Boost elimination follow-up: some modernization and code clean-up

See merge request !1500

Boost elimination: chrono, mutex, condition_variable

See merge request !1501

Replace boost::thread with std::thread

See merge request !1497

[lint] Allow std::variant to appear in codebase.

See merge request !1581

Fixed the comment for ContextualCheckTransaction to be more accurate

See merge request !1584

The comment implied that this function is called only for "non-coinbase"
transactions,  however it actually is called for *all* transactions in
ContextualCheckBlock() (including coinbase).

The comment was misleading.  Fixed.

All of the compilers and libraries we now target support this C++17
language feature. The last holdout was our macOS support, but since
we upgraded to 10.14 as the base deployment target in MR !1524,
std::variant is now supported on macOS as well.  As such, we can remove
this linting check from our codebase.

[backport] Replace MakeSpan helper with Span deduction guide

See merge request !1543

[doc] Clarify general expectation and policy on paid merge requests

See merge request !1575

Calin Culianu authored Jun 07, 2022 and freetrader committed Jun 07, 2022

This takes a bunch of changes from Core, heavily modified to work ok with
cmake, and upgrades the statically-linked Qt library we use for release
and/or depends from Qt 5.9.7 to Qt 5.15.3 (latest patched Qt5).
This brings lots of bugfixes and better support for HighDPI displays,
as well as correct OSX dark mode support (yay!).  By now, Qt 5.9.7 is
ancient and we really should be shipping the latest patched Qt with our
static binary releases.

Notes:

- We had to upgrade OpenSSL from 1.0.1 to 1.1.1n because Qt 5.15.3
  requires this library and refuses to work with older OpenSSL
  Ideally the openssl update would be a separate MR but the newer
  openssl refuses to work with our older Qt 5.9.8 !! Argh! So I had to
  roll these into the same MR.
- bison now is required on the host system when building using depends.
- OSX version works but requires clang-8.0.0 and OSX SDK 10.15
  (see: !1524).
  
Test Plan
---------

- build all platform with gitian
- verify ...

[ci] Unpin mkdocs-material dependency after upstream fix

See merge request !1578

[ci] Hotfix for 'pages' job: pin mkdocs-material to 8.2.16

Closes #430

See merge request !1576

[backport] secp256k1: Remove OpenSSL testing support

See merge request !1572

[backport] test: replace hashlib.ripemd160 with an own implementation

Closes #428

See merge request !1571

See discussion in issue #427.  The performance regression was
introduced by commit: 493e7d6d

It can be seen using benchmark: `BanManAddressBan`

The above commit made serialization of V1 addresses slower because it involved
un-needed heap allocations (std::vector) when some code paths could have
simply benefitted from doing everything on the stack (std::array).

After this commit, the performance of the `BanManAddressBan` is back to
levels similar to before 493e7d6d.

[qa] Bump version to 24.1.1 and move 24.1.0 Release Notes to archive

See merge request !1567

[doc] Release Notes for Bitcoin Cash Node v24.1.0

See merge request !1565

[doc] Include section on Merge Requests

See merge request !1564

[rpc] Deprecate setexcessiveblock RPC

Closes #374

See merge request !1566

Unused preprocessor definitions

See merge request !1560

Calin Culianu authored May 28, 2022 and freetrader committed May 28, 2022

This is required for the upgrade to Qt 5.15.3 to work on OSX.
It is a pre-requisite for MR !1523.

We upgrade the clang we use to build the OSX binaries from 6.0.1 -> 8.0.0,
thus bringing improvements to generated code and increased support for C++17.

Additionally, we were forced to update the OSX minimum system version from
10.12 (Sierra) to 10.14 (Mojave). This increases C++17 coverage and is also
required by the new Qt 5.15.3 in !1523. This means that it's possible that
the pre-built app we ship won't run on older macs, however at this point Mojave
is itself 4 versions back, so setting that as the minimum brings some benefits.
Namely.. dark mode will work once we upgrade to Qt 5.15.3, which is a huge
UI win for most users.

Test Plan
---------

- You need to update your downloaded MacOSX SDK from 10.14 -> 10.15
  (see instructions in commit's `doc/gitian-building.md` and `depends/README.md`)
- Add this to `gitian-builder/inputs...

Calin Culianu authored May 27, 2022 and freetrader committed May 27, 2022

This allows us to avoid the expensive MTP check for the May 15, 2022 upgrade
and instead use pindexPrev->nHeight directly which is much faster.
We did this for phonon and other previous upgrades (but not Axion
for "reasons").

Note that we hard-coded the activation height for scalenet to 10,006
since we will reorg this chain anyway in the future bqck to height 10,000.
We chose 10,006 because other nodes may be relying on MTP time to activate
upgrade8 still, and we anticipate that after a reorg height 10,007 will
be the first block to accept the new rules (the activation height is always
1 less than the first fork block for new rules).

It's safe to retroactively do this to scalenet because upgrade8 actually
was a "reverse-soft-fork", or a loosening of consensus rules. So the scalenet
chain is valid using the new rules because it was valid under the older,
stricter rules. To keep things simple we simply retroactively pretend that
scalenet always allowed native introspection and other upgrade opcodes after
height 10,006.

We also removed the `-upgrade8activationtime` flag from BCHN.

This means we had to update 2 functional tests to remove the use of the
upgrade8activationtime flag (which simply tested for block rejection
pre-activation).

These tests were good to have but since the new rules are now active on
the real chain, and since we had to remove the activation time argument,
the two minor functional tests cases relying on this argument were removed.

Test Plan
---------

- Review the hard-coded heights are 1 before the fork block (activation is
  weird and we always hard-code the height that is 1-before the fork block!)
- `ninja all check-all`
- Optional: Run a full node on mainnet and sync it from scratch
  (can be done before release)
- Optional2: Run a full scalenet node off this MR and sync it from scratch
  to be paranoid

File permissions

See merge request !1557

Summary
---

This is in preparation for the upcoming release.  It was decided in
Slack by maintainers that a minor version bump was warranted on account
of this version containing a significant update to a major subsystem:
the network subsystem was updated to support torv3.

Test Plan
---

`ninja clean && ninja && src/bitcoind --version` (verify version is updated)

This data is used for UI purposes only to estimate blockchain
download/IBD progress. It is obtained by looking at the debug.log for
the respective chain and looking at the latest UpdateTip line to get the
tx count and block time.  The estimated TPS was calculated by taking:

`(last_txn_count - this_txn_count) / (last_time - this_time)`

This adds a checkpoint for mainnet, testnet3, and testnet4 on the
activation block of the May 15, 2022 upgrade (that is, the first block
mined under new rules).

Summary
---

This is a code quality commit -- it is inappropriate to use class uint256
as if it were a C-style array of uint8_t (even though the underlying
representation is indeed as such).  While the code in question is not
technically wrong as per the C++ standard, it is unsafe in case we ever
decide to change the representation of class uint256.  It is also
inelegant.

It is best to use the accessor methods .data() or .begin() to access
the underlying uint8_t bytes. This commit does just that.

Test Plan
---

- ninja all check-all

The DAA activated when height = X, but the actual activation block was X
+ 1.  The reason why this is has to do with how activation works (you
activate new rules at height X, but the block following X has the
new rules in place!).

A number of our checkpoints in early days of BCH suffered from off-by-1
errors here (see: MRs !1539 and !1540).  This MR is a follow-up
to those to correct the DAA activation block checkpoint being 1
block too early.

While it doesn't matter much practically, it does makes sense for historical
reasons to fix the checkpoints to be the actual block that was the forking
block for a particular rules change, and not the preceding block which was not
using the new rules yet.  One might imagine someone reading our code and
using these hashes for some other tool in the future and since code is a
form of documentation, it's best to document these historical blocks of
interest precisely.

Test plan:

- ninja all check
- Verify that the proposed changed hashes are on the mainnet and/or
  testnet3 chains respectively, at the specified height (use a block
  explorer or a full node RPC to accomplish this).

This is a follow-up to MR !1539 which corrected the UAHF checkpoint
block for mainnet.  This does similar for testnet3.  The old checkpoints
don't matter much now since we have since added more checkpoints after
them, but it's nice to be correct and consistent.  This is a purely
aesthetic/QA MR.

Test plan:

- ninja check-all
- Use a block explorer or BTC & BCH testnet3 node RPCs to verify that
  the existing hash for height 1155875 is
  `00000000f17c850672894b9a75b63a1e72830bbd5f4c8889b5c1a80e7faef138` on
   both chains (this is wrong as a checkpoint!).
- Use a block explorer or BTC & BCH testnet3 node RPCs to verify that
  the new hash in this MR for height 1155876 is
  `00000000000e38fef93ed9582a7df43815d5c2ba9fd37ef70c9a0ea4a285b8f5` and
  that this block exists only on BCH and not on BTC.

[qa] Correct the checkpoint for first post-UAHF block

See merge request !1539

[doc] Update im_uname public key

Closes #283

See merge request !1538

[doc] Include missing dependency install step

Closes #350

See merge request !1534

[doc] Remove outdated and misleading leveldb docs

Closes #247

See merge request !1536

[doc] Restore and update bitcoin.conf documentation

Closes #264

See merge request !1535

This patch fixes a compile error encountered on newer Ubuntu when building the
(now ancient) Qt 5.9.7.  The issue was observed on Ubuntu 21.10 and 22.04,
where a missing include caused compile errors.

This patch just fixes the error by including <limits>.

See BCHN issue #417

build: Fix boost depends build with glibc >= 2.34

See merge request !1528

[backport] Improvements on ADDR caching

See merge request !1522

Don't start node twice

See merge request cculianu/bitcoin-cash-node!25

[ci] Fix for recent breakage of CI due to a jinja2 update

See merge request !1521