This commit is (intentionally) adding a broken test. The test is broken
because it registering a subscriber object that can go out of scope
while events are still being sent.
To run the broken test and reproduce the bug:
- Remove comment /** and */
- ./configure --with-sanitizers=address
- export ASAN_OPTIONS=detect_leaks=0
- make
- while ./src/test/test_bitcoin -t validationinterface_tests/unregister_validation_interface_race --catch_system_errors=no ; do true; done
9f5608c289 test: check for matching object hashes in wait_for_getdata (Danny Lee)
Pull request description:
Previously, `wait_for_getdata` only looked for the presence of a recent `"getdata"` message. Additionally checking the object hashes inside the message should make tests involving `wait_for_getdata` more robust.
`p2p_sendheaders.py` already overrides `wait_for_getdata` do this check; we can use the same approach consistently across all tests that call `wait_for_getdata`.
This PR is progress towards #18614 , but closing that issue would also involve some additional changes to `wait_for_getheaders`.
ACKs for top commit:
theStack:
ACK 9f5608c289🍻
Tree-SHA512: 8e7f95881c19631db014d4bb2399fea0d14686a32542f6ca3b60809744b0d684eac4e4c107c87143991f3cd0c2d4ab09d0c17486239768a9b40bee25f2e4d54a
fa60afc4fb wallet: Add BlockUntilSyncedToCurrentChain to dumpwallet (MarcoFalke)
Pull request description:
dumpwallet includes the block hash in the output, so this method depends on the chainstate. According to the developer notes e84a5f0004/doc/developer-notes.md (L1095) it must include a `BlockUntilSyncedToCurrentChain`.
This is a minor fix and does not need backport, I think.
It fixes test failures such as https://travis-ci.org/github/bitcoin/bitcoin/jobs/675487097#L2657 , which can only happen in master because the test was not backported.
ACKs for top commit:
promag:
Code review ACK fa60afc4fb.
ryanofsky:
Code review ACK fa60afc4fb
meshcollider:
utACK fa60afc4fb
Tree-SHA512: 8df70b06b226b2cdf880dec9264adb72d66fd81b09b404fd1665a79e5f5236d26122eebf15df00fe71ee292b5c91b2dc23a0a42b2aa50a8d690604b23832723f
faff9e4bb4 test: Remove unused, undocumented and misleading CScript.__add__ (MarcoFalke)
Pull request description:
See the corresponding pull #18612
ACKs for top commit:
laanwj:
ACK faff9e4bb4 provided it passes Travis
Tree-SHA512: 5d9c4d5b6453c70b24a6960d3b42834e9b31f6dbb99ac47a6abfd85f2739d5372563e7188c22aceabeee1c37eb218bf580848356f4a77268d65f178a9419b269
fa1fdb02fc bench: Replace ::mempool globabl with test_setup.mempool (MarcoFalke)
fab1170964 bench: Remove requirement that all benches use RegTestingSetup (MarcoFalke)
Pull request description:
The benches have always set up one global testing setup. This makes it hard to pick no testing setup at all or one with different params.
Fix this by removing any global state setup from the main `bench.cpp` and leave the setup to each individual bench.
One reason to have one global testing setup is to set the datadir location to a tempdir to avoid reading or writing in the default datadir location. But #13687 should prevent this already.
Top commit has no ACKs.
Tree-SHA512: 7c98aea7725a20f4b9225221f4279b9e9f7257ed5c14712ad01ea80d87c3b0fed760b40f413892498bbb354a917ee02d4c575cbe8423a403b86755e8ee11f33b
8508473094 Avoid non-trivial global constants in SHA-NI code (Pieter Wuille)
Pull request description:
This is a potential solution for #18456.
It seems that the compiler cannot turn `_mm_set_epi64x(<constant>,<constnant>)` into a constant itself, and thus emits a global initializer for the `MASK`, `INIT0`, and `INIT1` global constants in the sha-ni SHA256 implementation.
Change this by turning them into dumb byte arrays, loading them into an SSE variable whenever needed.
Tested on a SHA-NI capable machine. I do not observe any obvious performance impact (but this is hard to measure, it's already very fast...).
ACKs for top commit:
laanwj:
Code review ACK 8508473094
elichai:
ACK 8508473094
Tree-SHA512: 07049cf1a33624c22df2be48b814d5636c037b368861eb13ee073bdce2b7c902a56e96518218961f55a2a1631a40825ded6dbbc28d7fe0e7fec267d704e39112
21fa0a44ab [docs] use consistent naming for possible_overwrite (John Newbery)
2685c214cc [tests] small whitespace fixup (John Newbery)
e9936966c0 scripted-diff: Rename PRUNED to SPENT in coins tests (John Newbery)
c205979031 [docs] Improve commenting in coins.cpp|h (John Newbery)
Pull request description:
- Add full commenting for spentness / DIRTYness / FRESHness and which combinations are valid
- Remove the 'pruned' terminology, which doesn't make sense since per-txout chainstate db was merged (#10195).
- Rename `potential_overwrite` to `possible_overwrite` to standardize terminology (there were previously examples of both, which made searching the codebase difficult).
- Make other minor improvements to the comments
ACKs for top commit:
jonatack:
Re-ACK 21fa0a4 per `git diff 98bee55 21fa0a4` the only change since my previous review is the following code commenting diff in `src/coins.cpp::L177-179`; rebuilt/ran unit tests anyway as a sanity check on the unit test changes.
Tree-SHA512: 391e01588ef5edb417250080cec17361f982c4454bc5f8c6d78bbd528c68a2bb94373297760691295c24660ce1022ad3ef7599762f736c8eed772ce096d38c3d
b91e4ae0d8 Do not expose and consider -logthreadnames when it does not work (Hennadii Stepanov)
Pull request description:
There are conditions when the `HAVE_THREAD_LOCAL` macro is undefined what causes the `-logthreadnames` option does not work -- instead of thread names empty strings `[]` only are printed in the `debug.log` file.
This PR does not exposes the `-logthreadnames` option in such cases.
Refs:
- #16059
- #18652
ACKs for top commit:
MarcoFalke:
ACK b91e4ae0d8, looked at the diff, didn't test
Tree-SHA512: 3bd58e5ea603c69686589ddc94d6fa441cab4f712004378f2f1661e12638804ca03cfb6426e6393e55b6a095b325f3161d3c5371af05d7fc79d6d328227bf40c
ccccd51908 script: Remove undocumented and unused operator+ (MarcoFalke)
Pull request description:
This operator has no documented use case and is also unused outside of test code. The test code and all other (imaginary) code that might use this operator is written more clear and concise by the existing CScript push operators for opcodes and data.
Removing the operator is also going to protect against accidentally reintroducing bugs like this 6ff5f718b6 (diff-8458adcedc17d046942185cb709ff5c3L1135) (last time it was used).
ACKs for top commit:
laanwj:
ACK ccccd51908
Tree-SHA512: 43898ac77e4d9643d9f8ac6f8f65497a4f0bbb1fb5dcaecc839c3719aa36181ba77befb213e59a9f33a20a29e0173a0e9c4763b1930940b32c3d1598b3e39af9
8334ee31f8 scripts: add MACHO LAZY_BINDINGS test to test-security-check.py (fanquake)
7b99c7454c scripts: add MACHO Canary check to security-check.py (fanquake)
Pull request description:
7b99c7454c uses `otool -Iv` to check for `___stack_chk_fail` in the macOS binaries. Similar to the [ELF check](https://github.com/bitcoin/bitcoin/blob/master/contrib/devtools/security-check.py#L105). Note that looking for a triple underscore prefixed function (as opposed to two for ELF) is correct for the macOS binaries. i.e:
```bash
otool -Iv bitcoind | grep chk
0x00000001006715b8 509 ___memcpy_chk
0x00000001006715be 510 ___snprintf_chk
0x00000001006715c4 511 ___sprintf_chk
0x00000001006715ca 512 ___stack_chk_fail
0x00000001006715d6 517 ___vsnprintf_chk
0x0000000100787898 513 ___stack_chk_guard
```
8334ee31f8 is a follow up to #18295 and adds test cases to `test-security-check.py` that for some reason I didn't add at the time. I'll sort out #18434 so that we can run these tests in the CI.
ACKs for top commit:
practicalswift:
ACK 8334ee31f8: Mitigations are important. Important things are worth asserting :)
jonasschnelli:
utACK 8334ee31f8.
Tree-SHA512: 1aa5ded34bbd187eddb112b27278deb328bfc21ac82316b20fab6ad894f223b239a76b53dab0ac1770d194c1760fcc40d4da91ec09959ba4fc8eadedb173936a
b155fcda51 doc: fix typo in configure.ac (fanquake)
20a30922fb doc: note why we can't use thread_local with glibc back compat (fanquake)
Pull request description:
Given that we went through a [gitian build](https://github.com/bitcoin/bitcoin/pull/18681) to remember why this is the case, we might as well make a note of it in configure.ac.
[From #18681](https://github.com/bitcoin/bitcoin/pull/18681#issuecomment-615526634):
Looking at the Linux build log, this has failed with:
```bash
Checking glibc back compat...
bitcoind: symbol __cxa_thread_atexit_impl from unsupported version GLIBC_2.18
bitcoind: failed IMPORTED_SYMBOLS
bitcoin-cli: symbol __cxa_thread_atexit_impl from unsupported version GLIBC_2.18
bitcoin-cli: failed IMPORTED_SYMBOLS
bitcoin-tx: symbol __cxa_thread_atexit_impl from unsupported version GLIBC_2.18
bitcoin-tx: failed IMPORTED_SYMBOLS
bitcoin-wallet: symbol __cxa_thread_atexit_impl from unsupported version GLIBC_2.18
bitcoin-wallet: failed IMPORTED_SYMBOLS
test/test_bitcoin: symbol __cxa_thread_atexit_impl from unsupported version GLIBC_2.18
test/test_bitcoin: failed IMPORTED_SYMBOLS
bench/bench_bitcoin: symbol __cxa_thread_atexit_impl from unsupported version GLIBC_2.18
bench/bench_bitcoin: failed IMPORTED_SYMBOLS
qt/bitcoin-qt: symbol __cxa_thread_atexit_impl from unsupported version GLIBC_2.18
```
`__cxa_thread_atexit_impl` is used for [thread_local variable destruction](https://sourceware.org/glibc/wiki/Destructor%20support%20for%20thread_local%20variables):
> To implement this support, glibc defines __cxa_thread_atexit_impl exclusively for use by libstdc++ (which has the __cxa_thread_atexit to wrap around it), that registers destructors for thread_local variables in a list. Upon thread or process exit, the destructors are called in reverse order in which they were added.
As suggested, this only became available in glibc 2.18. From the [2.18 release notes](https://sourceware.org/legacy-ml/libc-alpha/2013-08/msg00160.html):
> * Add support for calling C++11 thread_local object destructors on thread
and program exit. This needs compiler support for offloading C++11
destructor calls to glibc.
ACKs for top commit:
hebasto:
ACK b155fcda51
Tree-SHA512: 5b9567e4a70598a4b0b91956f44ae0d93091db17c84cbf9817dac6cfa992c97d3438a8b1bb66644c74891f2149e44984daed445d22de93ca8858c5b0eabefb40
3c21db7b78 [test] add 8 error messages to feature_segwit and change version to type (Gloria Zhao)
Pull request description:
Followup to [this](https://github.com/bitcoin/bitcoin/pull/15169/files#r303673472) comment on functional test feature_segwit.py verifying that unsigned witness transactions are invalid.
(1) Changes 8 error messages from "mandatory-script-verify-flag" to "non-mandatory-script-verify-flag" and with more specific error messages.
(2) Edits comments that incorrectly describe the test, namely that the `v` variable corresponds to using P2WSH versus P2WPKH, not witness versions.
ACKs for top commit:
MarcoFalke:
ACK 3c21db7b78🍾
Tree-SHA512: 3734ea3762667636c4fb20f5285634ab94d6b3527b7390fcc5e41b4582829dfe0099beabeaed42098613d168ede3385a6ffcd73989d1fa9dbd18004f5e9cf083
P2WPKH witness program without signature -> throws "hash mismatch" error
P2WSH witness program without signature -> throws "empty witness" error
same errors for P2SH_P2WPKH and P2SH_P2WSH respectively when passed redeemScript but no signature
P2SH_P2WPKH and P2SH_P2WSH with no signature fail with "Operation not valid with current stack size" when not signed due to missing input
change VER to TYPE and constants WIT_V0 to P2WPKH=0 and WIT_V1 to P2WSH=1
fac2fc4dd8 test: Increase debugging to hunt down mempool_reorg intermittent failure (MarcoFalke)
Pull request description:
Top commit has no ACKs.
Tree-SHA512: 4094b44afaa623e58b69f8d0332e60f0150b9ae2fd8bb265210d85546d887672ab8a3435cd9b086be14f69ab5b17e0f9fae06bd8aec1e7947ca766dd72b577c4
c743718558 test: add further BIP37 size limit checks to p2p_filter.py (Sebastian Falbesoner)
Pull request description:
This is a follow-up PR to #18628. In addition to the hash-functions limit test introduced with commit fa4c29bc1d, it adds checks for the following size limits as defined in [BIP37](https://github.com/bitcoin/bips/blob/master/bip-0037.mediawiki):
ad message type `filterload`:
> The filter itself is simply a bit field of arbitrary byte-aligned size. The maximum size is **36,000 bytes**.
ad message type `filteradd`:
> The data field must be smaller than or equal to **520 bytes** in size (the maximum size of any potentially matched object).
Also introduces new constants for the limits (or reuses the max script size constant in case for the `filteradd` limit).
Also fixes#18711 by changing the misbehaviour check on "filteradd without filterset" (introduced with #18544) below to also use the more commonly used `assert_debug_log` method.
ACKs for top commit:
MarcoFalke:
ACK c743718558
robot-visions:
ACK c743718558
jonasschnelli:
utACK c743718558. Seems to fix it: https://bitcoinbuilds.org/index.php?build=2524
Tree-SHA512: a03e7639263eb36a381922afb4e1d0ed2ae286f2ad2e7bbd922509a043ddf6cfd08747e01d54d29bfb8f54b66908f653974b9c347e4ca4f43332b586778893be
b68e717967 build: Set libevent minimum version to 2.0.21 (Hennadii Stepanov)
Pull request description:
The non-`pkg-config` path is ignored as there is a hope to get rid of all of them in #18307.
As xenial has [libevent 2.0.21](https://packages.ubuntu.com/xenial-updates/libevent-2.0-5) only, the default bionic Docker image is used in the _"[no depends, only system libs, sanitizers: thread (TSan), no wallet]"_ CI test.
ACKs for top commit:
theStack:
utACK b68e717967
laanwj:
ACK b68e717967
Tree-SHA512: 9825c42aeb166165e99fe5eaf74dbb47c2b51aecdbe53c5ae949fe126e1b8e8b6fe8d228fdde4e8daa4243e5907954202f42eb23c71629e4b2b92a7d4eb892e4
a1d5b12ec0 Merge getreceivedby tally into GetReceived function (Andrew Toth)
Pull request description:
This PR merges the tally code of `getreceivedbyaddress` and `getreceivedbylabel` into a single function `GetReceived`. This reduces repeated code and makes it similar to `listreceivedbyaddress` and `listreceivedbylabel`, which use the function `ListReceived`. It will also make the change in #14707 simpler and easier to review.
ACKs for top commit:
theStack:
re-ACK a1d5b12ec0
meshcollider:
utACK a1d5b12ec0
Tree-SHA512: 43d9cd92f7c2c6a8b9c7509aa85a9b9233a6cfec1c43a9062e3bdfb83515413d1feafa8938c828351278ba22bd31c47e62ab5341e4bddc2493103b094d73b047
709998467e rpc: doc: Fix and extend getblockstats examples (Adam Soltys)
Pull request description:
This pull fixes the example curl command for `getblockstats` which doesn't work as is because it's missing a comma between the params and has single quotes around the second parameter.
It also adds an additional example of getting block stats by hash by using a known workaround (#15412) to get bitcoin-cli to treat the hash parameter as JSON instead of a string since there is ongoing deliberation about how or whether to fix the root issue (#15448).
ACKs for top commit:
theStack:
ACK 709998467e
Tree-SHA512: 84a5b7f449f06fff785bc0afbc1a7dfd55454bc76c52a8945e91556f87f3edfdc5a1780faab8fcfd6c415b734295b7c67d2e04ba7b6cfa91a77758af5dda53ae
a9ecbdfcaa test: add more inactive filter tests to p2p_filter.py (Sebastian Falbesoner)
5eae034996 net: limit BIP37 filter lifespan (active between 'filterload' and 'filterclear') (Sebastian Falbesoner)
Pull request description:
This PR fixes https://github.com/bitcoin/bitcoin/issues/18483. On the master branch, there is currently _always_ a BIP37 filter set for every peer: if not a specific filter is set through a `filterload` message, a default match-everything filter is instanciated and pointed to via the `CBloomFilter` default constructor; that happens both initially, when the containing structure `TxRelay` is constructed:
c0b389b335/src/net.h (L812)
and after a loaded filter is removed again through a `filterclear` message:
c0b389b335/src/net_processing.cpp (L3201)
The behaviour was introduced by commit 37c6389c5a (an intentional covert fix for [CVE-2013-5700](https://github.com/bitcoin/bitcoin/pull/18515), according to gmaxwell).
This default match-everything filter leads to some unintended side-effects:
1. `getdata` request for filtered blocks (i.e. type `MSG_FILTERED_BLOCK`) are always responded to with `merkleblock`s, even if no filter was set by the peer, see issue #18483 (strictly speaking, this is a violation of BIP37) c0b389b335/src/net_processing.cpp (L1504-L1507)
2. if a peer sends a `filteradd` message without having loaded a filter via `filterload` before, the intended increasing of the banscore never happens (triggered if `bad` is set to true, a few lines below) c0b389b335/src/net_processing.cpp (L3182-L3186)
This PR basically activates the `else`-branch code paths for all checks of `pfilter` again (on the master branch, they are dead code) by limiting the pointer's lifespan: instead of always having a filter set, the `pfilter` is only pointing to a `CBloomFilter`-instance after receiving a `filterload` message and the instance is destroyed again (and the pointer nullified) after receiving a `filterclear` message.
Here is a before/after comparison in behaviour:
| code part / scenario | master branch | PR branch |
| --------------------------------------------- | ---------------------------------- | ---------------------------------------------------- |
| `getdata` processing for `MSG_FILTERED_BLOCK` | always responds with `merkleblock` | only responds if filter was set via `filterload` |
| `filteradd` processing, no filter was loaded | nothing | peer's banscore increases by 100 (i.e. disconnect) |
On the other code parts where `pfilter` is checked there is no change in the logic behaviour (except that `CBloomFilter::IsRelevantAndUpdate()` is unnecessarily called and immediately returned in the master branch).
Note that the default constructor of `CBloomFilter` is only used for deserializing the received `filterload` message and nowhere else. The PR also contains a functional test checking that sending `getdata` for filtered blocks is ignored by the node if no bloom filter is set.
ACKs for top commit:
MarcoFalke:
re-ACK a9ecbdfcaa, only change is in test code 🕙
Tree-SHA512: 1a656a6d74ccaf628e7fdca063ba63fbab2089e0b6d0a11be9bbd387c2ee6d3230706ff8ffc1a55711481df3d4547137dd7c9d9184d89eaa43ade4927792d0b6
92fe537cf7 test: fix intermittent race condition in interface_bitcoin_cli.py (Jon Atack)
c648e636b2 test: add wait_for_cookie_credentials() to test framework (Jon Atack)
Pull request description:
This PR adds a `wait_for_cookie_credentials()` method to the test framework and calls it before the `-rpcwait` tests, to avoid an intermittent race condition on the CI run with Valgrind where the cookie file isn't written yet when the CLI call with `-rpcwait` arrives to `get_auth_cookie()`.
To reproduce/test, build with
```diff
diff --git a/src/httprpc.cpp b/src/httprpc.cpp
index 60c4d06f12..3dd06c4758 100644
--- a/src/httprpc.cpp
+++ b/src/httprpc.cpp
@@ -291,6 +291,7 @@ static bool InitRPCAuthentication()
bool StartHTTPRPC()
{
LogPrint(BCLog::RPC, "Starting HTTP RPC server\n");
+ UninterruptibleSleep(std::chrono::seconds{11});
if (!InitRPCAuthentication())
```
then run the test normally and with valgrind
```
test/functional/interface_bitcoin_cli.py -l debug
valgrind test/functional/interface_bitcoin_cli.py -l debug
```
Thanks to Marco Falke for all the help. Closes#18684.
Top commit has no ACKs.
Tree-SHA512: 1b76635b5b1d6b05138affef7ab788aa3bc3fc75b0c69ba778ecdf81063cfe02a8dd7667cfd63a6c6e19b2dac47d7a8b755e334d8af5c0ab9d4026808ee96c83
by calling wait_for_cookie_credentials() to ensure the cookie file is written
and auth credentials available for testing the CLI -rpcwait option before the
RPC connection is up.
fc289b7898 wallet: Refactor WalletRescanReserver to use wallet reference (João Barbosa)
Pull request description:
Simple refactor to `WalletRescanReserver` to use wallet reference instead of pointer.
Complements #18259.
ACKs for top commit:
MarcoFalke:
ACK fc289b7898
Tree-SHA512: b03e33f2d9df2870436aa3284137fd022dd89ea96a1b170fa27f8685ad4f986e6c4ba5975a84966c30d18430a4014d7d8740a1dff2f985c9ef8226ed18e69db9
9df32e820d scripted-diff: test: replace command with msgtype (Sebastian Falbesoner)
Pull request description:
This is a follow-up PR to https://github.com/bitcoin/bitcoin/pull/18533, which changed the naming of `strCommand` to `msg_type` in the network processing code. The same approach is done here for the function test framework, to get rid of the wrong "command" terminology for network mesage types. (Commands are usually used in the CLI or RPC context, so using the same name in the network message context would only be confusing.)
The commit was created through the following steps:
1. search for all occurences of the string "command" within the folder `test/functional`
```git grep -i command test/functional > command_finds```
2. manually sort out all false-positives, i.e. occurences of "command" which describe commands in the correct sense (mostly CLI or RPC related, also some with Socks5)
3. put the remaining occurences into a scripted-diff (a quite simple one, actually) that renames "command" to "msgtype" in the concerned files.
The name `msgtype` was intentionally chosen without the underscore `_` as classes beginning with `msg_` define concrete types of messages.
ACKs for top commit:
MarcoFalke:
ACK 9df32e820d . Makes sense that tests use the same naming as Bitcoin Core. See `NetMsgType` here: https://doxygen.bitcoincore.org/namespace_net_msg_type.html
Tree-SHA512: cd0ee08a382910b7f10ce583acdaf4f8a39f9ba4a22434a914415727eedd98bac538de9bf6633574d5eb86f62558bc8dcb638a3289d99b04f8481f34e7a9a0c7
