c295cba5a2 depends: zeromq: disable draft classes and methods (fanquake)
0072237b9e depends: xproto: configure flags cleanup (fanquake)
6a8ada3a4f depends: qrencode: configure flags cleanup (fanquake)
86beb8cdc4 depends: fontconfig: configure flags cleanup (fanquake)
e656d95ec7 depends: libxcb: configure flags cleanup (fanquake)
e439388b35 depends: libXau: configure flags cleanup (fanquake)
Pull request description:
Related to #16354.
This PR adds additional configure flags to packages in depends to explicitly disable features we aren't using; similar to #16183. It also fixes passing `--without-tools` to `qrencode`.
I've added `--disable-drafts` to `zeromq`:
```bash
Build and install draft classes and methods [default=yes]
```
I'm not entirely sure how far we want to take this. i.e in the `zeromq` package we explicitly pass `--without-libsodium`, even though it's disabled by default.
Do we also want to explicitly pass all the other `--without` flags? :
```bash
--with-libgssapi_krb5 require libzmq build with libgssapi_krb5
[default=no]
--with-libsodium use libsodium instead of built-in tweetnacl
[default=no]
--with-pgm build libzmq with PGM extension. Requires pkg-config
[default=no]
--with-norm build libzmq with NORM protocol extension,
optionally specifying norm path [default=no]
--with-vmci build libzmq with VMCI transport [default=no]
```
ACKs for top commit:
dongcarl:
ACK c295cba5a2
Tree-SHA512: df6d38b863b4008ed2cb06c97eb0e21eaa4b5fde552876065ba7f3c87bf6e372e5b954a51bf3fde2151cfb6d2c022227d34337fc6e50ce0caa1d518abbd2412a
74e3876989 Release notes: add previously undocumented changes (David A. Harding)
7e1634a927 Release notes: edit previously-detached notes (David A. Harding)
e7415a5a95 Doc: move detached release notes into release-notes.md (David A. Harding)
Pull request description:
Merges in detached release notes, edits each change down to a single paragraph bullet point (or, in a couple cases, two individual bullet points in separate sections each with a single paragraph). Adds notes for some undocumented changes I found reviewing `git log --merges`. Also tries something new: adds the PR number(s) after each entry to make it easier for both reviewers and end-user readers to look up the details behind each change. (If the PR numbers are unwanted, they're easy to remove either in this PR or later in the release process.)
I also checked the 0.18 branch but I didn't find anything in the current release notes that had been backported.
A particular focus in my editing was trying to keep things concise, particularly by pointing to RPC documentation when available (or upcoming, as in #16629). I do suspect that one downside of detached notes is that people write longer summaries than they would if they knew there were already 300 other lines of release notes. :-)
The first commit only moves notes, puts them in bullet form, adjusts indentation appropriately, and drops unneeded headers. It can be reviewed with `git diff --color-moved=dimmed-zebra` for a little bit of a speedup, but unfortunately I wasn't smart enough to split my copy/pasting and line wrapping into separate commits, so it's not a transparently move-only change.
ACKs for top commit:
fanquake:
ACK 74e3876989 - Thanks for doing this. It's nice to have a gauge of what's going to be in the `0.19.0` release.
meshcollider:
ACK 74e3876989
Tree-SHA512: 676668765849d5a67520dd8ac49de85ac1bfb5ba2dc09504e75db77d79c7e2c58b5cee16c58591ec575cb3682e630231baba7fd07565d19f8d02243e06fcb9ab
a4765bd77f gui: Ensure tx send error highlight is visible (bpay)
Pull request description:
Rebased and squashed #14956.
> If sending to multiple recipients and one of the recipient fields is malformed, the highlighted field may not be visible due to being scrolled out of view. This results in a confusing lack of error feedback.
> Avoid this problem by ensuring the first field containing an error is scrolled into view when Send is clicked.
You can see the behavior here: https://imgur.com/a/QZG5TQc
How to test:
Add a few recipients and give any of them an invalid address or amount. Scroll the invalid recipient out of view and hit Send. With this change, the GUI will scroll to show the invalid recipient, with master it will not, "hiding" the error.
ACKs for top commit:
jonatack:
Tested ACK a4765bd77f on Linux Debian with Qt 5.11.3. Change is that I had added an inadvertent typo in my make bash alias; fixed.
hebasto:
ACK a4765bd77f, tested on Debian 9.9 with system Qt 5.7.1.
Tree-SHA512: a5653ca44d6d540214bdb424b0b75a06a5872cff41b0cd8cffd9cef99ebf04a17a3652e561139ac75315b39c3347e5f7ae304fa35e14b48bdae4768a416df9b0
We shouldn't rely on this sync call to get an accurate view of txn
state, if a tx conflicts with one in mapTx we are going to update
our wallet dependencies in AddToWalletIfInvolvingMe while conflicting
txn get connected. If it doesn't conflict with one of our dependencies
we are not going to track it anyway.
This is a cleanup, as this SyncTransaction is redundant with the
following one for confirmation which is triggering the MarkConflicted
logic. We keep the loop because set of conflicted txn isn't same as txn
included in block.
Instead of relying on combination of hashBlock and nIndex
values to manage tx in its lifecycle, we introduce 4
status : CONFIRMED, UNCONFIRMED, CONFLICTED, ABANDONED.
hashBlock and nIndex magic values should only be used at
serialization/deserialization for backward-compatibility.
At block disconnection, we know flag txn as UNCONFIRMED where
previously they kept their states until being override by a
block connection or abandontransaction call. This is a change
in behavior for which user may have to call abandon twice
if transaction is disconnected and not accepted back in the mempool.
We assert status transitioning right in AddToWallet. Doing so
flagged a misbehavior in ComputeTimeSmart unit test where same
tx is confirmed twice in different block. To avoid inconsistencies
we unconfirmed tx before new connection in different block. We
also remove a cs_main lock in test, as AddToWallet and its
callees don't rely on locked chain.
In Qt 5.12 and before the QFontMetrics::width() is used and it is
deprecated since Qt 13.0.
In Qt 5.11 the QFontMetrics::horizontalAdvance() was introduced.
If sending to multiple recipients and one of the recipient fields is malformed,
the highlighted field may not be visible due to being scrolled out of view,
leading to a confusing lack of error feedback when clicking Send. To avoid this
problem ensure the first field containing an error is scrolled into view
when Send is clicked.
2483266c59 packages.md: document depends build targets (Russell Yanofsky)
be27161ee4 Clarify need to specify --prefix with depends (Russell Yanofsky)
Pull request description:
There seems to be some confusion about exactly how to use depends, when to pass a prefix to `./configure` etc (see #16367, #16654).
I've cherry-picked two of russ's commits out of #16367, as they are clear stand-alone improvements and we don't have to wait for #16367 to improve the depends documentation.
ACKs for top commit:
Sjors:
utACK 2483266
hebasto:
ACK 2483266c59, I have reviewed the code and it looks OK, I agree it can be merged.
jonasschnelli:
ACK 2483266c59
Tree-SHA512: a198c288248f573519a3b0ef384626b61cc803803280af9a448c28466e3d9949bed0332af6618dac19e81c5a6e9694afa83d976b176fd13c32a6c2c3fea3fc1f
bca388db0d gui: remove unused PlatformStyle::TextColorIcon (fanquake)
Pull request description:
This is unused after #16612.
ACKs for top commit:
emilengler:
Concept ACK bca388d
practicalswift:
ACK bca388db0d -- diff looks correct & less cruft is better
hebasto:
ACK bca388db0d, I have reviewed the code and it looks OK, I agree it can be merge.
jonasschnelli:
utACK bca388db0d
Tree-SHA512: ae5c5cfc02cf2e6f138e75e227320f4c430b93bd529a8ead7ba971ceb5bbc997cdb721c6fd8a46a5a897008f0bc6e519422d6beef240b9d2e28ecf4f8cbbcc6a
f7220e85fd build: update RapidCheck Makefile (Jon Atack)
Pull request description:
- update RapidCheck to the latest commit on master
- install only the extras needed by Bitcoin Core, e.g. ENABLE_BOOST_TEST instead of INSTALL_ALL_EXTRAS
- remove cmake warning by providing `-B` arg:
```
CMake Warning:
No source or binary directory provided. Both will be assumed to be the
same as the current working directory, but note that this warning will
become a fatal error in future CMake releases.
```
Tested with `cd depends && make RAPIDCHECK=1` on Linux Debian.
ACKs for top commit:
fanquake:
ACK f7220e85fd - Tested building using depends. `ENABLE_BOOST_TEST` instead of `INSTALL_ALL_EXTRAS` seems fine.
Tree-SHA512: 8033aa25014a87125cd0aa0885a892649039ef888688f4a855d23af21a350792a4081b181b8222ed293b5070dff39c438e9c3f80294eda0c931076a8153403ea
b9ee63c71b Make descriptor test deterministic (David Reikher)
Pull request description:
This is an improvement to a test, inspired by #14343 - removing non determinism from a test.
The test `descriptor_test` is non-deterministic, as it relies on the `MaybeUseHInsteadOfApostrophy` function which randomly either swaps all apostrophes with 'h' or doesn't at all in a descriptor. This fix makes both cases always run, if an apostrophe is found in a test descriptor.
This does not reduce test coverage but removes the non-determinism.
Additionally, the `MaybeUseHInsteadOfApostrophy` function removed the checksum if found at the end of a descriptor when the apostrophes are swapped by 'h's, since after being swapped the checksum is no longer correct. I instead added re-calculation of the checksum using the `DescriptorChecksum` function, which adds coverage for the case of a descriptors having 'h's instead of apostrophes and a checksum. This was previously lacking.
To achieve this I had to move `DescriptorChecksum` and `PolyMod` out of the anonymous namespace in descriptor.cpp to make `DescriptorChecksum` accessible in descriptor_tests.cpp.
All tests complete successfully (functional as well as unit tests).
ACKs for top commit:
achow101:
Code Review ACK b9ee63c71b
Tree-SHA512: 992c73a6644a07bfe7c72301ee2666f3c4845a012aaedd7a099a05cea8bdac84fa8280b28e44a7856260c00c0be1a6f1b6768f5694c2a22edf4c489e53fec424
faebf62714 rpc: Use Join helper in rpc/util (MarcoFalke)
fa8cd6f9c1 util: Add Join helper to join a list of strings (MarcoFalke)
Pull request description:
We have a lot of enumerations in the code and sometimes those enumerations need to be mentioned in the RPC or command line documentation. Previously, each caller would have a couple of lines inline to join the strings or the joined string is hardcoded in the documentation. A helper to join strings would make code such as https://github.com/bitcoin/bitcoin/pull/16629#discussion_r315852446 less verbose and easier to read.
Also, warnings commonly accumulate in complex RPCs, since a warning doesn't lead to an early return. A helper to join those warnings would make code such as https://github.com/bitcoin/bitcoin/pull/16394/files#r309324997 less verbose and easier to read.
ACKs for top commit:
practicalswift:
ACK faebf62714
Tree-SHA512: 80f2db86a05c63b686f510585c1c631250271a8958fd71fafaac91559ffd2ec25d609bf7d53412ba27f87eff5893ac9dd9c2f296fc0c73581556e1d6a734a36f
153d9dd9ac refactor: replace qLowerBound & qUpperBound with std:: upper_bound & lower_bound (fanquake)
59373e3e94 refactor: replace qSort with std::sort (fanquake)
fea33cbbdf refactor: replace qStableSort with std::stable_sort (fanquake)
Pull request description:
`qStablesort`, `qSort`, `qLowerBound` and `qUpperBound` have been marked as obsolete since at least Qt 5.9: [Obsolete Members for QtAlgorithms](https://doc.qt.io/qt-5.9/qtalgorithms-obsolete.html).
This pull request replaces their usage with the suggested `std::` replacements.
This also removes some warning spam when compiling against newer Qt (5.13.0 via brew):
```bash
CXX qt/libbitcoinqt_a-walletcontroller.o
qt/transactiontablemodel.cpp:96:52: warning: 'qLowerBound<QList<TransactionRecord>::iterator, uint256, TxLessThan>' is deprecated: Use std::lower_bound [-Wdeprecated-declarations]
QList<TransactionRecord>::iterator lower = qLowerBound(
qt/transactiontablemodel.cpp:98:52: warning: 'qUpperBound<QList<TransactionRecord>::iterator, uint256, TxLessThan>' is deprecated: Use std::upper_bound [-Wdeprecated-declarations]
QList<TransactionRecord>::iterator upper = qUpperBound(
```
```bash
CXX qt/libbitcoinqt_a-moc_walletcontroller.o
qt/bantablemodel.cpp:64:13: warning: 'qStableSort<QList<CCombinedBan>::iterator, BannedNodeLessThan>' is deprecated: Use std::stable_sort [-Wdeprecated-declarations]
qStableSort(cachedBanlist.begin(), cachedBanlist.end(), BannedNodeLessThan(sortColumn, sortOrder));
```
```bash
CXX qt/libbitcoinqt_a-sendcoinsentry.o
qt/recentrequeststablemodel.cpp:205:5: warning: 'qSort<QList<RecentRequestEntry>::iterator, RecentRequestEntryLessThan>' is deprecated: Use std::sort [-Wdeprecated-declarations]
qSort(list.begin(), list.end(), RecentRequestEntryLessThan(column, order));
```
ACKs for top commit:
hebasto:
ACK 153d9dd9ac
promag:
ACK 153d9dd9ac.
jonasschnelli:
utACK 153d9dd9ac
Tree-SHA512: 22f7290ed798ce8b0f5f313405377845d4c8e48dc8687be7464e27fff53363b451a40e9e18910a8c3b4b9d4dcc236a366c92e7d171fcb8576c48f149a1886c26
6011c9d72d QA: fix rpc_setban.py race (Jonas Schnelli)
Pull request description:
The new `rpc_setban.py` test failes regularly on CIs due to a race between injecting the ban and testing the log "on the other side".
The problem is, that the test immediately after the `addnode` command on node0 checks for the `dropped (banned)` entry on node1 (without giving some time).
Adding a 2 seconds sleep seems to solve the race (I guess there is no better event-driven delay).
Example of a failed test: https://bitcoinbuilds.org/index.php?ansilog=bf743910-103f-4b54-9a97-960c471061bd.log#l2906
Top commit has no ACKs.
Tree-SHA512: 680f8ea3e5ddb07e93f824f1aeff4a459e25e6c14715a39fc7670e50506d7cf25925348672c5c2d8ba3e1243ccf5effbc2456bcd094fb96868349f8d26e008f1
2dbfb37b40 Fix Char as Bool in interfaces (Jeremy Rubin)
Pull request description:
In a few places in src/wallet/wallet.h, we use a char when semantically we want a bool.
This is kind of an issue because it means we can unserialize the same transaction with different fFromMe flags (as differing chars) and evaluate the following section in wallet/wallet.cpp
```c++
if (wtxIn.fFromMe && wtxIn.fFromMe != wtx.fFromMe)
{
wtx.fFromMe = wtxIn.fFromMe;
fUpdated = true;
}
```
incorrectly (triggering an fUpdated where both fFromMe values represent true, via different chars).
I don't think this is a vulnerability, but it's just a little messy and unsemantic, and could lead to issues with stored wtxIns not being findable in a map by their hash.
The serialize/unserialize code for bool internally uses a char, so it should be safe to make this substitution.
NOTE: Technically, this is a behavior change -- I haven't checked too closely that nowhere is depending on storing information in this char. Theoretically, this could break something because after this change a tx unserialized with such a char would preserve it's value, but now it is converted to a ~true~ canonical bool.
ACKs for top commit:
achow101:
Code review ACK 2dbfb37b40
meshcollider:
Code review ACK 2dbfb37b40
Tree-SHA512: 8c0dc9cf672aa2276c694facbf50febe7456eaa8bf2bd2504f81a61052264b8b30cdb5326e1936893adc3d33504667aee3c7e207a194c71d87b3e7b5fe199c9d
Changed MaybeUseHInsteadOfApostrophy to UseHInsteadOfApostrophe.
This function now always replaces apostrophes with 'h'.
The original Check function was renamed to DoCheck and it's
called with an additional parameter which tells it to either
leave the prv and pub arguments as is or replace the apostrophes
with 'h'. The test runs with apostrophes replaced in prv only,
pub only, prv and pub and without replacement at all. Replacement
of apostrophes in a descriptor and then running DoCheck is conditional
on whether apostrophes are found in that descriptor.
Additionally, instead of dropping the checksum recalculate it
after replacing apostrophes with 'h' in the function UseHInsteadOfApostrophe
using the GetDescriptorChecksum function. That way, this also
introduces an indirect unit test to GetDescriptoChecksum.
c84ff23c6d build: ignore macOS make deploy artefacts & add them to clean-local (fanquake)
Pull request description:
ACKs for top commit:
hebasto:
ACK c84ff23c6d, tested on Linux Mint 19.2:
dongcarl:
ACK c84ff23c6d
Tree-SHA512: 5a6525a5e433a7279807cc02d3f47bb7f4bd5da8142a945b21d3bfa71938cac823569b34b071e21525a817a187fb02a562bb1c5a9761e2019a2b0900a51a8315
17d178fb94 doc: add release-notes for getmempoolentry weight field addition (fanquake)
9c9cc2bd20 qa: Add RPC tests for weight in mempool entry (Daniel Edgecumbe)
54aaa7883c RPC: add weight to mempool entry output (Daniel Edgecumbe)
Pull request description:
Rebase of #14649 (which itself was a rebase of #11256).
Squash the two test related commits, and swapped out `size` usage for `vsize`.
Added a commit with release notes.
ACKs for top commit:
emilengler:
Concept ACK 17d178f
instagibbs:
utACK 17d178fb94
meshcollider:
utACK 17d178fb94
Tree-SHA512: 1d354c9837e0ad0afa40325de9329b9e62688d5eab4d9e1cf9b46d8ae29d08f42d903ab37a41751c2ea8f9034231b21095881b1f5d911cb542b8b06bc85dc7cd
The bitcoin user needs read access to the configuration file, but write
access is not needed. It is not considered best practice to make
configuration directories and files owned by the services reading them.
Rather than making the config dir world-readable, which potentially
leaks RPC credentials, the group of the directory is changed to the one
the service is executed as.
