
  58 | +        self.log.info('Check that ping is sent after connection is established')
  59 | +        no_pong_node = self.nodes[0].add_p2p_connection(NodeNoPong())
  60 | +        self.mock_forward(3)
  61 | +        assert no_pong_node.last_message.pop('ping').nonce != 0
    troygiorshev commented at 12:43 AM on June 24, 2020:

    Why pop here, as opposed to just accessing the value? Seems like having (message_count['ping'] == 1 and 'ping' not in last_message) == True could be confusing.

    Maybe decrementing message_count['ping'] every pop is an easier solution, being that this idea is used throughout the test.

            assert no_pong_node.last_message['ping'].nonce != 0

    Or maybe

            assert no_pong_node.last_message.get('ping').nonce != 0

    or

            assert no_pong_node.last_message.pop('ping').nonce != 0
        no_pong_node.message_count['ping'] -= 1
    MarcoFalke commented at 1:57 PM on July 10, 2020:

    I wasn't aware that message_count exists, nor am I using it in this test. Given that it is almost unused right now and causes (potential) confusion, I'd say to remove it

    MarcoFalke commented at 2:03 PM on July 10, 2020:

    the reason I use pop is to ensure the value is discarded after a read and has no way to poison follow-up test cases

  63. in test/functional/p2p_ping.py:73 in fa3365430c 
      68 | +
  69 | +        self.log.info('Reply without ping')
  70 | +        with self.nodes[0].assert_debug_log([
  71 | +                'pong peer=0: Unsolicited pong without ping, 0 expected, 0 received, 8 bytes',
  72 | +        ]):
  73 | +            no_pong_node.send_and_ping(msg_pong())
    troygiorshev commented at 12:46 AM on June 24, 2020:

    Why send_and_ping and not send_message, here and throughout? Doesn't seem like it would make a difference, and maybe it's best that we don't have extraneous pings and pongs.

    MarcoFalke commented at 1:55 PM on July 10, 2020:

    I don't like to use assert_debug_log as a way to synchronize/poll for a change

    send_and_ping is a way to flush all messages in the buffer (on both sides, send and receive) and ensure they are processed (and any errors have been logged).

    MarcoFalke commented at 2:04 PM on July 10, 2020:

    Also, sending a ping and receiving a ping should never interfere, unless I am missing something obvious

  64. in test/functional/p2p_ping.py:12 in fa3365430c 
       7 | +
   8 | +import time
   9 | +
  10 | +from test_framework.messages import (
  11 | +    msg_pong,
  12 | +)
    troygiorshev commented at 12:53 AM on June 24, 2020:

    nit: single line for a single import?

    MarcoFalke commented at 1:58 PM on July 10, 2020:

    Good point. Can be fixed in a follow-up

  65. troygiorshev commented at 1:30 AM on June 24, 2020: contributor

    ACK fa3365430c5fb57d7c0b5f2bce9fbbe290be93c3

    Reviewed, verified everything runs.

    I enjoyed how each sub-test flows into the next one. Hard to see at first but they really work together.

    A few questions.

  66. MarcoFalke commented at 2:05 PM on July 10, 2020: member

    Thanks for the review everyone. If there are ways to improve the test, I am happy to review them in a follow-up

  67. MarcoFalke merged this on Jul 10, 2020
  68. MarcoFalke closed this on Jul 10, 2020
  69. MarcoFalke deleted the branch on Jul 10, 2020
  70. sidhujag referenced this in commit c5d9d2b110 on Jul 10, 2020
  71. deadalnix referenced this in commit 0f688cf687 on Feb 3, 2021
  72. kittywhiskers referenced this in commit f6d0b79ab5 on Jul 13, 2021
  73. kittywhiskers referenced this in commit 0a2e3957b2 on Jul 13, 2021
  74. kittywhiskers referenced this in commit e1199999bc on Jul 15, 2021
  75. kittywhiskers referenced this in commit ee30f09225 on Jul 15, 2021
  76. kittywhiskers referenced this in commit 2360ac919f on Jul 16, 2021
  77. kittywhiskers referenced this in commit aa36b710ce on Aug 1, 2021
  78. DrahtBot locked this on Feb 15, 2022
Contributors
MarcoFalkeDrahtBotlaanwjpracticalswiftglozowjonatackfridokustroygiorshev
Labels
P2P
Linked (view graph)
#19580 test: Remove message_count
github-metadata-mirror

This is a metadata mirror of the GitHub repository bitcoin/bitcoin. This site is not affiliated with GitHub. Content is generated from a GitHub metadata backup.
generated: 2026-05-02 15:14 UTC
This site is hosted by @0xB10C
More mirrored repositories can be found on mirror.b10c.me