v30.3-roots.1 · Commit 12: [PATCH 12/68] release: add pre-tag artifact rehearsal
AGENTS.md
Before: /dev/null · After: AGENTS.md
File metadata
new file mode 100644 index 0000000000000000000000000000000000000000..a593ae772cc2b58b690cc1a406bce164d8abca91
Download original patch · Release notes
Diff rows 1–400 of 472. Line numbers refer to the original before/after files.
+ Added− RemovedUnmarked lines are context.
Scroll the code area horizontally to read long lines. The two line-number columns refer to before and after.
AGENTS.md
ADDED
|
@@ -0,0 +1,472 @@
|
|
|
1
|
+
# AGENTS.md
|
|
2
|
+
|
|
3
|
+
This file is the operational guide for AI agents working in this repository. It
|
|
4
|
+
applies to the whole tree unless a more specific `AGENTS.md` is added below it.
|
|
5
|
+
|
|
6
|
+
## Project identity and non-negotiable invariants
|
|
7
|
+
|
|
8
|
+
This repository is **Bitcoin Roots**, not an unmodified Bitcoin Core or Bitcoin
|
|
9
|
+
Knots checkout. It is a C++20 full-node, wallet, GUI, and utility suite based directly on
|
|
10
|
+
Bitcoin Core v30.3, with selected historical policy features inherited from Bitcoin Knots
|
|
11
|
+
`29.3.knots20260507`.
|
|
12
|
+
|
|
13
|
+
Read `README.md` before changing behavior. In particular:
|
|
14
|
+
|
|
15
|
+
- Bitcoin Roots stays compatible with Bitcoin Core consensus.
|
|
16
|
+
- Conservative transaction relay and mempool policy is intentional and
|
|
17
|
+
configurable. A transaction rejected by local policy can still be valid by
|
|
18
|
+
consensus; blocks containing it must remain acceptable.
|
|
19
|
+
- This revision deliberately does **not** enforce RDTS/BIP110 consensus rules.
|
|
20
|
+
Do not add, imply, or silently enable those rules, and do not confuse the
|
|
21
|
+
startup warning/consent code with consensus enforcement.
|
|
22
|
+
- Consensus, validation, serialization, wallet, networking, and cryptographic
|
|
23
|
+
changes are security-critical. Make the smallest justified change and add
|
|
24
|
+
regression coverage. Never “clean up” consensus-adjacent code incidentally.
|
|
25
|
+
|
|
26
|
+
Much of `doc/`, comments, scripts, and test output still uses the inherited
|
|
27
|
+
names “Bitcoin Core” or “Bitcoin Knots.” Do not mass-rebrand this text. Determine
|
|
28
|
+
whether a name is a historical/upstream reference, a protocol compatibility
|
|
29
|
+
identifier, or user-facing Roots branding before changing it. The current
|
|
30
|
+
`doc/release-notes.md` is inherited and contains RDTS migration language that is
|
|
31
|
+
not an authoritative statement of Roots policy. For fork identity and policy,
|
|
32
|
+
prefer, in order: current source behavior and tests, root `README.md`, current
|
|
33
|
+
build configuration, then inherited documentation.
|
|
34
|
+
|
|
35
|
+
## Repository map
|
|
36
|
+
|
|
37
|
+
- `src/`: production C++ and internal libraries.
|
|
38
|
+
- `bitcoind.cpp`, `bitcoin-cli.cpp`, `bitcoin-tx.cpp`, `bitcoin-util.cpp`, and
|
|
39
|
+
`bitcoin-wallet.cpp` are command-line entry points.
|
|
40
|
+
- `qt/` contains the `bitcoin-qt` GUI and GUI tests.
|
|
41
|
+
- `node/` owns P2P/RPC server orchestration and node-specific behavior.
|
|
42
|
+
- `kernel/` is the validation/consensus engine being separated from the node.
|
|
43
|
+
- `consensus/`, `script/`, `primitives/`, and much of `validation.cpp` are
|
|
44
|
+
consensus-sensitive.
|
|
45
|
+
- `policy/` and mempool code define local admission, relay, and mining policy;
|
|
46
|
+
policy must not leak into block-validity decisions.
|
|
47
|
+
- `wallet/` contains wallet databases, coin selection, signing, and wallet RPC.
|
|
48
|
+
- `rpc/` contains non-wallet RPC implementations; `interfaces/` contains
|
|
49
|
+
abstract boundaries between node, wallet, and GUI.
|
|
50
|
+
- `common/` is higher-level shared application code; `util/` is lower-level,
|
|
51
|
+
broadly reusable platform/support code; `crypto/` is standalone primitives.
|
|
52
|
+
- `index/`, `zmq/`, `ipc/`, `init/`, and `logging/` are feature-focused areas.
|
|
53
|
+
- `test/`, `wallet/test/`, and `qt/test/` contain C++ unit tests.
|
|
54
|
+
- `test/functional/`: Python integration tests that launch isolated regtest
|
|
55
|
+
nodes and drive RPC/P2P behavior. `test_framework/` is their shared harness.
|
|
56
|
+
- `test/lint/`: repository lint checks; its runner is a Rust crate.
|
|
57
|
+
- `test/fuzz/` and `src/test/fuzz/`: fuzz runner and C++ fuzz targets.
|
|
58
|
+
- `cmake/`, root `CMakeLists.txt`, and `src/**/CMakeLists.txt`: the primary build
|
|
59
|
+
system. This tree no longer uses Autotools for the main build.
|
|
60
|
+
- `depends/`: reproducible pinned third-party dependency builder, used for
|
|
61
|
+
release and cross builds.
|
|
62
|
+
- `ci/` and `.github/workflows/`: canonical CI configurations. CI intentionally
|
|
63
|
+
exercises different feature/compiler/platform combinations.
|
|
64
|
+
- `contrib/`: developer, packaging, tracing, deployment, and maintenance tools.
|
|
65
|
+
- `doc/`: build, runtime, interface, design, and contributor documentation.
|
|
66
|
+
- `share/`: example configuration, desktop integration, icons, and RPC auth tool.
|
|
67
|
+
|
|
68
|
+
Embedded `src/secp256k1`, `src/leveldb`, `src/crc32c`, `src/crypto/ctaes`, and
|
|
69
|
+
`src/minisketch` are maintained as upstream subtrees. Avoid direct cosmetic or
|
|
70
|
+
cross-cutting edits there. Fix upstream when possible and use
|
|
71
|
+
`test/lint/git-subtree-check.sh` when intentionally updating a subtree.
|
|
72
|
+
|
|
73
|
+
### Library boundaries
|
|
74
|
+
|
|
75
|
+
Follow `doc/design/libraries.md`. The important dependency direction is:
|
|
76
|
+
|
|
77
|
+
- `crypto` has no internal dependencies.
|
|
78
|
+
- `consensus` depends only on `crypto`.
|
|
79
|
+
- `util` depends only on `crypto` and must remain suitable for kernel users.
|
|
80
|
+
- `common` depends only on `util`, `consensus`, and `crypto`.
|
|
81
|
+
- `kernel` depends only on `util`, `consensus`, and `crypto`; internally only
|
|
82
|
+
`node` should depend on `kernel`.
|
|
83
|
+
- Node, wallet, and GUI implementations must not link directly to one another.
|
|
84
|
+
Communicate through abstract types in `src/interfaces/`.
|
|
85
|
+
|
|
86
|
+
When adding production C++ files, add them to the appropriate library target in
|
|
87
|
+
`src/CMakeLists.txt` or the nearest feature `CMakeLists.txt`. When adding a unit
|
|
88
|
+
test file, register it in `src/test/CMakeLists.txt`,
|
|
89
|
+
`src/wallet/test/CMakeLists.txt`, or `src/qt/test/CMakeLists.txt`.
|
|
90
|
+
|
|
91
|
+
## Toolchain and dependencies
|
|
92
|
+
|
|
93
|
+
The minimums documented by this checkout are CMake 3.22, Python 3.10, GCC 11.1
|
|
94
|
+
or Clang 16, Boost 1.74, and libevent 2.1.8. The code is compiled as C++20.
|
|
95
|
+
The default build is `RelWithDebInfo` (`-O2 -g` on typical Unix compilers).
|
|
96
|
+
|
|
97
|
+
Minimal Debian/Ubuntu setup for a headless descriptor-wallet build:
|
|
98
|
+
|
|
99
|
+
```bash
|
|
100
|
+
sudo apt-get install build-essential cmake pkgconf python3 \
|
|
101
|
+
libevent-dev libboost-dev libsqlite3-dev capnproto libcapnp-dev
|
|
102
|
+
```
|
|
103
|
+
|
|
104
|
+
Common optional packages are `python3-zmq`, `libzmq3-dev`,
|
|
105
|
+
`libminiupnpc-dev`, `systemtap-sdt-dev`, `libqrencode-dev`, and Qt development
|
|
106
|
+
packages. See `doc/build-unix.md` and `doc/dependencies.md` for the current full
|
|
107
|
+
list and the macOS/BSD/Windows-specific documents under `doc/` for those hosts.
|
|
108
|
+
|
|
109
|
+
Descriptor wallets use SQLite and are enabled by default. Core 30.3 does not
|
|
110
|
+
build Berkeley DB legacy wallets. Migration tests use pinned previous-release
|
|
111
|
+
fixtures where needed. A node-only build avoids the wallet database with
|
|
112
|
+
`-DENABLE_WALLET=OFF`.
|
|
113
|
+
|
|
114
|
+
Do not install dependencies or run package-manager commands without user
|
|
115
|
+
authorization. Prefer reporting the exact missing dependency. The `depends/`
|
|
116
|
+
system is the source of pinned release dependencies and toolchain files; read
|
|
117
|
+
`depends/README.md` before using it.
|
|
118
|
+
|
|
119
|
+
## Configure and build
|
|
120
|
+
|
|
121
|
+
Always build out of source. The conventional directory is `build/`:
|
|
122
|
+
|
|
123
|
+
```bash
|
|
124
|
+
cmake -B build
|
|
125
|
+
cmake --build build -j 4
|
|
126
|
+
```
|
|
127
|
+
|
|
128
|
+
Use a conservative fixed `-j` value on memory-limited hosts; compilation needs
|
|
129
|
+
roughly 1.5 GiB or more and highly parallel C++ builds can exhaust RAM. Inspect
|
|
130
|
+
available configuration switches with `cmake -B build -LH`.
|
|
131
|
+
|
|
132
|
+
Useful configurations:
|
|
133
|
+
|
|
134
|
+
```bash
|
|
135
|
+
# Debug symbols, assertions, lock-order and lock-contention instrumentation
|
|
136
|
+
cmake -B build-debug -DCMAKE_BUILD_TYPE=Debug
|
|
137
|
+
|
|
138
|
+
# Headless node without wallet or GUI dependencies
|
|
139
|
+
cmake -B build-node -DENABLE_WALLET=OFF -DBUILD_GUI=OFF
|
|
140
|
+
|
|
141
|
+
# GUI (Qt 6 by default; use -DWITH_QT_VERSION=5 for optional Qt 5)
|
|
142
|
+
cmake -B build-gui -DBUILD_GUI=ON
|
|
143
|
+
|
|
144
|
+
# Treat warnings as errors for CI-like compilation
|
|
145
|
+
cmake -B build -DWERROR=ON
|
|
146
|
+
|
|
147
|
+
# Generate compile_commands.json for tooling
|
|
148
|
+
cmake -B build -DCMAKE_EXPORT_COMPILE_COMMANDS=ON
|
|
149
|
+
```
|
|
150
|
+
|
|
151
|
+
The `dev-mode` preset enables nearly every optional target and dependency and
|
|
152
|
+
uses `build_dev_mode/`; it is useful only when those dependencies are installed:
|
|
153
|
+
|
|
154
|
+
```bash
|
|
155
|
+
cmake --preset dev-mode
|
|
156
|
+
cmake --build build_dev_mode -j 4
|
|
157
|
+
```
|
|
158
|
+
|
|
159
|
+
Native Core IPC/libmultiprocess support is retained. Use `-DENABLE_IPC=OFF`
|
|
160
|
+
for a deliberately limited profile when Cap'n Proto is unavailable; do not
|
|
161
|
+
claim that profile proves IPC functionality. Treat the generated CMake summary
|
|
162
|
+
as truth when checking which optional components a preset enables.
|
|
163
|
+
|
|
164
|
+
Build only the target needed during iteration, for example:
|
|
165
|
+
|
|
166
|
+
```bash
|
|
167
|
+
cmake --build build --target bitcoind -j 4
|
|
168
|
+
cmake --build build --target test_bitcoin -j 4
|
|
169
|
+
cmake --build build --target bitcoin-qt -j 4
|
|
170
|
+
```
|
|
171
|
+
|
|
172
|
+
Binaries are emitted under `build/bin/`. Installation is optional:
|
|
173
|
+
`cmake --install build`. Never install system-wide unless explicitly requested.
|
|
174
|
+
|
|
175
|
+
## Running safely
|
|
176
|
+
|
|
177
|
+
Use regtest and a disposable, explicit data directory for development. Never
|
|
178
|
+
point tests or experimental builds at a user's real mainnet wallet/datadir.
|
|
179
|
+
|
|
180
|
+
```bash
|
|
181
|
+
build/bin/bitcoind -regtest -datadir=/absolute/path/to/disposable-dir \
|
|
182
|
+
-daemon -fallbackfee=0.0002
|
|
183
|
+
build/bin/bitcoin-cli -regtest -datadir=/absolute/path/to/disposable-dir \
|
|
184
|
+
getblockchaininfo
|
|
185
|
+
build/bin/bitcoin-cli -regtest -datadir=/absolute/path/to/disposable-dir stop
|
|
186
|
+
```
|
|
187
|
+
|
|
188
|
+
By default the data directory is `~/.bitcoin` on Linux,
|
|
189
|
+
`~/Library/Application Support/Bitcoin` on macOS, and `%LOCALAPPDATA%\Bitcoin`
|
|
190
|
+
on Windows. Chain state is below `regtest/`, `signet/`, `testnet3/`, or
|
|
191
|
+
`testnet4/`; `bitcoin.conf` is normally at the datadir root. Command-line values
|
|
192
|
+
override configuration values. Do not edit `settings.json` while the program is
|
|
193
|
+
running; changes may be ignored or overwritten.
|
|
194
|
+
|
|
195
|
+
`bitcoind -help` is the authority for options in the built revision. RPC help is
|
|
196
|
+
available through `bitcoin-cli help [command]`. Prefer graceful shutdown via
|
|
197
|
+
the `stop` RPC and wait for the process to exit before reusing a datadir.
|
|
198
|
+
|
|
199
|
+
## Testing and validation
|
|
200
|
+
|
|
201
|
+
Choose the narrowest test that proves the changed behavior, then expand based
|
|
202
|
+
on risk. Do not claim success without running the relevant command on the final
|
|
203
|
+
working tree.
|
|
204
|
+
|
|
205
|
+
### C++ unit and utility tests
|
|
206
|
+
|
|
207
|
+
```bash
|
|
208
|
+
ctest --test-dir build --output-on-failure
|
|
209
|
+
build/bin/test_bitcoin --list_content
|
|
210
|
+
build/bin/test_bitcoin -l all -t getarg_tests
|
|
211
|
+
build/bin/test_bitcoin -l all -t getarg_tests/doubledash
|
|
212
|
+
```
|
|
213
|
+
|
|
214
|
+
`ctest` also runs tests from embedded subprojects and utility tests. Use
|
|
215
|
+
`ctest --test-dir build -N` to list registered tests and `-R <regex>` to select
|
|
216
|
+
one. Failure details are saved in `build/Testing/Temporary/LastTest.log`.
|
|
217
|
+
|
|
218
|
+
### Functional tests
|
|
219
|
+
|
|
220
|
+
Configure and build first. Run from the repository root so paths and globs are
|
|
221
|
+
unambiguous:
|
|
222
|
+
|
|
223
|
+
```bash
|
|
224
|
+
build/test/functional/test_runner.py feature_rbf.py
|
|
225
|
+
build/test/functional/test_runner.py --jobs=4
|
|
226
|
+
build/test/functional/test_runner.py --extended
|
|
227
|
+
```
|
|
228
|
+
|
|
229
|
+
An individual generated test script can also be invoked directly, for example
|
|
230
|
+
`build/test/functional/feature_rbf.py`. The default suite is already large;
|
|
231
|
+
`--extended` is larger still. Run the focused test while iterating, then the
|
|
232
|
+
appropriate suite before handoff.
|
|
233
|
+
|
|
234
|
+
Functional tests launch isolated regtest nodes. Their first output line gives
|
|
235
|
+
the temporary directory. On failure inspect:
|
|
236
|
+
|
|
237
|
+
- `<tmpdir>/test_framework.log`
|
|
238
|
+
- `<tmpdir>/nodeN/regtest/debug.log`
|
|
239
|
+
- merged output from
|
|
240
|
+
`build/test/functional/combine_logs.py -c <tmpdir> | less -r`
|
|
241
|
+
|
|
242
|
+
Useful flags are `--nocleanup`, `--tracerpc`, `--jobs=N`, `--tmpdir=...`, and
|
|
243
|
+
the per-test `-l debug`. Run the runner with `-h` for the current set. Do not
|
|
244
|
+
kill all host `bitcoind` processes as a routine cleanup step; that can destroy
|
|
245
|
+
an unrelated user's running node. Identify and terminate only test-owned PIDs.
|
|
246
|
+
|
|
247
|
+
When changing RPC/P2P behavior, add or update a functional test. Reuse helpers
|
|
248
|
+
from `test/functional/test_framework/`, use deterministic assertions, and avoid
|
|
249
|
+
wall-clock sleeps. If the framework already exposes a synchronization helper,
|
|
250
|
+
use it.
|
|
251
|
+
|
|
252
|
+
### Lint and formatting
|
|
253
|
+
|
|
254
|
+
The full lint runner requires Rust plus the tools listed in
|
|
255
|
+
`test/lint/README.md` and `ci/lint/04_install.sh`:
|
|
256
|
+
|
|
257
|
+
```bash
|
|
258
|
+
(cd test/lint/test_runner && cargo fmt && cargo clippy && \
|
|
259
|
+
RUST_BACKTRACE=1 cargo run)
|
|
260
|
+
```
|
|
261
|
+
|
|
262
|
+
Run selected lints with, for example:
|
|
263
|
+
|
|
264
|
+
```bash
|
|
265
|
+
(cd test/lint/test_runner && \
|
|
266
|
+
RUST_BACKTRACE=1 cargo run -- --lint=doc --lint=trailing_whitespace)
|
|
267
|
+
```
|
|
268
|
+
|
|
269
|
+
Format only changed C++ lines, not whole legacy files:
|
|
270
|
+
|
|
271
|
+
```bash
|
|
272
|
+
git diff -U0 --no-color | contrib/devtools/clang-format-diff.py -p1 -i
|
|
273
|
+
```
|
|
274
|
+
|
|
275
|
+
Python formatting/lint behavior is defined by `.style.yapf` and the lint runner
|
|
276
|
+
(which invokes pinned tools such as Ruff). Check `git diff --check` for
|
|
277
|
+
whitespace errors.
|
|
278
|
+
|
|
279
|
+
### Fuzzing, sanitizers, and benchmarks
|
|
280
|
+
|
|
281
|
+
```bash
|
|
282
|
+
cmake --preset=libfuzzer
|
|
283
|
+
cmake --build build_fuzz -j 4
|
|
284
|
+
FUZZ=process_message build_fuzz/bin/fuzz /path/to/corpus
|
|
285
|
+
|
|
286
|
+
cmake -B build-asan -DSANITIZERS=address,undefined
|
|
287
|
+
cmake --build build-asan -j 4
|
|
288
|
+
|
|
289
|
+
cmake -B build-bench -DBUILD_BENCH=ON
|
|
290
|
+
cmake --build build-bench --target bench_bitcoin -j 4
|
|
291
|
+
build-bench/bin/bench_bitcoin
|
|
292
|
+
```
|
|
293
|
+
|
|
294
|
+
The fuzz seed corpora live in the separate `bitcoin-core/qa-assets` repository.
|
|
295
|
+
Do not add arbitrary corpora or crash artifacts to this repository. Address and
|
|
296
|
+
thread sanitizers are mutually incompatible. Use the suppressions in
|
|
297
|
+
`test/sanitizer_suppressions/` as documented in `doc/developer-notes.md`.
|
|
298
|
+
Performance claims should include a focused `bench_bitcoin` comparison and, for
|
|
299
|
+
user-visible paths such as IBD, an end-to-end measurement where practical.
|
|
300
|
+
|
|
301
|
+
## Debugging workflow
|
|
302
|
+
|
|
303
|
+
1. Reproduce with the smallest deterministic unit or functional test.
|
|
304
|
+
2. Record the exact configure flags, executable, arguments, and datadir.
|
|
305
|
+
3. Read the relevant `debug.log`; enable categories with `-debug=<category>`
|
|
306
|
+
and severity with `-loglevel=<level>`. Runtime categories can be toggled via
|
|
307
|
+
the `logging` RPC. Qt `qDebug()` output uses the `qt` category.
|
|
308
|
+
4. Use a Debug or sanitizer build when the symptom warrants it.
|
|
309
|
+
5. Form a concrete root-cause hypothesis before editing and add a regression
|
|
310
|
+
test that fails for the original bug.
|
|
311
|
+
|
|
312
|
+
Common commands:
|
|
313
|
+
|
|
314
|
+
```bash
|
|
315
|
+
gdb --args build/bin/bitcoind -regtest -datadir=/path/to/test-datadir
|
|
316
|
+
gdb build/bin/test_bitcoin
|
|
317
|
+
build/bin/test_bitcoin --catch_system_errors=no -t suite/test
|
|
318
|
+
```
|
|
319
|
+
|
|
320
|
+
A Debug build enables `DEBUG_LOCKORDER` and `DEBUG_LOCKCONTENTION`. Use
|
|
321
|
+
`-debug=lock` or the runtime `logging` RPC for contention records. For a unit
|
|
322
|
+
test that needs a persistent temporary datadir, pass application arguments
|
|
323
|
+
after `--`, for example:
|
|
324
|
+
|
|
325
|
+
```bash
|
|
326
|
+
build/bin/test_bitcoin -t getarg_tests/doubledash -- \
|
|
327
|
+
-testdatadir=/absolute/path/to/test-output -printtoconsole=1
|
|
328
|
+
```
|
|
329
|
+
|
|
330
|
+
Do not “fix” flaky functional tests by increasing sleeps. Check stale test
|
|
331
|
+
processes, port conflicts, the generated cache under `build/test/cache`, and
|
|
332
|
+
the combined logs. Only remove generated test cache after confirming its exact
|
|
333
|
+
path and that no test is using it.
|
|
334
|
+
|
|
335
|
+
## Coding rules that matter most
|
|
336
|
+
|
|
337
|
+
The complete rules are in `doc/developer-notes.md`; functional-test rules are
|
|
338
|
+
in `test/functional/README.md`. Preserve these essentials:
|
|
339
|
+
|
|
340
|
+
- Four spaces, no tabs (except Makefile recipes), LF endings, final newline,
|
|
341
|
+
and no trailing whitespace. CMake/YAML commonly use two spaces as specified
|
|
342
|
+
by `.editorconfig`.
|
|
343
|
+
- New C++ uses `snake_case` variables/namespaces, `m_` members, `g_` globals,
|
|
344
|
+
`UPPER_CASE` constants, and `PascalCase` classes/functions/methods. Do not
|
|
345
|
+
rename broad legacy surfaces merely to match modern style.
|
|
346
|
+
- Prefer RAII, value returns, `std::optional` for genuine optional values,
|
|
347
|
+
`nullptr`, list initialization, explicit namespace qualification, named
|
|
348
|
+
casts, and `++i`.
|
|
349
|
+
- Keep includes minimal and in the order enforced by project lint. Include the
|
|
350
|
+
corresponding header first in implementation files where the project does so.
|
|
351
|
+
- Use project synchronization annotations/macros and established lock order.
|
|
352
|
+
Do not call unknown/external code while holding a lock without analyzing
|
|
353
|
+
reentrancy and order. Run a Debug build for synchronization changes.
|
|
354
|
+
- Use `Assert`, `Assume`, `CHECK_NONFATAL`, or explicit error handling according
|
|
355
|
+
to the distinctions documented under “Assertions and Checks”; never rely on
|
|
356
|
+
assertions for untrusted input validation.
|
|
357
|
+
- Monetary RPC values must use `AmountFromValue`/`ValueFromAmount`, not floating
|
|
358
|
+
point or hand parsing. RPC methods are lowercase; new argument/result fields
|
|
359
|
+
use established names or `snake_case`. Treat omitted and `null` arguments
|
|
360
|
+
consistently.
|
|
361
|
+
- Keep GUI work off the GUI thread when it can block. Preserve node/wallet/GUI
|
|
362
|
+
separation through interfaces.
|
|
363
|
+
- Python functional tests use `snake_case`, named constants instead of magic
|
|
364
|
+
values, precise `assert_equal`-style helpers, and framework synchronization.
|
|
365
|
+
- Add comments for why and for invariants, not line-by-line narration. Keep
|
|
366
|
+
patches reviewable; avoid unrelated refactors and formatting churn.
|
|
367
|
+
|
|
368
|
+
For new user-facing strings, follow the translation helpers already used by
|
|
369
|
+
the surrounding component. Do not manually edit translated
|
|
370
|
+
`src/qt/locale/bitcoin_*.ts` files. `bitcoin_en.ts` is generated with the
|
|
371
|
+
`translate` target as described in `doc/translation_process.md`.
|
|
372
|
+
|
|
373
|
+
## Change-specific checklist
|
|
374
|
+
|
|
375
|
+
- **Consensus/validation/script/serialization:** obtain explicit intent; verify
|
|
376
|
+
compatibility with historical data and peers; add unit, functional, and/or
|
|
377
|
+
fuzz coverage; test reindex/load paths when relevant. Preserve the Roots
|
|
378
|
+
consensus-neutral invariant.
|
|
379
|
+
- **Policy/mempool/mining:** prove the rule affects admission/relay/template
|
|
380
|
+
policy only unless a consensus change was explicitly requested. Test both
|
|
381
|
+
rejection and block acceptance where that boundary is at risk.
|
|
382
|
+
- **RPC:** update implementation, help text, argument/result schemas, and
|
|
383
|
+
functional tests together. Run `bitcoin-cli help <rpc>` and relevant doc lint.
|
|
384
|
+
- **Wallet/database:** test descriptor-wallet behavior by default and legacy
|
|
385
|
+
BDB behavior only with a compatible BDB build. Cover upgrade, unload/reload,
|
|
386
|
+
backup, and watch-only/signing implications as applicable.
|
|
387
|
+
- **P2P/networking:** test malformed and adversarial input, disconnect/ban
|
|
388
|
+
behavior, permissions, and resource bounds. Prefer P2P test-framework helpers
|
|
389
|
+
and add fuzz coverage for parsers.
|
|
390
|
+
- **GUI:** build with `-DBUILD_GUI=ON`, run `test_bitcoin-qt` when enabled, and
|
|
391
|
+
manually check the affected interaction without a real wallet/datadir.
|
|
392
|
+
- **Build/CI/platform:** configure at least one affected minimal combination;
|
|
393
|
+
remember options interact. Consult `ci/test/00_setup_env_*.sh` rather than
|
|
394
|
+
assuming the default build represents CI.
|
|
395
|
+
- **User-visible feature/config/API:** add `doc/release-notes-<PR>.md` when a PR
|
|
396
|
+
number is known. Do not invent a number; report the need in the handoff.
|
|
397
|
+
|
|
398
|
+
## Generated files, docs, and repository hygiene
|
|
399
|
+
|
|
400
|
+
- Never edit files in a build directory as source; reconfigure/regenerate them.
|