661ac15a4a appveyor: Run functional tests on appveyor (Chun Kuan Lee)
2148c36b6e tests: Make it possible to run functional tests on Windows (Chun Kuan Lee)
Pull request description:
This PR do the following things:
- Make functional tests compatible with Windows
- Print color output in functional tests for Windows 10
- Run util and functional tests on appveyor
- Do not run symlink tests on Windows
Note:
- The wallet_multiwallet.py fail is unrelated to the test framework, it's a bug related to c++ code or maybe dependencies. `bitcoind` would exit with 0xC0000005(Access violation) during shutdown occasionally. Disable this for now.
- Not using `--failfast` because this is still in experimental. We should track if there is any other error.
- Disable ZMQ tests because the python zmq library could cause access violation sometimes.
- Disable `feature_notifications` because Bitcoin Core handles the command in different thread, whicha can cause a race condition.
Tree-SHA512: b76db137d264e62a5c130e1cbca7a2ca002a7a0f4153fa0b92c1ea6c9c09ef0533e11c49bdbd566c472d8ff59f245758feb5e5a6ec6cb6bb66a1c67bab5fa48a
5eb20f81d9 Consistently use ParseHashV to validate hash inputs in rpc (Ben Woosley)
Pull request description:
ParseHashV validates the length and encoding of the string and throws
an informative RPC error on failure, which is as good or better than
these alternative calls.
Note I switched ParseHashV to check string length first, because
IsHex tests that the length is even, and an error like:
"must be of length 64 (not 63, for X)" is much more informative than
"must be hexadecimal string (not X)" in that case.
Split from #13420
Tree-SHA512: f0786b41c0d7793ff76e4b2bb35547873070bbf7561d510029e8edb93f59176277efcd4d183b3185532ea69fc0bbbf3dbe9e19362e8017007ae9d51266cd78ae
1eb9a9b524 [RPC] Remove warning for removed estimatefee RPC (John Newbery)
Pull request description:
The RPC was removed in a previous version, but a warning was
left for users to use the estimatesmartfee RPC. Remove that warning now
that estimatefee has been gone for over one version.
Tree-SHA512: 7fb440a354a5058f9e95930306d7fe0c1cba6563b9a44b7388a17d9e5c3cff42023f5aa1728fd94a1a11249ea4a8615a8a891afe4fa81ae46b61c2aa08e9cc47
30973e9844 [REST] improve performance for JSON calls (Antoine Le Calvez)
Pull request description:
JSON calls do not use the raw data generated for the .bin and .hex calls.
By moving the raw data creation into the .bin and .hex switch branches, JSON calls' performance is improved.
Light benchmarking indicates that fetching 2000 JSON headers is ~25% faster, fetching large JSON blocks is ~4% faster.
Tree-SHA512: 433552c89bac2469d041b48a4a991d5443e4026a3ad7dc5621685386029f22826484218642fa5130c268349a55524ecbc4e30d64c867bd6632e0edd24370cf11
a23a7f60aa wallet: Avoid potential use of unitialized value bnb_used in CWallet::CreateTransaction(...) (practicalswift)
Pull request description:
Avoid use of uninitialized value `bnb_used` in `CWallet::CreateTransaction(...)`.
Tree-SHA512: 22faf0711ae35af44d9a0ab7f251bc01661ac88b40ad7b0a87a510427b46bbc8caf16868cab2e0a05e7d8518e93ce666d6bd1d48d3707d37bab2c0fb56a0a4a2
JSON calls do not use the raw data generated for
the .bin and .hex calls. By moving the raw data
creation into the .bin and .hex switch branches,
JSON calls become faster.
addwitnessaddress is deprecated. Remove the call to that RPC from
wallet_dump.py and improve testing of all types of address (legacy,
p2sh-segwit and bech32)
The RPC was removed in a previous version, but a warning was
left for users to use the estimatesmartfee RPC. Remove that warning now
that estimatefee has been gone for over one version.
fa910e4301 init: Remove deprecated args from hidden args (MarcoFalke)
Pull request description:
The args have been deprecated since 0.17 (maybe longer) and since we reject unknown args, there is no need to add deprecated args to the list of hidden args and then hand-craft an error message if a user provides them.
Tree-SHA512: 3a3191439ab0d7969fb72801d097bd86998524f84b3819380224f746cbe4b0f57beec1ad34744424f6587038035b0ddf418ad13171a8d9c3b97b4f3b7b3222a3
52beb9ed88 Add autogen.sh in ARM Cross-compilation (Walter)
Pull request description:
autogen for the config files was missing.
Tree-SHA512: 90e5fdc3392b64ff239b8f359a94a39a3e992640a67ad7ed80a1a5d765aabf95e4dc2f3f4af5ee2aefbe1d3effd7cef35a57b0fad0eaaef8346a07e460538e62
b6718e373e tests: Use MakeUnique to construct objects owned by unique_ptrs (practicalswift)
Pull request description:
A subset of #14211 ("Use MakeUnique to construct objects owned by unique_ptrs") as suggested by @MarcoFalke in https://github.com/bitcoin/bitcoin/pull/14211#issuecomment-423324019.
Use `MakeUnique` to construct objects owned by `unique_ptr`s.
Rationale:
* `MakeUnique` ensures exception safety in complex expressions.
* `MakeUnique` gives a more concise statement of the construction.
Tree-SHA512: 1228ae6ce7beb178d79142c4e936b728178ccaa8aa35c6d8feeb33d1a667abfdd010c59996a9d833594611e913877ce5794e75953d11d9b1fdbac04aa491d9cf
3ccfa34b32 convert C-style (void) parameter lists to C++ style () (Arvid Norberg)
Pull request description:
In C, an empty parameter list, `()`, means the function takes any arguments, and `(void)` means the function does not take any parameters.
In C++, an empty parameter list means the function does not take any parameters.
So, C++ still supports `(void)` parameter lists with the same semantics, why change to `()`?
1. removing the redundant `void` improves signal-to-noise ratio of the code
2. using `(void)` exposes a rare inconsistency in that a template taking a template `(T)` parameter list, cannot be instantiated with `T=void`
Tree-SHA512: be2897b6c5e474873aa878ed6bac098382cd21866aec33752fe40b089a6331aa6263cae749aba1b4a41e8467f1a47086d32eb74abaf09927fd5a2f44a4b2109a
a2eb6f5405 [rpc] Add getnodeaddresses RPC command (chris-belcher)
Pull request description:
Implements issue https://github.com/bitcoin/bitcoin/issues/9463
New getnodeaddresses call gives access via RPC to the peers known by the node. It may be useful for bitcoin wallets to broadcast their transactions over tor for improved privacy without using the centralized DNS seeds. getnodeaddresses is very similar to the getaddr p2p method.
Please advise me on the best approach for writing an automated test. By my reading the getaddr p2p method also isn't really tested.
Tree-SHA512: ad03abf518847476495b76a2f5394b8030aa86654429167fa618e21460abb505c10ef9817ec1b80472320d41d0aff5dc94a8efce023aaaaf5e81386aa92b852b
fd5c95cc4e doc: Add historical release notes for 0.16.3 (Wladimir J. van der Laan)
Pull request description:
Release notes for 0.16.3 have been kind of rushed, feel free to make suggestions here.
Tree-SHA512: 9e6abb6cdd152595c9264f347b88adb1404dee4fad9b7e8e540948dbe9438ec0fc68f8cd4f6dbed1837897662277d96ff570c04e0d7de5d59d45bf965019387f
9b4a36effc [qa] Test for duplicate inputs within a transaction (Suhas Daftuar)
b8f801964f Fix crash bug with duplicate inputs within a transaction (Suhas Daftuar)
Pull request description:
Tree-SHA512: 8c7ea34c7fa44188d86c04a690a7cbf8e9deda71ab1f7ca6d11de1f2abb3dd7222627071f86d0d39689a8b302ba9af142f0202466a67e30cd54aed3a08d4eb14
New getnodeaddresses call gives access via RPC to the peers known by
the node. It may be useful for bitcoin wallets to broadcast their
transactions over tor for improved privacy without using the
centralized DNS seeds. getnodeaddresses is very similar to the getaddr
p2p method.
Tests the new rpc call by feeding IP address to a test node via the p2p
protocol, then obtaining someone of those addresses with
getnodeaddresses and checking that they are a subset.
0ca4c8b3c6 Changed functional tests which do not require wallets to run without (sanket1729)
Pull request description:
Addresses #14216 . Changed Changed `get_deterministic_priv_key()` to return named tuple`(address, key)`
I have tried to be exhaustive as possible in maximum coverage for non-wallet mode without affecting any coverage for wallet mode.
However, I could not check the tests in wallet mode because of timeout issues. Hopefully, travis job checks those.
Tests `feature_block.py`, `feature_logging.py` and `feature_reindex.py` were skipping despite having no direct dependency on any wallet functions. So, I have also disabled the `skip_test_no_wallet()` for those files too.
Tree-SHA512: 8f84bd8400a732d4266c7518d5cbcf1eb761f623a64a74849e0470142c8ef22cb75364474ddae75d9213c3d16659a52917b5ed979a313695da6abd16c4fd7445
f1bd03eb01 [depends, zmq, doc] upgrade zeromq to 4.2.5 and avoid deprecated zeromq api functions (mruddy)
Pull request description:
Upgrade the ZeroMQ dependency from version 4.2.3 to the latest stable version 4.2.5.
This PR Follows the lead of https://github.com/bitcoin/bitcoin/pull/11986.
I upgraded both patch files to correspond to the version `4.2.5` libzmq files.
I assume doing so is still necessary and correct.
Without updating the patch line numbers, things appear to work, but you get extra log messages while building `depends` because things don't exactly match, e.g.:
```
/bitcoin/depends> make zeromq
Extracting zeromq...
/bitcoin/depends/sources/zeromq-4.2.5.tar.gz: OK
Preprocessing zeromq...
patching file src/windows.hpp
Hunk #1 succeeded at 58 (offset 3 lines).
patching file src/thread.cpp
Hunk #1 succeeded at 307 with fuzz 2 (offset 87 lines).
Hunk #2 succeeded at 323 with fuzz 2 (offset 90 lines).
```
Updating the patches seemed cleaner, so I did it. Note that libzmq had some whitespace changes, so that's why the updated patches do too.
More info: https://github.com/zeromq/libzmq/releases/tag/v4.2.5
tags: libzmq, zmq, 0mq
Tree-SHA512: 78659dd276b5311e40634b1bbebb802ddd6b69662ba3c84995ef1e3795c49a78b1635112c7fd72a405ea36e2cc3bdeb84e6d00d4e491a349bba1dafff50e2fa5
be54f42e5f use integer division instead of double conversion and multiplication for computing amounts (Arvid Norberg)
Pull request description:
use integer division instead of double conversion and multiplication for computing amounts. This will most likely generate identical code.
My main argument in favour of this change is one of purity, that we should not rely on implicit conversion from `CAmount` -> `double` and back again. Today this implicit conversion can happen because `CAmount` is just a typedef to `int64_t`. However, I envision a future where `CAmount` is a proper type that does not allow suspicious implicit conversions like these.
Tree-SHA512: a70966623ac6e82410ac94d26cf44e2b7b7a4dbaa514d68ae1f0369aaee1bc2851d05a5e365291b005fe0941428e6139dc62bcfdd0b2f66720706fefe0eb92f1
d641c29a5a travis: Run feature_dbcrash functional tests in cron job (Chun Kuan Lee)
c0d947d725 tests: Reorder tests and move most of extended tests up to normal tests (Chun Kuan Lee)
Pull request description:
The travis should run almost all jobs unless it takes really long time, however it does not take too long for now. So it's time for moving it to normal job.
(The test sort is to see how many conflict will this cause, will drop it if there are too many)
The first commit can be reviewed by `git diff --color-moved=plain`
Tree-SHA512: db6bd5b1f19de2f729012adda6ed00ca989071fd40a20710c0ff2579b5bd008edcf7421c1ad56d5f0752354e7df408f58351129d35a1ab7f4a6caa9d315df2ec
