Repository navigation
backport: replace Autotools with CMake (bitcoin#30454, #30664, #31013, #31395) - #7481
Conversation
|
✅ Final review complete — no blockers (commit bc908d12) · triage: normal |
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: e732253572
ℹ️ About Codex in GitHub
Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
|
Note Reviews pausedIt looks like this branch is under active development. To avoid overwhelming you with review comments due to an influx of new commits, CodeRabbit has automatically paused this review. You can configure this behavior by changing the Use the following commands to manage reviews:
Use the checkboxes below for quick actions:
WalkthroughThis PR replaces the dummy CMake project with a complete Dash Core build system. It adds configuration and platform detection, dependency and toolchain integration, core libraries and executables, wallet/GUI/ZMQ/IPC support, tests, benchmarks, fuzzing, coverage, deployment scripts, vcpkg metadata, CI workflows, and Windows build documentation. Testing instructions are updated for CMake and CTest. Estimated code review effort: 5 (Critical) | ~150 minutes Suggested reviewers: Pre-merge checks:
|
There was a problem hiding this comment.
Actionable comments posted: 9
🧹 Nitpick comments (9)
CMakePresets.json (1)
6-21: 📐 Maintainability & Code Quality | 🔵 Trivial | ⚡ Quick winDeclare an out-of-source
binaryDirfor both presets.Without it, the output location depends on the caller’s working directory. Set a deterministic ignored build path, e.g.
${sourceDir}/build/${presetName}, socmake --presetcannot populate the checkout. CMake 3.21 permits omittedbinaryDirand falls back to regular CMake behavior. (cmake.org)Proposed fix
"generator": "Visual Studio 17 2022", "architecture": "x64", + "binaryDir": "${sourceDir}/build/${presetName}", "toolchainFile": "$env{VCPKG_ROOT}\\scripts\\buildsystems\\vcpkg.cmake",Also applies to: 24-39
🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In `@CMakePresets.json` around lines 6 - 21, Add a deterministic out-of-source binaryDir to both presets, using the preset name under the source directory (for example, ${sourceDir}/build/${presetName}). Keep the existing preset configuration unchanged so each cmake --preset invocation writes only to its own ignored build directory.CMakeLists.txt (4)
764-782: 📐 Maintainability & Code Quality | 🔵 Trivial | ⚡ Quick winDash-specific options are missing from the configure summary.
ENABLE_MINER,ENABLE_STACKTRACES, andENABLE_CRASH_HOOKSare the Dash-only switches restored in this PR (and stacktraces can be silently flipped off by the detection block at lines 625-638), so they're exactly the values a user needs echoed back.♻️ Proposed addition
message(" USDT tracing ........................ ${WITH_USDT}") +message(" in-wallet miner ..................... ${ENABLE_MINER}") +message(" stack traces ........................ ${ENABLE_STACKTRACES}") +message(" crash hooks ......................... ${ENABLE_CRASH_HOOKS}") message(" QR code (GUI) ....................... ${WITH_QRENCODE}")🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In `@CMakeLists.txt` around lines 764 - 782, Add the Dash-specific configuration switches ENABLE_MINER, ENABLE_STACKTRACES, and ENABLE_CRASH_HOOKS to the configure summary near the existing optional-feature messages. Echo each variable’s final value, including any value changes made by the stacktrace detection logic, so the summary reflects the effective configuration.
674-693: 📐 Maintainability & Code Quality | 🔵 Trivial | 💤 Low valueDead
wrap_exceptions_flagsaccumulation.The
VAR wrap_exceptions_flagsresult is never consumed (the literal-Wl,-wrap,...options are added explicitly below) and is then unset. Since onlyRESULT_VARmatters here, droppingVARmakes the probe intent clearer.🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In `@CMakeLists.txt` around lines 674 - 693, Remove the unused VAR wrap_exceptions_flags argument from the try_append_linker_flag call while preserving RESULT_VAR CRASH_HOOKS_WRAPPED_CXX_ABI and the existing target_link_options behavior. Also remove the now-unnecessary unset(wrap_exceptions_flags) statement.
382-394: 📐 Maintainability & Code Quality | 🔵 Trivial | 💤 Low value
SANITIZERSis consumed but never declared as a cache variable.It works via
-DSANITIZERS=...(untyped cache entry), but it won't show up in the configure summary tooling /ccmake/cmake-guiand has no help text, unlike the neighboringAPPEND_*cache strings.♻️ Proposed change
+set(SANITIZERS "" CACHE STRING "Comma-separated list of sanitizers to enable, e.g. \"address,undefined\".") + add_library(sanitize_interface INTERFACE)🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In `@CMakeLists.txt` around lines 382 - 394, Declare SANITIZERS as a typed cache string with descriptive help text near the neighboring APPEND_* cache variables, preserving its existing use in the sanitizer flag check and allowing it to appear in configure summaries, ccmake, and cmake-gui.
117-126: 📐 Maintainability & Code Quality | 🔵 Trivial | 💤 Low value
WARN_INCOMPATIBLE_BDB=OFFno longer suppresses the incompatible-BDB warning.The version warning is emitted unconditionally and only the follow-up hint respects
WARN_INCOMPATIBLE_BDB, so the option's help text ("Warn when using a Berkeley DB version other than 4.8.") doesn't match behavior.🛠️ Proposed fix
if(NOT BerkeleyDB_VERSION VERSION_EQUAL 4.8) - message(WARNING "Found Berkeley DB (BDB) other than 4.8.\n" - "BDB (legacy) wallets opened by this build will not be portable!" - ) if(WARN_INCOMPATIBLE_BDB) - message(WARNING "If this is intended, pass \"-DWARN_INCOMPATIBLE_BDB=OFF\".\n" + message(WARNING "Found Berkeley DB (BDB) other than 4.8.\n" + "BDB (legacy) wallets opened by this build will not be portable!\n" + "If this is intended, pass \"-DWARN_INCOMPATIBLE_BDB=OFF\".\n" "Passing \"-DWITH_BDB=OFF\" will suppress this warning." ) endif() endif()🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In `@CMakeLists.txt` around lines 117 - 126, Update the incompatible Berkeley DB warning conditional around BerkeleyDB_VERSION so the entire warning, including its follow-up hint, is emitted only when WARN_INCOMPATIBLE_BDB is enabled. Preserve the existing version check and warning messages when the option remains enabled.cmake/ccache.cmake (1)
26-33: 📐 Maintainability & Code Quality | 🔵 Trivial | 💤 Low valueAdd the missing
include(TryAppendCXXFlags).This module calls
try_append_cxx_flags()but relies on the rootCMakeLists.txthaving included the defining module first. The include guard makes an explicit include free and keeps the module self-contained.🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In `@cmake/ccache.cmake` around lines 26 - 33, Make cmake/ccache.cmake self-contained by explicitly including the module that defines try_append_cxx_flags before the WITH_CCACHE block. Add include(TryAppendCXXFlags) near the top, relying on its include guard and leaving the existing flag configuration unchanged.cmake/module/ProcessConfigurations.cmake (2)
92-107: 📐 Maintainability & Code Quality | 🔵 Trivial | 💤 Low value
flagis interpolated as an unanchored regex.
remove_cxx_flag_from_all_configs(-DNDEBUG)would also strip any flag containing that substring (e.g.-DNDEBUG_SOMETHING), and a caller passing a flag with regex metacharacters would misbehave. Anchoring the pattern makes the intent exact.♻️ Proposed change
- list(FILTER flags EXCLUDE REGEX "${flag}") + list(FILTER flags EXCLUDE REGEX "^${flag}$")🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In `@cmake/module/ProcessConfigurations.cmake` around lines 92 - 107, Update remove_cxx_flag_from_all_configs so list(FILTER ... EXCLUDE REGEX ...) matches the entire flag rather than an unanchored substring. Anchor the interpolated flag pattern and escape any regex metacharacters in flag before filtering, preserving removal of only the exact requested compiler flag.
120-129: 📐 Maintainability & Code Quality | 🔵 Trivial | 💤 Low valueModule has an implicit dependency on
core_interface_debugexisting.
include(ProcessConfigurations)silently requires the caller to have already createdcore_interface_debug(rootCMakeLists.txtline 250). A short comment or anif(NOT TARGET ...)fatal error would make the contract explicit for future includes.🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In `@cmake/module/ProcessConfigurations.cmake` around lines 120 - 129, Add an explicit prerequisite check before the target_compile_definitions call for core_interface_debug: detect when the target does not exist and fail with a clear fatal error explaining that callers must create it before including ProcessConfigurations. Keep the existing configuration definitions unchanged.cmake/script/GenerateHeaderFromJson.cmake (1)
12-21: 🚀 Performance & Scalability | 🔵 Trivial | ⚡ Quick winShared root cause: duplicated, per-byte
file(APPEND)hex-dump loop in both header generators. Both scripts issue onefile(APPEND)call per input byte, which is both a DRY violation and slow for larger fixture files at configure time.
cmake/script/GenerateHeaderFromJson.cmake#L12-L21: accumulate the formatted bytes into a single string and write it with onefile(APPEND)/file(WRITE)call instead of one call per byte; consider extracting the shared loop into a common CMake macro/module reused by both generators.cmake/script/GenerateHeaderFromRaw.cmake#L11-L20: apply the same batching fix and share the extracted macro from the JSON variant.🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In `@cmake/script/GenerateHeaderFromJson.cmake` around lines 12 - 21, The duplicated byte-formatting loops in GenerateHeaderFromJson.cmake and GenerateHeaderFromRaw.cmake perform one file append per byte. Extract the shared loop into a reusable CMake macro/module, have both generators accumulate formatted bytes into one string, and write the complete output with a single file(APPEND) or file(WRITE) call; update both listed sites accordingly.
🤖 Prompt for all review comments with AI agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
Inline comments:
In `@cmake/ccache.cmake`:
- Around line 8-14: Replace the readlink-based resolution in the ccache
masquerade check with CMake 3.22’s portable file(REAL_PATH) command. Resolve
both CMAKE_CXX_COMPILER and CCACHE_EXECUTABLE before comparing them in the
existing CCACHE_EXECUTABLE/NOT WITH_CCACHE condition, preserving the warning
behavior when either path reaches the same ccache executable through symlinks.
In `@cmake/module/FindLibevent.cmake`:
- Around line 43-80: Guard the post-discovery setup in both VCPKG and pkg-config
branches with if(Libevent_FOUND): only call check_evhttp_connection_get_peer,
create the libevent aliases, and apply the related mark_as_advanced entries when
discovery succeeded. Keep the existing package discovery and required-variable
handling unchanged.
In `@cmake/module/GenerateSetupNsi.cmake`:
- Around line 5-17: Update generate_setup_nsi to use Dash executable names for
the existing BITCOIN_*_NAME variables, replacing the bitcoin-* values with their
dash-* equivalents, and emit dash-win64-setup.nsi instead of
bitcoin-win64-setup.nsi while preserving the template variable names.
In `@CMakeLists.txt`:
- Around line 222-230: Apply APPEND_CFLAGS to the C compilation rule by
appending it to CMAKE_C_COMPILE_OBJECT, alongside the existing CXX rule updates.
Also append APPEND_CXXFLAGS to CMAKE_OBJCXX_COMPILE_OBJECT so Objective-C++
compilation receives the configured C++ flags, preserving the existing ordering
at the end of each command line.
- Around line 617-623: Update the warn_interface INTERFACE_COMPILE_OPTIONS setup
to transform each entry in warn_interface_flags into its own compile-language
generator expression, preserving each expression as a single list element. Do
not wrap the entire semicolon-separated flag list in one genex; keep the
existing CXX/OBJCXX language scope and cleanup behavior.
In `@doc/build-windows-msvc.md`:
- Line 49: Correct the typo “istructions” to “instructions” in the documentation
sentence describing the Debug and Release configurations.
- Line 27: Update every PowerShell code fence in doc/build-windows-msvc.md to
use the powershell language label, including the fences near the referenced
sections, so each opening fence follows the ```powershell format required by
MD040.
- Around line 11-18: Add a setup and validation step in the Windows build guide
before the first cmake --list-presets or build command to define VCPKG_ROOT for
the installed vcpkg directory and verify that it is set to a valid location.
Ensure the existing vs2022 and vs2022-static preset commands run only after this
variable is available.
In `@src/qt/CMakeLists.txt`:
- Around line 56-93: Move masternodemodel.cpp and masternodemodel.h from the
ENABLE_WALLET-gated source block into the unconditional bitcoinqt source list,
while keeping masternodelist.cpp wallet-gated. Ensure the CMake target includes
MasternodeModel when ENABLE_WALLET=OFF, matching the unconditional
BITCOIN_QT_BASE_CPP and BITCOIN_QT_H lists.
---
Nitpick comments:
In `@cmake/ccache.cmake`:
- Around line 26-33: Make cmake/ccache.cmake self-contained by explicitly
including the module that defines try_append_cxx_flags before the WITH_CCACHE
block. Add include(TryAppendCXXFlags) near the top, relying on its include guard
and leaving the existing flag configuration unchanged.
In `@cmake/module/ProcessConfigurations.cmake`:
- Around line 92-107: Update remove_cxx_flag_from_all_configs so list(FILTER ...
EXCLUDE REGEX ...) matches the entire flag rather than an unanchored substring.
Anchor the interpolated flag pattern and escape any regex metacharacters in flag
before filtering, preserving removal of only the exact requested compiler flag.
- Around line 120-129: Add an explicit prerequisite check before the
target_compile_definitions call for core_interface_debug: detect when the target
does not exist and fail with a clear fatal error explaining that callers must
create it before including ProcessConfigurations. Keep the existing
configuration definitions unchanged.
In `@cmake/script/GenerateHeaderFromJson.cmake`:
- Around line 12-21: The duplicated byte-formatting loops in
GenerateHeaderFromJson.cmake and GenerateHeaderFromRaw.cmake perform one file
append per byte. Extract the shared loop into a reusable CMake macro/module,
have both generators accumulate formatted bytes into one string, and write the
complete output with a single file(APPEND) or file(WRITE) call; update both
listed sites accordingly.
In `@CMakeLists.txt`:
- Around line 764-782: Add the Dash-specific configuration switches
ENABLE_MINER, ENABLE_STACKTRACES, and ENABLE_CRASH_HOOKS to the configure
summary near the existing optional-feature messages. Echo each variable’s final
value, including any value changes made by the stacktrace detection logic, so
the summary reflects the effective configuration.
- Around line 674-693: Remove the unused VAR wrap_exceptions_flags argument from
the try_append_linker_flag call while preserving RESULT_VAR
CRASH_HOOKS_WRAPPED_CXX_ABI and the existing target_link_options behavior. Also
remove the now-unnecessary unset(wrap_exceptions_flags) statement.
- Around line 382-394: Declare SANITIZERS as a typed cache string with
descriptive help text near the neighboring APPEND_* cache variables, preserving
its existing use in the sanitizer flag check and allowing it to appear in
configure summaries, ccmake, and cmake-gui.
- Around line 117-126: Update the incompatible Berkeley DB warning conditional
around BerkeleyDB_VERSION so the entire warning, including its follow-up hint,
is emitted only when WARN_INCOMPATIBLE_BDB is enabled. Preserve the existing
version check and warning messages when the option remains enabled.
In `@CMakePresets.json`:
- Around line 6-21: Add a deterministic out-of-source binaryDir to both presets,
using the preset name under the source directory (for example,
${sourceDir}/build/${presetName}). Keep the existing preset configuration
unchanged so each cmake --preset invocation writes only to its own ignored build
directory.
🪄 Autofix (Beta)
Fix all unresolved CodeRabbit comments on this PR:
- Push a commit to this branch (recommended)
- Create a new PR with the fixes
ℹ️ Review info
⚙️ Run configuration
Configuration used: Repository UI
Review profile: CHILL
Plan: Pro Plus
Run ID: 4ac2e910-9426-48d3-be53-31596dc5cb13
📒 Files selected for processing (72)
.gitignoreCMakeLists.txtCMakePresets.jsonREADME.mdcmake/bitcoin-config.h.incmake/ccache.cmakecmake/cov_tool_wrapper.sh.incmake/crc32c.cmakecmake/introspection.cmakecmake/leveldb.cmakecmake/minisketch.cmakecmake/module/AddBoostIfNeeded.cmakecmake/module/AddWindowsResources.cmakecmake/module/CheckSourceCompilesAndLinks.cmakecmake/module/FindBerkeleyDB.cmakecmake/module/FindLibevent.cmakecmake/module/FindMiniUPnPc.cmakecmake/module/FindNATPMP.cmakecmake/module/FindQt5.cmakecmake/module/FindUSDT.cmakecmake/module/FlagsSummary.cmakecmake/module/GenerateHeaders.cmakecmake/module/GenerateSetupNsi.cmakecmake/module/GetTargetInterface.cmakecmake/module/Maintenance.cmakecmake/module/ProcessConfigurations.cmakecmake/module/TestAppendRequiredLibraries.cmakecmake/module/TryAppendCXXFlags.cmakecmake/module/TryAppendLinkerFlag.cmakecmake/module/WarnAboutGlobalProperties.cmakecmake/script/Coverage.cmakecmake/script/CoverageFuzz.cmakecmake/script/CoverageInclude.cmake.incmake/script/GenerateBuildInfo.cmakecmake/script/GenerateHeaderFromJson.cmakecmake/script/GenerateHeaderFromRaw.cmakecmake/script/macos_zip.shcmake/tests.cmakecontrib/guix/README.mddepends/Makefiledepends/funcs.mkdepends/hosts/default.mkdepends/packages/qt.mkdepends/patches/qt/qt.prodepends/toolchain.cmake.indoc/CMakeLists.txtdoc/Doxyfile.indoc/build-windows-msvc.mdsrc/CMakeLists.txtsrc/Makefile.qt.includesrc/bench/CMakeLists.txtsrc/crypto/CMakeLists.txtsrc/dashbls/CMakeLists.txtsrc/dashbls/depends/relic/CMakeLists.txtsrc/dashbls/depends/relic/include/relic_conf.h.cmake.insrc/dashbls/src/CMakeLists.txtsrc/ipc/CMakeLists.txtsrc/qt/CMakeLists.txtsrc/qt/test/CMakeLists.txtsrc/test/CMakeLists.txtsrc/test/fuzz/CMakeLists.txtsrc/test/fuzz/util/CMakeLists.txtsrc/test/util/CMakeLists.txtsrc/univalue/CMakeLists.txtsrc/util/CMakeLists.txtsrc/wallet/CMakeLists.txtsrc/wallet/test/CMakeLists.txtsrc/wallet/test/fuzz/CMakeLists.txtsrc/zmq/CMakeLists.txttest/CMakeLists.txttest/functional/test_runner.pyvcpkg.json
💤 Files with no reviewable changes (1)
- depends/patches/qt/qt.pro
thepastaclaw
left a comment
There was a problem hiding this comment.
Preliminary review — Codex only
The default native CMake path may build, but five confirmed blockers leave supported optional and deployment paths unusable: the MinGW installer is malformed, multiprocess configuration and installed binaries are broken, and CMake drops Dash's public consensus library. Two commit-history cleanups are also warranted for durable backport provenance and logical WERROR commit grouping.
Validated blockers were found in the Codex precheck. Sonnet is deferred until a fresh Codex revalidation clears the blocker gate.
Review provenance
- Codex reviewers:
gpt-5.6-sol— general (completed),gpt-5.6-sol— dash-core-commit-history (completed),gpt-5.6-sol— backport-reviewer (completed) - Verifier:
gpt-5.6-sol— verifier - Sonnet: not run (deferred by blocker gate)
🔴 5 blocking | 🟡 2 suggestion(s)
2 additional finding(s) omitted (not in diff).
🤖 Prompt for all review comments with AI agents
These findings are from an automated code review. Verify each finding against the current code and only fix it if needed.
In `cmake/module/GenerateSetupNsi.cmake`:
- [BLOCKING] cmake/module/GenerateSetupNsi.cmake:9-15: NSIS script references unstaged Bitcoin executable names
The deploy command stages each target using `$<TARGET_FILE_NAME:...>`, and the targets have Dash output names such as `dash-qt.exe`, `dashd.exe`, and `dash-cli.exe`. This generator instead substitutes `bitcoin-qt.exe`, `bitcoind.exe`, `bitcoin-cli.exe`, and the other upstream names into `share/setup.nsi.in`. NSIS therefore cannot find the staged files and cannot build the installer. `PACKAGE_TARNAME` also registers the upstream `bitcoin` URI class instead of preserving Dash's Autotools package name, `dashcore`.
In `cmake/module/Maintenance.cmake`:
- [BLOCKING] cmake/module/Maintenance.cmake:75-84: Windows deploy script never specifies an installer output file
`share/setup.nsi.in` contains no `OutFile` directive, and this command invokes `makensis` without injecting one. The existing Autotools recipe explicitly appends `OutFile "$@"` before invoking NSIS. Without an equivalent directive, NSIS has no output path and cannot produce the custom command's declared `bitcoin-win64-setup.exe`, so the `deploy` target cannot complete.
In `src/test/CMakeLists.txt`:
- [BLOCKING] src/test/CMakeLists.txt:191-210: Missing prerequisite: bitcoin#28921
The imported `WITH_MULTIPROCESS` block references `ipc_test.cpp`, `ipc_test.capnp`, and `ipc_tests.cpp`, but none exists in the PR head. Upstream bitcoin#30454 added this CMake block after bitcoin#28921 had already introduced those files, so copying the block without its prerequisite makes `-DWITH_MULTIPROCESS=ON` fail during CMake generation with the default `BUILD_TESTS=ON`. Backport and adapt the IPC test sources, or omit this test block until that prerequisite is present.
In `src/CMakeLists.txt`:
- [BLOCKING] src/CMakeLists.txt:400-410: Multiprocess node is installed under a name its IPC code cannot spawn
Unlike the other Dash executables, `bitcoin-node` has no `OUTPUT_NAME`, so CMake installs it as `bitcoin-node`. The IPC implementation constructs a sibling path using the literal name passed to `spawnProcess`, and `rpc/node.cpp` passes `dash-node`; an installed CMake build therefore tries to launch a file that does not exist. The corresponding `bitcoin-gui` target in `src/qt/CMakeLists.txt` also lacks the `dash-gui` output name required by `init/bitcoin-gui.cpp` and the documented Autotools behavior.
- [BLOCKING] src/CMakeLists.txt:93-119: Missing prerequisite: bitcoin#29648
bitcoin#30454 was developed after bitcoin#29648 removed Bitcoin Core's public consensus library, but Dash still provides and documents that API. The imported CMake code creates only an internal `STATIC EXCLUDE_FROM_ALL` archive and provides no shared `libdashconsensus`, installed `script/bitcoinconsensus.h`, generated and installed `libdashconsensus.pc`, or `HAVE_CONSENSUS_LIB` definition. This is not equivalent to Dash's default Autotools build and silently skips the public API checks guarded by `HAVE_CONSENSUS_LIB` in `script_tests.cpp`. Preserve Dash's public shared-library target, export definitions, header and pkg-config installation, and API tests in the CMake build.
In `<commit:86f540b>`:
- [SUGGESTION] <commit:86f540b>:1: Record the upstream provenance of the CMake import
The initial commit imports and adapts the CMake system from Bitcoin Core but its permanent commit message contains no upstream PR, source commit, or cherry-pick provenance. Record bitcoin/bitcoin#30454 and the exact merge or source hash used so future maintainers can compare Dash's adaptation with its upstream origin directly from `git log`.
In `<commit:c5915df>`:
- [SUGGESTION] <commit:c5915df>:1: Combine the adjacent WERROR parity commits
Commits `c5915df` and `e732253` are adjacent parts of the same logical change: making CMake's warning, WERROR, and optimization behavior match Autotools. The first commit also contains an unrelated explanatory change in `test/CMakeLists.txt`. Squash the WERROR changes into one commit and move the test-config comment into `fdfafe8`, where that wiring was fixed.
e732253 to
1feb706
Compare
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: 1feb706873
ℹ️ About Codex in GitHub
Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
There was a problem hiding this comment.
Actionable comments posted: 5
🤖 Prompt for all review comments with AI agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
Inline comments:
In @.github/workflows/build-src-cmake.yml:
- Line 42: Update the workflow’s container-image input handling so the image
producer resolves inputs.container-path to an immutable `@sha256` digest before
invoking reusable jobs, including the job using the container image. Reject or
fail when the resolved reference is not digest-pinned, and pass the validated
digest reference rather than forwarding needs.container.outputs.path unchanged.
- Line 68: Stop re-parsing CI inputs as shell source: in
.github/workflows/build-src-cmake.yml at lines 68, 86, 106, and 113, pass
depends-dep-opts through env and invoke make with a validated argument array,
while reading build-target from the environment rather than interpolating it
into shell text. In ci/dash/build_src_cmake.sh lines 36-42, remove bash -c and
invoke CMake and cmake --build using quoted paths and argument arrays.
- Around line 45-49: Update the workflow trigger and checkout configuration
around the checkout step and build execution so untrusted pull-request head code
is not executed in the privileged workflow. Prefer running this build on
pull_request rather than pull_request_target; if checkout remains necessary
under pull_request_target, isolate it as read-only and set persist-credentials
to false before running ci/dash/build_src_cmake.sh and tests.
In `@doc/build-windows-msvc.md`:
- Around line 24-29: Update the “Clone Dash Repository” instructions after the
git clone command to explicitly change into the cloned dash directory before any
subsequent build or CMake commands are run.
- Line 82: Update the build-performance guidance in doc/build-windows-msvc.md to
remove the recommendation for excluding the entire Dash repository from
Microsoft Defender. If exclusion advice is retained, scope it to a temporary
build directory only, make the security trade-off explicit, and require
compliance with local security policy.
🪄 Autofix (Beta)
Fix all unresolved CodeRabbit comments on this PR:
- Push a commit to this branch (recommended)
- Create a new PR with the fixes
ℹ️ Review info
⚙️ Run configuration
Configuration used: Repository UI
Review profile: CHILL
Plan: Pro Plus
Run ID: e5c9da9c-14c7-46fa-83ce-00af43e7ed6c
📒 Files selected for processing (21)
.github/workflows/build-src-cmake.yml.github/workflows/build.ymlCMakeLists.txtci/dash/build_src_cmake.shci/dash/matrix.shci/dash/test_integrationtests_cmake.shci/dash/test_unittests_cmake.shci/test/00_setup_env_native_cmake.shcmake/bitcoin-config.h.incmake/introspection.cmakecmake/module/TryAppendLinkerFlag.cmakecmake/script/GenerateBuildInfo.cmakedoc/build-windows-msvc.mdsrc/CMakeLists.txtsrc/crypto/CMakeLists.txtsrc/qt/test/CMakeLists.txtsrc/test/CMakeLists.txtsrc/test/fuzz/CMakeLists.txtsrc/zmq/CMakeLists.txttest/CMakeLists.txttest/functional/test_runner.py
🚧 Files skipped from review as they are similar to previous changes (12)
- src/zmq/CMakeLists.txt
- test/functional/test_runner.py
- src/qt/test/CMakeLists.txt
- src/test/fuzz/CMakeLists.txt
- test/CMakeLists.txt
- src/crypto/CMakeLists.txt
- cmake/script/GenerateBuildInfo.cmake
- cmake/module/TryAppendLinkerFlag.cmake
- cmake/introspection.cmake
- src/test/CMakeLists.txt
- CMakeLists.txt
- src/CMakeLists.txt
1feb706 to
47e47e8
Compare
There was a problem hiding this comment.
Actionable comments posted: 1
🤖 Prompt for all review comments with AI agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
Inline comments:
In @.github/workflows/build-src-cmake.yml:
- Around line 32-36: Update the workflow header documentation to accurately
reflect that the CMake job runs the integration smoke test via
test_integrationtests_cmake.sh, while only the full functional test suite
remains in the Autotools jobs; preserve the existing workflow steps.
🪄 Autofix (Beta)
Fix all unresolved CodeRabbit comments on this PR:
- Push a commit to this branch (recommended)
- Create a new PR with the fixes
ℹ️ Review info
⚙️ Run configuration
Configuration used: Repository UI
Review profile: CHILL
Plan: Pro Plus
Run ID: ae657ff6-af94-47fd-955d-430c509ae5e6
📒 Files selected for processing (9)
.github/workflows/build-src-cmake.yml.github/workflows/build.ymlCMakeLists.txtci/dash/build_src_cmake.shci/dash/matrix.shci/dash/test_integrationtests_cmake.shci/dash/test_unittests_cmake.shci/test/00_setup_env_native_cmake.shsrc/crypto/CMakeLists.txt
🚧 Files skipped from review as they are similar to previous changes (3)
- .github/workflows/build.yml
- ci/test/00_setup_env_native_cmake.sh
- CMakeLists.txt
The Windows resource targets referenced bitcoind-res.rc, bitcoin-cli-res.rc, bitcoin-tx-res.rc and bitcoin-wallet-res.rc; this tree carries dashd-res.rc, dash-cli-res.rc, dash-tx-res.rc and dash-wallet-res.rc, so no Windows build could be generated at all. bitcoin-node and bitcoin-gui were installed under their upstream names even though rpc/node.cpp spawns 'dash-node' and init/bitcoin-gui.cpp declares EXE_NAME 'dash-gui', and bitcoin-chainstate was built but never added to installable_targets. GenerateSetupNsi.cmake substituted the upstream executable names and the bitcoin URI class into share/setup.nsi.in, so the generated NSIS script referenced files the deploy target never stages. It also never emitted an OutFile directive, which the Autotools recipe pipes in ahead of the script, leaving makensis with no output path. masternodemodel.cpp is part of BITCOIN_QT_BASE_CPP under Autotools but was grouped with masternodelist.cpp inside if(ENABLE_WALLET). rpcconsole.cpp, clientfeeds.cpp and proposalinfo.cpp are all non-wallet sources that use MasternodeModel, so a GUI build without wallet support lost its definition. Finally, Dash still provides the public libdashconsensus API that upstream removed in bitcoin#29648, so build it as a shared library with its own compilation of the consensus and crypto sources (BUILD_BITCOIN_INTERNAL, DISABLE_OPTIMIZED_SHA256), install the header and pkg-config file, and define HAVE_CONSENSUS_LIB so script_tests exercises the public API. The internal static library goes back to its own name to avoid colliding with it. Found by CodeRabbit, Codex and thepastaclaw review on PR dashpay#7481. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
The block referenced ipc_test.cpp, ipc_test.capnp and ipc_tests.cpp, none of which exist in this tree; upstream added them in bitcoin#28921, which Dash has not backported. Configuring with -DWITH_MULTIPROCESS=ON therefore failed during generation, since BUILD_TESTS defaults to ON. Found by Codex and thepastaclaw review on PR dashpay#7481. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
README told readers to run ctest from the repository root, where in-source builds are refused and no tests exist; the MSVC guide never set VCPKG_ROOT that its own presets reference, never entered the cloned directory, and carried an upstream typo. The linux64_cmake workflow header also still claimed functional tests were left entirely to Autotools, which stopped being true when the smoke list was added. Found by CodeRabbit and Codex review on PR dashpay#7481. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
There was a problem hiding this comment.
Actionable comments posted: 1
Caution
Some comments are outside the diff and can’t be posted inline due to platform limitations.
⚠️ Outside diff range comments (1)
doc/build-windows-msvc.md (1)
63-74: 🎯 Functional Correctness | 🟠 Major | ⚡ Quick winUse separate build directories for the two vcpkg triplets.
vs2022andvs2022-staticset differentVCPKG_TARGET_TRIPLETvalues for the samebuilddirectory. Configure the second workflow with a different directory, e.g.build-vs2022andbuild-vs2022-static, or deletebuildbefore switching.🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In `@doc/build-windows-msvc.md` around lines 63 - 74, Update the Windows build instructions for the vs2022 and vs2022-static presets to use separate build directories, such as build-vs2022 and build-vs2022-static, consistently across configure, build, test, and install commands. Alternatively, explicitly instruct users to delete the existing build directory before switching presets.
🤖 Prompt for all review comments with AI agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
Inline comments:
In `@doc/build-windows-msvc.md`:
- Around line 35-39: Update the VCPKG_ROOT example in the build instructions to
avoid hardcoding the Visual Studio Community edition. Use an edition-independent
placeholder for the Visual Studio install directory or instruct users to
determine their own bundled vcpkg path before configuring.
---
Outside diff comments:
In `@doc/build-windows-msvc.md`:
- Around line 63-74: Update the Windows build instructions for the vs2022 and
vs2022-static presets to use separate build directories, such as build-vs2022
and build-vs2022-static, consistently across configure, build, test, and install
commands. Alternatively, explicitly instruct users to delete the existing build
directory before switching presets.
🪄 Autofix (Beta)
Fix all unresolved CodeRabbit comments on this PR:
- Push a commit to this branch (recommended)
- Create a new PR with the fixes
ℹ️ Review info
⚙️ Run configuration
Configuration used: Repository UI
Review profile: CHILL
Plan: Pro Plus
Run ID: e9cf604b-af1c-4baa-8c3b-ec419ff1384f
📒 Files selected for processing (12)
.github/workflows/build-src-cmake.ymlCMakeLists.txtREADME.mdcmake/bitcoin-config.h.incmake/introspection.cmakecmake/module/GenerateSetupNsi.cmakecmake/module/Maintenance.cmakedoc/build-windows-msvc.mdsrc/CMakeLists.txtsrc/crypto/CMakeLists.txtsrc/qt/CMakeLists.txtsrc/test/CMakeLists.txt
🚧 Files skipped from review as they are similar to previous changes (7)
- README.md
- .github/workflows/build-src-cmake.yml
- src/qt/CMakeLists.txt
- cmake/module/Maintenance.cmake
- src/crypto/CMakeLists.txt
- cmake/bitcoin-config.h.in
- CMakeLists.txt
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: bd328dd146
ℹ️ About Codex in GitHub
Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: fc0a91c10c
ℹ️ About Codex in GitHub
Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
thepastaclaw
left a comment
There was a problem hiding this comment.
Preliminary review — Codex only
The latest changes correctly fix the consensus-header installation, checkout-v6 configuration, Windows/debug parity, shared Doxyfile handling, and Qt test plugin staging. Two blockers remain: the CMake consensus library still changes the established ABI runtime name, and the new depends fallback consumes a built-artifact output that the producer workflow neither declares nor uploads. The remaining suggestions concern inaccurate MSVC guidance and commit organization.
Validated blockers were found in the Codex precheck. Sonnet is deferred until a fresh Codex revalidation clears the blocker gate.
Review provenance
- Codex reviewers:
gpt-5.6-sol— general (completed),gpt-5.6-sol— dash-core-commit-history (completed),gpt-5.6-sol— backport-reviewer (completed) - Verifier:
gpt-5.6-sol— verifier - Sonnet: not run (deferred by blocker gate)
🔴 2 blocking | 🟡 7 suggestion(s)
6 additional finding(s) omitted (not in diff).
🤖 Prompt for all review comments with AI agents
These findings are from an automated code review. Verify each finding against the current code and only fix it if needed.
In `src/CMakeLists.txt`:
- [BLOCKING] src/CMakeLists.txt:129-133: Preserve libdashconsensus's versioned runtime name
Dash's retained Autotools target builds `libdashconsensus.la` without a libtool version override, which gives the public library ABI-major runtime names such as `libdashconsensus.so.0` and `libdashconsensus.0.dylib`. The new CMake `SHARED` target has no `SOVERSION`, so it installs an unversioned runtime library instead. Existing applications linked against ABI major 0 cannot load that CMake-installed replacement; set `SOVERSION 0` so both build systems publish the same runtime identity.
In `.github/workflows/build.yml`:
- [BLOCKING] .github/workflows/build.yml:232: Provide the depends artifact output before consuming it
This line passes `needs.depends-linux64.outputs.built-artifact`, but the exact-head `.github/workflows/build-depends.yml` declares only `key`, `host`, and `dep-opts` and contains no artifact upload. The expression therefore resolves to an empty value. When the cache restore misses, `build-src-cmake.yml` skips the download and deliberately exits with an error, so the new fallback never transfers the depends prefix produced earlier in the same run. Add the producer output and upload step in the same PR, or remove the consumer until that prerequisite is present.
In `<commit:86f540b>`:
- [SUGGESTION] <commit:86f540b>:1: Record the upstream provenance of the CMake import
Commit `86f540b432a22eee9126a781c6824401dadb50d2` imports more than 6,000 lines of Bitcoin Core's CMake migration with an empty body. Reword it to identify bitcoin/bitcoin#30454 and upstream merge `338bc2cd261ba3daf7fb494f8cb4a534762e292c`, and explain that the imported tree was adapted to coexist with Dash's retained Autotools build. This provenance should remain discoverable in permanent history.
In `<commit:24a9dc0>`:
- [SUGGESTION] <commit:24a9dc0>:1: Make the libbitcoinkernel removal atomic
Commit `24a9dc04c81` deletes `src/kernel/CMakeLists.txt` and its parent wiring without mentioning libbitcoinkernel in its message. The following commit `fc106e0864f` removes `BUILD_KERNEL_LIB` and describes dropping the target after its implementation file is already gone. Put the target file, option, status output, and parent wiring removal in one dedicated commit so the history and bisect boundary accurately describe the change.
In `<commit:86af396>`:
- [SUGGESTION] <commit:86af396>:1: Split the omnibus CMake review-fix commit by subsystem
Commit `86af39687b4` combines independent changes to Windows resources and packaging, multiprocess executable names, chainstate installation, no-wallet Qt wiring, and the public libdashconsensus API. These paths will be searched and bisected independently. Split the commit into focused Windows packaging, Qt no-wallet, and consensus-library changes, then place subsequent corrections with the corresponding focused commit.
In `<commit:ca2750c>`:
- [SUGGESTION] <commit:ca2750c>:1: Squash the PACKAGE_TARNAME cleanup into the preceding fixes
Commit `ca2750c901a` consolidates duplicate `PACKAGE_TARNAME` values introduced by the immediately preceding `86af39687b4` commit and also adds an unrelated dependency-order comment to that commit's consensus block. Fold the metadata cleanup into the Windows packaging change and place the consensus comment with the focused library-restoration change.
In `doc/build-windows-msvc.md`:
- [SUGGESTION] doc/build-windows-msvc.md:35-39: Use an edition-independent vcpkg root in the example
Both Windows presets dereference `VCPKG_ROOT`, but this example hardcodes the Visual Studio Community installation path. That path is invalid for Professional, Enterprise, Build Tools, and custom installations. Preserve an existing environment-provided value when available; otherwise instruct the user to set it to the `VC\vcpkg` directory under their actual Visual Studio installation and verify that `scripts\buildsystems\vcpkg.cmake` exists.
In `<commit:e921601>`:
- [SUGGESTION] <commit:e921601>:1: Combine the adjacent WERROR parity commits
Adjacent commits `e92160130b2` and `4604a5f750c` divide the same Autotools warning, WERROR, language-scoping, and sphlib optimization parity work. The first also includes a fuzz-configuration explanation unrelated to its warning fix. Combine the warning changes into one WERROR parity commit and move the fuzz explanation to the earlier test-wiring commit.
In `<commit:bd328dd>`:
- [SUGGESTION] <commit:bd328dd>:1: Fold the documentation corrections into their introducing commits
Commit `bd328dd1466` corrects three unrelated documentation sections introduced earlier in this unmerged stack: the README CTest command, the MSVC clone and vcpkg instructions, and the CMake workflow's test-scope description. Fold each correction into the commit that introduced the corresponding text so permanent blame points to the commit responsible for the final guidance.
|
This pull request has conflicts, please rebase. |
The Windows resource targets referenced bitcoind-res.rc, bitcoin-cli-res.rc, bitcoin-tx-res.rc and bitcoin-wallet-res.rc; this tree carries dashd-res.rc, dash-cli-res.rc, dash-tx-res.rc and dash-wallet-res.rc, so no Windows build could be generated at all. bitcoin-node and bitcoin-gui were installed under their upstream names even though rpc/node.cpp spawns 'dash-node' and init/bitcoin-gui.cpp declares EXE_NAME 'dash-gui', and bitcoin-chainstate was built but never added to installable_targets. GenerateSetupNsi.cmake substituted the upstream executable names and the bitcoin URI class into share/setup.nsi.in, so the generated NSIS script referenced files the deploy target never stages. It also never emitted an OutFile directive, which the Autotools recipe pipes in ahead of the script, leaving makensis with no output path. masternodemodel.cpp is part of BITCOIN_QT_BASE_CPP under Autotools but was grouped with masternodelist.cpp inside if(ENABLE_WALLET). rpcconsole.cpp, clientfeeds.cpp and proposalinfo.cpp are all non-wallet sources that use MasternodeModel, so a GUI build without wallet support lost its definition. Finally, Dash still provides the public libdashconsensus API that upstream removed in bitcoin#29648, so build it as a shared library with its own compilation of the consensus and crypto sources (BUILD_BITCOIN_INTERNAL, DISABLE_OPTIMIZED_SHA256), install the header and pkg-config file, and define HAVE_CONSENSUS_LIB so script_tests exercises the public API. The internal static library goes back to its own name to avoid colliding with it. Found by CodeRabbit, Codex and thepastaclaw review on PR dashpay#7481. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
The block referenced ipc_test.cpp, ipc_test.capnp and ipc_tests.cpp, none of which exist in this tree; upstream added them in bitcoin#28921, which Dash has not backported. Configuring with -DWITH_MULTIPROCESS=ON therefore failed during generation, since BUILD_TESTS defaults to ON. Found by Codex and thepastaclaw review on PR dashpay#7481. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
README told readers to run ctest from the repository root, where in-source builds are refused and no tests exist; the MSVC guide never set VCPKG_ROOT that its own presets reference, never entered the cloned directory, and carried an upstream typo. The linux64_cmake workflow header also still claimed functional tests were left entirely to Autotools, which stopped being true when the smoke list was added. Found by CodeRabbit and Codex review on PR dashpay#7481. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
fc0a91c to
eac1703
Compare
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: eac17034af
ℹ️ About Codex in GitHub
Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
thepastaclaw
left a comment
There was a problem hiding this comment.
Preliminary review — Codex only
The depends-artifact handoff is fixed at this head, but the CMake installation still changes libdashconsensus's established runtime identity, so changes remain required. The ten-comment budget retains that blocker, seven cumulative suggestions, and the two highest-priority test-coverage gaps; lower-priority benchmark and documentation completeness claims are adjudicated below.
Source: reviewer backends: gpt-5.6-sol (general, dash-core-commit-history, backport-reviewer); final verifier backend: gpt-5.6-sol. openclaw-agent/cliproxy/gpt-5.6-sol is orchestration-only and is not reviewer evidence.
Validated blockers were found in the Codex precheck. Opus is deferred until a fresh Codex revalidation clears the blocker gate.
Review provenance
- Codex reviewers:
gpt-5.6-sol— general (completed),gpt-5.6-sol— dash-core-commit-history (completed),gpt-5.6-sol— backport-reviewer (completed) - Verifier:
gpt-5.6-sol— verifier - Sonnet: not run (deferred by blocker gate)
🔴 1 blocking | 🟡 8 suggestion(s)
6 additional finding(s) omitted (not in diff).
2 carried-forward finding(s) already raised on this PR; not re-posting as new inline comments.
🤖 Prompt for all review comments with AI agents
These findings are from an automated code review. Verify each finding against the current code and only fix it if needed.
In `<commit:1b2b49f>`:
- [SUGGESTION] <commit:1b2b49f>:1: Record the upstream provenance of the CMake import
Commit `1b2b49fe80436ea0758a2422edd522014c3c08dd` imports and adapts more than 6,000 lines from Bitcoin Core's CMake migration but has an empty body. Record bitcoin/bitcoin#30454 and upstream merge `338bc2cd261ba3daf7fb494f8cb4a534762e292c`, together with the intentional adaptation to retain Autotools, so future maintainers can compare the permanent Dash commit directly with its source.
In `<commit:eb91067>`:
- [SUGGESTION] <commit:eb91067>:1: Make the libbitcoinkernel removal atomic
Commit `eb9106715a9` deletes `src/kernel/CMakeLists.txt` under a message about unrelated Dash build failures, while `02705514cad` later removes `BUILD_KERNEL_LIB`, its fuzz override, and its status output under the target-removal explanation. Put the target file deletion, parent wiring, option, override, and status removal in one dedicated commit so the supported target set and bisect boundary remain coherent.
In `<commit:3b5bf6c>`:
- [SUGGESTION] <commit:3b5bf6c>:1: Split the omnibus CMake review-fix commit by subsystem
Commit `3b5bf6c18c2` combines unrelated Windows resources and packaging, multiprocess executable names, chainstate installation, no-wallet Qt wiring, and restoration of the public libdashconsensus API. Split these into focused Windows packaging, executable/install, Qt no-wallet, and consensus-library commits so independent subsystems can be searched and bisected without traversing one omnibus review-fix commit.
In `<commit:7d81202>`:
- [SUGGESTION] <commit:7d81202>:1: Squash the PACKAGE_TARNAME cleanup into the preceding fixes
Commit `7d8120255a1` consolidates `PACKAGE_TARNAME` definitions introduced by the preceding Windows-packaging restoration and also adds an unrelated dependency-order comment to the consensus-library block. Fold the metadata cleanup into the focused packaging change and place the source-order explanation with the focused libdashconsensus restoration.
In `<commit:22a1812>`:
- [SUGGESTION] <commit:22a1812>:1: Combine the adjacent WERROR parity commits
Adjacent commits `22a1812737b` and `9b4a772cd36` divide the same Autotools warning, WERROR, language-scoping, and sphlib optimization parity work. The first also adds an unrelated fuzz-configuration explanation. Combine the warning and optimization changes into one WERROR-parity commit and move the fuzz explanation to `34601915001`, where the CMake test wiring is corrected.
In `<commit:c8961f6>`:
- [SUGGESTION] <commit:c8961f6>:1: Fold the documentation corrections into their introducing commits
Commit `c8961f66eae` combines corrections to three unrelated sections introduced earlier in this unmerged stack: the root README's CTest command, the MSVC clone/vcpkg instructions, and the CMake workflow's test-scope comment. Fold each correction into the commit that introduced the corresponding text so permanent blame points directly to the commit responsible for the final guidance.
In `src/test/CMakeLists.txt`:
- [SUGGESTION] src/test/CMakeLists.txt:79-171: Include all unit suites in the CMake test target
The CMake `test_bitcoin` target omits 12 suites still compiled by `BITCOIN_TESTS`: `evo_db_tests.cpp`, `governance_inv_tests.cpp`, the three governance vote suites, `llmq_blockprocessor_tests.cpp`, `llmq_invalid_type_tests.cpp`, `masternode_payments_tests.cpp`, `spork_tests.cpp`, `translation_tests.cpp`, `txrequest_tests.cpp`, and `unordered_lru_cache_tests.cpp`. All files exist, and upstream bitcoin#30454 itself includes the translation and transaction-request suites. The new CMake CI job can therefore report a successful full unit-test run while skipping Dash evo, governance, LLMQ, masternode, spork, and cache regressions.
In `src/test/fuzz/CMakeLists.txt`:
- [SUGGESTION] src/test/fuzz/CMakeLists.txt:50-112: Restore the omitted upstream fuzz targets
Bitcoin#30454 includes `partially_downloaded_block.cpp` and `txrequest.cpp` in the fuzz executable and `mempool.cpp` in the fuzz utility library. All three files exist in Dash and remain wired into the Autotools fuzz build, but the new CMake source lists omit them. Add the two harnesses to `src/test/fuzz/CMakeLists.txt` and `mempool.cpp` to `src/test/fuzz/util/CMakeLists.txt` so CMake fuzz builds retain the imported network, transaction-request, and mempool coverage.
In `src/CMakeLists.txt`:
- [BLOCKING] src/CMakeLists.txt:128-133: Preserve libdashconsensus's versioned runtime name
(existing thread: https://github.com/dashpay/dash/pull/7481#discussion_r3696264407)
The retained Autotools target publishes ABI-major runtime names such as `libdashconsensus.so.0` and `libdashconsensus.0.dylib`. This new shared CMake target sets no `SOVERSION`, so it installs an unversioned runtime library that cannot replace the Autotools-built package for applications already linked against ABI major 0. Set `SOVERSION 0` on `dashconsensus` to preserve the public library's established runtime identity.
In `doc/build-windows-msvc.md`:
- [SUGGESTION] doc/build-windows-msvc.md:35-39: Use an edition-independent vcpkg root in the example
(existing thread: https://github.com/dashpay/dash/pull/7481#discussion_r3696264423)
Both Windows presets dereference `VCPKG_ROOT`, but the example overwrites it with the Visual Studio Community path. That path is invalid for Professional, Enterprise, Build Tools, custom installations, and users with an existing standalone vcpkg installation. Preserve an existing value when available; otherwise direct users to the `VC\vcpkg` directory beneath their actual Visual Studio installation, and validate that `scripts\buildsystems\vcpkg.cmake` exists before invoking a preset.
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: ea51e5e122
ℹ️ About Codex in GitHub
Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
thepastaclaw
left a comment
There was a problem hiding this comment.
Preliminary review — Codex only
The latest commit restores the previously omitted unit and fuzz sources, but two blockers remain: CMake changes libdashconsensus's established runtime identity, and multi-config CTest runs pass utility executable paths under environment-variable names Dash does not read. The ten-finding budget also retains the cumulative commit-history findings, the MSVC vcpkg guidance issue, and the omitted descriptor benchmark; two lower-priority valid suggestions were omitted due to the budget.
Source: reviewer backends: gpt-5.6-sol (general, dash-core-commit-history, backport-reviewer); final verifier backend: gpt-5.6-sol. openclaw-agent/cliproxy/gpt-5.6-sol is orchestration-only and not reviewer evidence.
Validated blockers were found in the Codex precheck. Opus is deferred until a fresh Codex revalidation clears the blocker gate.
Review provenance
- Codex reviewers:
gpt-5.6-sol— general (completed),gpt-5.6-sol— dash-core-commit-history (completed),gpt-5.6-sol— backport-reviewer (completed) - Verifier:
gpt-5.6-sol— verifier - Sonnet: not run (deferred by blocker gate)
🔴 2 blocking | 🟡 7 suggestion(s)
6 additional finding(s) omitted (not in diff).
2 carried-forward finding(s) already raised on this PR; not re-posting as new inline comments.
🤖 Prompt for all review comments with AI agents
These findings are from an automated code review. Verify each finding against the current code and only fix it if needed.
In `cmake/tests.cmake`:
- [BLOCKING] cmake/tests.cmake:5-7: Pass Dash executable paths to utility tests
The utility runner reads only `DASHUTIL` and `DASHTX`, but this CTest command sets the upstream `BITCOINUTIL` and `BITCOINTX` variables. Single-config builds can succeed through the runner's `${BUILDDIR}/src/dash-*` fallback, masking the mismatch. With the documented Visual Studio presets, executables reside in configuration-specific directories such as `src/Release`; the fallback therefore cannot find them and `util_test_runner` fails. Pass the generated target paths through the Dash environment-variable names.
In `<commit:1b2b49f>`:
- [SUGGESTION] <commit:1b2b49f>:1: Record the upstream provenance of the CMake import
Commit `1b2b49fe80436ea0758a2422edd522014c3c08dd` imports and adapts more than 6,000 lines from Bitcoin Core's CMake migration but has an empty body. Record bitcoin/bitcoin#30454 and upstream merge `338bc2cd261ba3daf7fb494f8cb4a534762e292c`, together with the intentional adaptation to retain Autotools, so future maintainers can compare the permanent Dash commit directly with its source.
In `<commit:eb91067>`:
- [SUGGESTION] <commit:eb91067>:1: Make the libbitcoinkernel removal atomic
Commit `eb9106715a9` deletes `src/kernel/CMakeLists.txt` under a message about unrelated Dash build failures, while `02705514cad` later removes `BUILD_KERNEL_LIB`, its fuzz override, and its status output under the target-removal explanation. Put the target file deletion, parent wiring, option, override, and status removal in one dedicated commit so the supported target set and bisect boundary remain coherent.
In `<commit:3b5bf6c>`:
- [SUGGESTION] <commit:3b5bf6c>:1: Split the omnibus CMake review-fix commit by subsystem
Commit `3b5bf6c18c2` combines unrelated Windows resources and packaging, multiprocess executable names, chainstate installation, no-wallet Qt wiring, and restoration of the public libdashconsensus API. Split these into focused Windows packaging, executable/install, Qt no-wallet, and consensus-library commits so independent subsystems can be searched and bisected without traversing one omnibus review-fix commit.
In `<commit:7d81202>`:
- [SUGGESTION] <commit:7d81202>:1: Squash the PACKAGE_TARNAME cleanup into the preceding fixes
Commit `7d8120255a1` consolidates `PACKAGE_TARNAME` definitions introduced by the preceding Windows-packaging restoration and also adds an unrelated dependency-order comment to the consensus-library block. Fold the metadata cleanup into the focused packaging change and place the source-order explanation with the focused libdashconsensus restoration.
In `<commit:22a1812>`:
- [SUGGESTION] <commit:22a1812>:1: Combine the adjacent WERROR parity commits
Adjacent commits `22a1812737b` and `9b4a772cd36` divide the same Autotools warning, WERROR, language-scoping, and sphlib optimization parity work. The first also adds an unrelated fuzz-configuration explanation. Combine the warning and optimization changes into one WERROR-parity commit and move the fuzz explanation to `34601915001`, where the CMake test wiring is corrected.
In `<commit:c8961f6>`:
- [SUGGESTION] <commit:c8961f6>:1: Fold the documentation corrections into their introducing commits
Commit `c8961f66eae` combines corrections to three unrelated sections introduced earlier in this unmerged stack: the root README's CTest command, the MSVC clone/vcpkg instructions, and the CMake workflow's test-scope comment. Fold each correction into the commit that introduced the corresponding text so permanent blame points directly to the commit responsible for the final guidance.
In `src/bench/CMakeLists.txt`:
- [SUGGESTION] src/bench/CMakeLists.txt:26-28: Restore the descriptor benchmark to the CMake target
Upstream bitcoin#30454 includes `descriptors.cpp` in the benchmark target, and Dash's retained `src/Makefile.bench.include` still compiles it. The CMake list omits that file, so `bench_dash` lacks the high-priority `ExpandDescriptor` benchmark and its high-priority CTest sanity run no longer matches the Autotools benchmark inventory. Add `descriptors.cpp` to the target.
In `src/CMakeLists.txt`:
- [BLOCKING] src/CMakeLists.txt:129-133: Preserve libdashconsensus's versioned runtime name
(existing thread: https://github.com/dashpay/dash/pull/7481#discussion_r3696264407)
The retained Autotools target publishes ABI-major runtime names such as `libdashconsensus.so.0` and `libdashconsensus.0.dylib`. The new shared CMake target has no `SOVERSION`, so it installs an unversioned runtime library that cannot replace the Autotools-built package for applications already linked against ABI major 0. Set `SOVERSION 0` on `dashconsensus` to preserve the public library's established runtime identity.
In `doc/build-windows-msvc.md`:
- [SUGGESTION] doc/build-windows-msvc.md:34-38: Use an edition-independent vcpkg root in the example
(existing thread: https://github.com/dashpay/dash/pull/7481#discussion_r3696264423)
Both Windows presets dereference `VCPKG_ROOT`, but the example overwrites it with the Visual Studio Community path. That path is invalid for Professional, Enterprise, Build Tools, custom installations, and users with an existing standalone vcpkg installation. Preserve an existing value when available; otherwise direct users to the `VC\vcpkg` directory beneath their actual Visual Studio installation, and validate that `scripts\buildsystems\vcpkg.cmake` exists before invoking a preset.
There was a problem hiding this comment.
Re-review — Final validation — Phase 1 + Phase 2
Updated after discussion: 1 finding(s) of this review were settled in the inline threads.
- ✅ Withdrawn: 🟡 Enumerate the deferred upstream CMake test registrations
The final head contains no confirmed blocking build or runtime defects. One in-scope documentation suggestion remains: the PR should enumerate the deferred upstream CMake test and fuzz registrations instead of describing them only generically; completed CI checks are green while the remaining source-build jobs are still running.
🟡 1 0 suggestion(s)
Approved: nothing here blocks the merge. The non-blocking findings are optional: fix them forward, in this PR or a follow-up, without waiting for another review round.
Review provenance
Source: reviewer 1: muse-spark-1.3-contributor (agent: phase1-reviewer, role: general); reviewer 2: gpt-6.1-sol (agent: phase2-reviewer, role: general); reviewer 3: gpt-6.1-sol (agent: phase2-reviewer, role: backport-reviewer); reviewer 4: gpt-6.1-sol (agent: phase2-reviewer, role: dash-core-commit-history); final verifier: gpt-6.1-sol (agent: sol-verifier, role: final-verifier)
- Triage:
normalbygpt-6.1-sol(effort low) — This is a large, cross-cutting build-system, CI, packaging, and documentation migration, but the diff does not directly change consensus, funds movement, cryptography, peer-facing deserialization, or storage migrations. - Phase 1 reviewers:
muse-spark-1.3-contributor— general (completed, effort xhigh); agentphase1-reviewer - Phase 1 model:
muse-spark-1.3-contributor— not quota-gated; passed overgemini-3.8-flash-high(antigravity quota lookup failed),glm-5.3-flash(not used above high effort; tier asks max) - Fresh verifier:
gpt-6.1-sol— final-verifier; agentsol-verifier - Phase 2 reviewers:
gpt-6.1-sol— general (completed, effort high); agentphase2-reviewer,gpt-6.1-sol— backport-reviewer (completed, effort high); agentphase2-reviewer,gpt-6.1-sol— dash-core-commit-history (completed, effort high); agentphase2-reviewer
🤖 Prompt for all review comments with AI agents
These findings are from an automated code review (PastaClaw). How to work through them:
- Verify each finding against the code; fix only what is real, as narrowly as the finding.
- If a finding is wrong, decline it on its thread in one or two sentences; the bot answers once and stops.
- To defer one, link a tracking issue on its thread (or reply asking the bot to file one).
- Suggestions don't block the merge, but most are worth fixing in this PR; judge each on its merits.
- Keep the PR's scope; don't grow it to satisfy hardening suggestions.
- Carried items were raised in an earlier round: read the thread first, and don't re-fix what you already fixed, declined or deferred there. Disputed and deferred items are already handled and are not listed.
In `src/test/CMakeLists.txt`:
- [SUGGESTION → WITHDRAWN in discussion; do not fix] src/test/CMakeLists.txt:218-221: Enumerate the deferred upstream CMake test registrations
The final CMake target correctly avoids referencing sources that are absent from Dash, but the source comment documents only the IPC omission and the PR description describes the other omissions only generally. The upstream #30454 registration set also defers unit tests `cluster_linearize_tests.cpp`, `coinscachepair_tests.cpp`, `feefrac_tests.cpp`, `headers_sync_chainwork_tests.cpp`, `miniminer_tests.cpp`, `node_warnings_tests.cpp`, `policy_fee_tests.cpp`, and `timeoffsets_tests.cpp`; fuzz targets `bitdeque.cpp`, `block_index.cpp`, `cluster_linearize.cpp`, `crypto_chacha20poly1305.cpp`, `feefrac.cpp`, `feeratediagram.cpp`, `headerssync.cpp`, `mini_miner.cpp`, `p2p_handshake.cpp`, `package_eval.cpp`, `script_assets_test_minimizer.cpp`, `script_parsing.cpp`, and `timeoffsets.cpp`; and wallet fuzz source `wallet_bdb_parser.cpp`. Record this complete deferred inventory and identify non-adopted feature cases, or restore each registration when its source backport lands. This is documentation/completeness work, not a claim that the current target fails to generate.
| # NOTE: Upstream also builds an IPC test library here from ipc_test.cpp, | ||
| # ipc_test.capnp and ipc_tests.cpp. Dash has not backported those sources | ||
| # (bitcoin#28921), and referencing them makes -DWITH_MULTIPROCESS=ON fail | ||
| # during generation. Restore this block along with the test sources. |
There was a problem hiding this comment.
✅ Withdrawn at
0d520861; see the replies below.
🟡 Suggestion: Enumerate the deferred upstream CMake test registrations
The final CMake target correctly avoids referencing sources that are absent from Dash, but the source comment documents only the IPC omission and the PR description describes the other omissions only generally. The upstream bitcoin#30454 registration set also defers unit tests cluster_linearize_tests.cpp, coinscachepair_tests.cpp, feefrac_tests.cpp, headers_sync_chainwork_tests.cpp, miniminer_tests.cpp, node_warnings_tests.cpp, policy_fee_tests.cpp, and timeoffsets_tests.cpp; fuzz targets bitdeque.cpp, block_index.cpp, cluster_linearize.cpp, crypto_chacha20poly1305.cpp, feefrac.cpp, feeratediagram.cpp, headerssync.cpp, mini_miner.cpp, p2p_handshake.cpp, package_eval.cpp, script_assets_test_minimizer.cpp, script_parsing.cpp, and timeoffsets.cpp; and wallet fuzz source wallet_bdb_parser.cpp. Record this complete deferred inventory and identify non-adopted feature cases, or restore each registration when its source backport lands. This is documentation/completeness work, not a claim that the current target fails to generate.
source: gpt-6.1-sol (phase2-reviewer: general, backport-reviewer, dash-core-commit-history)
There was a problem hiding this comment.
Leaving this out. A list of files Dash does not have yet would go stale with every backport, and the registration is not something a later backport can forget: each of those upstream PRs adds its source to the test list in the same diff, so the line arrives with the file. span_tests.cpp was added that way in this push, after #8136 brought the file in.
🤖 Posted autonomously by Claude on behalf of pasta.
There was a problem hiding this comment.
I agree. Since each deferred upstream test registration arrives together with its source and the CMake source list is the authoritative registration point, maintaining a separate inventory here would be redundant and prone to staleness. I’m withdrawing this documentation suggestion.
Withdrawing this finding.
There was a problem hiding this comment.
Withdrawn (re-reviewed at 0d520861): The current CMake lists contain every test source present in Dash, and the PR description explains that absent upstream sources are deferred until their source backports arrive. The requested inventory is therefore not an actionable defect in this head.
|
This pull request has conflicts, please rebase. |
1 similar comment
|
This pull request has conflicts, please rebase. |
thepastaclaw
left a comment
There was a problem hiding this comment.
Re-review — Final validation — Phase 1 + Phase 2
Static verification of head 0d52086 found no remaining in-scope correctness, security, or build-system findings. Completed checks are successful; source-build, sanitizer, multiprocess, fuzz, platform-GUI, and review jobs are still queued or running, with no failed code checks currently reported.
🔴 0 blocking | 🟡 0 suggestion(s) | 💬 0 nitpick(s)
Review provenance
Source: reviewer 1: muse-spark-1.3-contributor (agent: phase1-reviewer, role: general); reviewer 2: gpt-6.1-sol (agent: phase2-reviewer, role: general); reviewer 3: gpt-6.1-sol (agent: phase2-reviewer, role: backport-reviewer); reviewer 4: gpt-6.1-sol (agent: phase2-reviewer, role: dash-core-commit-history); final verifier: gpt-6.1-sol (agent: sol-verifier, role: final-verifier)
- Triage:
normalbygpt-6.1-sol(effort low) — This is a large, intricate build-system, CI, packaging, and release-process migration, but it does not itself change consensus, funds movement, cryptography, signatures, peer-facing deserialization, or storage migrations. - Phase 1 reviewers:
muse-spark-1.3-contributor— general (completed, effort xhigh); agentphase1-reviewer - Phase 1 model:
muse-spark-1.3-contributor— not quota-gated; passed overgemini-3.8-flash-high(antigravity quota lookup failed),glm-5.3-flash(not used above high effort; tier asks max) - Fresh verifier:
gpt-6.1-sol— final-verifier; agentsol-verifier - Phase 2 reviewers:
gpt-6.1-sol— general (completed, effort high); agentphase2-reviewer,gpt-6.1-sol— backport-reviewer (completed, effort high); agentphase2-reviewer,gpt-6.1-sol— dash-core-commit-history (completed, effort high); agentphase2-reviewer
🤖 Prompt for all review comments with AI agents
These findings are from an automated code review. Verify the current code and confirm that no unresolved issues remain.
No unresolved findings remain from the prior review on this head.
|
This pull request has conflicts, please rebase. |
Import Bitcoin Core's CMake build system (merge commit 338bc2c), with its source lists and target names changed to the Dash tree: the top-level CMakeLists.txt, the cmake/ modules, the per-directory CMakeLists.txt under src/, and the depends changes that produce depends/<host>/toolchain.cmake. Left out: - vcpkg.json, CMakePresets.json and doc/build-windows-msvc.md. Dash Core does not build with MSVC yet, and CMake stops with an error there, so the manifest, the presets (both MSVC-only) and the guide would describe a build that cannot run. - Test and fuzz registrations whose sources Dash does not have, either because the upstream change is not backported yet or because it belongs to a feature Dash does not adopt (SegWit, RBF, Signet). - Upstream's CI, Guix and build-guide changes. Dash's own scripts and guides are switched to CMake in the later commits of this series, and the last one removes Autotools, as bitcoin#30664 did. The work the import needs before it builds Dash Core is in the next commit, so that this one stays comparable to its upstream source. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>
Bring the imported CMake build to parity with the Autotools build it replaces. The source lists and the config header were checked against src/Makefile*.am and configure.ac. Configuration: - Define the macros Dash sources test: ENABLE_MINER, the X11 SSSE3, AES-NI and ARM backends (with their sources), ENABLE_STACKTRACES and ENABLE_CRASH_HOOKS with libbacktrace from depends, HAVE_GETRANDOM, WITH_ZSTD and ENABLE_PLATFORM_GUI. Without them the build succeeds but behaves differently, for example every mining RPC is stubbed out. - Debug builds define DEBUG_CORE, which Dash uses in place of upstream's DEBUG. - Version 24.1.0, and the `git describe` build description that clientversion.cpp reads. - Windows targets the same API level and PE subsystem as configure.ac. - Linux binaries keep -Wl,--build-id=sha1, which the Guix release script needs to match them to their split debug files. - Appended C flags reach the X11 and secp256k1 subdirectories. - The pipe2 check is skipped on Darwin (bitcoin#36218). Warnings: a WERROR build works. The vendored sphlib C sources are built as their own -O3 library outside the C++ warning set, -Wundef is not an error, and the dashbls and immer headers are system includes. Targets: - Dash executable names, Windows resources and the NSIS installer. - The public libdashconsensus, which Dash keeps (upstream removed it in bitcoin#29648), with its libtool soname, Mach-O versions, install name and MinGW DLL name. It is not built for fuzzing. - dash-chainstate links a kernel library with the sources of libdashkernel_la_SOURCES. - dashbls is compiled from Dash's own source list, like src/Makefile.dashbls.include, with blst from depends through a new FindBLST module. The subtree's own CMake project downloads blst at configure time, so it cannot be used. ENABLE_MIMALLOC mirrors --disable-mimalloc. - Crash-hook --wrap flags are applied where the wrappers are linked. - coinjoin/walletman.cpp is in the wallet library, and the fuzz binary links test_util before the node library. Tests: every unit, fuzz, wallet, Qt and bench source in the tree is compiled. CTest also runs the dashbls and minisketch tests, which `make check` ran, and util_test_runner gets DASHUTIL and DASHTX. The multiprocess IPC test block is dropped: its sources come with bitcoin#28921, which is not backported. Maintenance targets: check-symbols and check-security use Dash's script paths, and check-stdlib-paths and osx_debug are added for Guix and CI. Co-Authored-By: UdjinM6 <UdjinM6@users.noreply.github.com> Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>
These are required when cross-compiling shared libraries such as the
kernel library.
Dash builds libdashconsensus as a shared library by default, so without
this the macOS cross build links it with the host's GNU ld instead of
ld64.lld ("unrecognised emulation mode: llvm"): -fuse-ld=lld and the
other depends LDFLAGS only reached executables.
(cherry picked from commit a8e0470)
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
…to prevent library conflict
CMake parses some paths from the spec of the C compiler, assuming it
will be the linker, resulting in the link to end up with
`-L/usr/lib/gcc/x86_64-w64-mingw32/12-win32` on debian bookworm if both
-win32 and -posix variants are installed, and -win32 is the default
alternative.
This results in the wrong C++ library being linked, missing
std::threads::hardware_concurrency and other threading functions.
To fix this, use the -posix variant of gcc as well when available. This
fixes a regression compared to autotools, where this scenario worked.
The Dash CI image installs both variants with -win32 as the default.
There the mismatched libstdc++ makes test_dash.exe deadlock under Wine
in its first test fixture ("RtlLeaveCriticalSection ... is not
acquired").
(cherry picked from commit ae56b32)
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Configure every CI job with CMake and depends' toolchain file instead of ./configure, as bitcoin#30454 did for upstream's CI. * ci/dash/build_src.sh configures build-ci/dashcore-$BUILD_TARGET with cmake and builds the job's GOAL with cmake --build. The build directory keeps the path the workflows already expect, so bundling and the test jobs are unchanged. The Autotools distdir round trip is gone along with `make dist`. Bench and fuzz binaries are still built everywhere, as Autotools did by default. CMake writes compile_commands.json itself, so the tidy job no longer runs Bear; the paths Bear's config excluded are dropped by the existing jq filter instead. * ci/dash/test_unittests.sh runs ctest instead of `make check`. * Each ci/test/00_setup_env_*.sh maps its configure flags to the matching CMake options. Compilers come from depends' toolchain unless the job used a different one at configure time; extra preprocessor, compiler and linker flags go through APPEND_CPPFLAGS, APPEND_CFLAGS, APPEND_CXXFLAGS and APPEND_LDFLAGS, so depends' own flags are kept. The nowallet job no longer builds a shared libdashkernel: the CMake build has no such target. The multiprocess Debug build keeps C code at -O2 and C++ at -O0, as before. * The ASan job builds against system libraries, so no toolchain file turns on the optional features those libraries provide. It names UPnP, NAT-PMP, zstd and USDT explicitly; ./configure detected them on its own. * The win64 job builds `install` instead of `deploy`. CMake's deploy target always runs makensis, which the CI image lacks; Autotools skipped the installer silently when makensis was missing, so CI never built it before either. Guix still builds the installer. * The fuzz and MSan jobs repeat the standard-library debug defines of their DEBUG=1 depends build in APPEND_CPPFLAGS. config.site applied those defines to every build; the toolchain file applies them to Debug builds only, and these jobs build RelWithDebInfo. * The sanitizer jobs now also instrument C code, including the mimalloc allocator in dashbls, whose random number generator relies on wrapping unsigned arithmetic. Suppress unsigned-integer-overflow for it. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>
Configure, build, check and install Dash Core with CMake and depends' toolchain file in the Guix build, as bitcoin#30454 did upstream (e821f0a). CONFIGFLAGS now holds CMake options, the host flags go through APPEND_CFLAGS, APPEND_CXXFLAGS and APPEND_LDFLAGS, the Windows installer and the macOS zip are produced by the deploy targets, and debug symbols are split with the configured build/split-debug.sh. Unlike upstream, macOS releases keep their dSYM debug archives through the osx_debug target, Windows releases still build without the miner, and the library directory is pinned to lib/ so the consensus library lands where the release archives expect it. Guix still exports C_INCLUDE_PATH and CPLUS_INCLUDE_PATH, so depends prefixes the darwin compilers with `env -u ...`. Name env by its full path: CMake does not search PATH for the first word of the OBJCXX compiler, and the Qt configure fails with "CMAKE_OBJCXX_COMPILER: env is not a full path and was not found in the PATH". Also move the Debian packaging and the GitHub Actions deploy image to CMake. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Binaries built with USDT tracing carry enough notes that `readelf -n` prints more than a pipe buffer. Both the build-id lookup (`awk ...; exit`) and the check (`grep -q`) stop reading at the first match, so readelf can die of SIGPIPE, which pipefail turns into a failure: a binary that has a build-id is reported as missing one, or the lookup aborts the script. Read all of the output instead. The check's result was also read with `$?` after a failing command substitution, which `set -e` never lets run, so the script exited without printing which binaries failed. Capture the status with `||` instead. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Replace the Autotools instructions in the build guides, the developer notes, the depends and test documentation and the agent guide with their CMake equivalents, as bitcoin#30454 did upstream: configure with `cmake -B build` (and depends' toolchain file), build with `cmake --build build`, run unit tests with ctest and functional tests from build/test/functional. Configure flags are mapped to the CMake options that replace them. Add a release note for the switch. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Remove configure.ac, autogen.sh, the Makefile.am files, the src/Makefile.*.include fragments, build-aux/m4, share/genbuild.sh and depends' config.site, now that CI, Guix and the documentation use CMake. This is Dash's counterpart of upstream merge commit ef6f49e: the same removal, applied to Dash's own set of Autotools files. Also drop what only Autotools used: the Bear configuration for the tidy job, the convenience Makefiles in src/, src/qt and src/test, the generated univalue source list, the .gitignore entries for in-tree build products, the Makefile sections in .editorconfig, and the Makefile.qt_locale.include that update-translations.py regenerated (the CMake build globs src/qt/locale/*.ts). Comments and messages that pointed at configure, configure.ac or `make check` now point at their CMake counterparts, and gen-manpages.py, gen-dash-conf.sh and ci/dash/build-docker.sh default to the build/ directory. The vendored subtrees keep their own build files, and depends packages still use their upstream Autotools builds. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>
|
This pull request has conflicts, please rebase. |
thepastaclaw
left a comment
There was a problem hiding this comment.
Re-review — Final validation — Phase 1 + Phase 2
The exact head completes the CMake cut-over and contains the previously verified Dash-specific target, dependency, packaging, CI, and history corrections; static inspection found no new in-scope defects. CI has no failed code checks, while several source-build jobs remain queued or in progress, so final automated validation is still pending.
🔴 0 blocking | 🟡 0 suggestion(s) | 💬 0 nitpick(s)
Review provenance
Source: reviewer 1: muse-spark-1.3-contributor (agent: phase1-reviewer, role: general); reviewer 2: gpt-6.1-sol (agent: phase2-reviewer, role: general); reviewer 3: gpt-6.1-sol (agent: phase2-reviewer, role: backport-reviewer); reviewer 4: gpt-6.1-sol (agent: phase2-reviewer, role: dash-core-commit-history); final verifier: gpt-6.1-sol (agent: sol-verifier, role: final-verifier)
- Triage:
normalbygpt-6.1-sol(effort low) — This is a large, intricate build-system, CI, packaging, and documentation migration, but it does not itself change consensus, funds movement, cryptography, signatures, peer-facing deserialization, or storage migrations. - Phase 1 reviewers:
muse-spark-1.3-contributor— general (completed, effort xhigh); agentphase1-reviewer - Phase 1 model:
muse-spark-1.3-contributor— not quota-gated; passed overgemini-3.8-flash-high(antigravity quota lookup failed),glm-5.3-flash(not used above high effort; tier asks max) - Fresh verifier:
gpt-6.1-sol— final-verifier; agentsol-verifier - Phase 2 reviewers:
gpt-6.1-sol— general (completed, effort high); agentphase2-reviewer,gpt-6.1-sol— backport-reviewer (completed, effort high); agentphase2-reviewer,gpt-6.1-sol— dash-core-commit-history (completed, effort high); agentphase2-reviewer
🤖 Prompt for all review comments with AI agents
These findings are from an automated code review. Verify the current code and confirm that no unresolved issues remain.
No unresolved findings remain from the prior review on this head.
|
To cleanup building directory after simultaneous removing of autoconf and adding cmake you may do: |
Issue being fixed or feature implemented
Bitcoin Core replaced Autotools with CMake (bitcoin#30454 added the CMake build and switched CI, Guix and the documentation; bitcoin#30664 removed Autotools). Since then every upstream backport carries build-system changes that Dash cannot apply, and keeping two build systems in step would double the maintenance for every source, flag or dependency change.
This PR switches Dash Core to CMake and removes the Autotools build.
What was done?
The series is nine commits, in this order:
partial Merge bitcoin/bitcoin#30454imports upstream's CMake tree at merge commit338bc2cd261: the top-levelCMakeLists.txt, thecmake/modules, the per-directoryCMakeLists.txtanddepends/toolchain.cmake.in, with source lists and target names changed to the Dash tree.build: adapt the CMake build to Dashbrings it to parity with the Autotools build it replaces:libdashconsensuswith its existing.so.0soname.-DENABLE_PLATFORM_GUI, enabled by depends'PLATFORM_GUI).check-symbols,check-security,check-stdlib-pathsandosx_debugmaintenance targets.-Wl,--build-id=sha1build-id, which the Guix release script needs to match them to their split debug files.ctestruns whatmake checkran, including the dashbls and minisketch tests.Merge bitcoin/bitcoin#31395applies depends' linker flags to shared libraries too, so the cross-compiledlibdashconsensuslinks with the depends linker.Merge bitcoin/bitcoin#31013uses the posix variant of the mingw C compiler as well as the C++ one. When both variants are installed and win32 is the default, as in the CI image, CMake otherwise links the win32 libstdc++, andtest_dash.exedeadlocks.ci:ci/dash/build_src.shconfigures every job with CMake, keeping thebuild-ci/dashcore-$BUILD_TARGETlayout the workflows bundle from. Jobs built on depends use its toolchain file. The ASan job builds against system libraries (ci: build linux64_asan against system libraries instead of depends #8106), so it has no toolchain file to turn on UPnP, NAT-PMP, zstd and USDT and names them in its own options.test_unittests.shrunsctest.ci/test/00_setup_env_*.shmaps its configure flags to CMake options, with extra flags passed throughAPPEND_CPPFLAGS,APPEND_CFLAGS,APPEND_CXXFLAGSandAPPEND_LDFLAGS.DEBUG=1depends build. The toolchain file applies those to Debug builds only, so the jobs pass them inAPPEND_CPPFLAGS.compile_commands.jsoninstead of Bear.guix: build releases with CMake:contrib/guix/libexec/build.shconfigures, builds, checks, installs and deploys with CMake, following upstream's e821f0a. macOS releases keep their dSYM debug archives, and Windows releases still disable the miner. The Debian packaging and the GitHub Actions deploy image use CMake too.guix: make the build-id check reliable: the check reads all ofreadelf's output, so binaries with USDT probe notes are no longer misreported as lacking a build-id underpipefail, and it prints the binaries that fail instead of exiting silently.doc:The build guides, developer notes, depends, test and fuzzing docs and the agent guide describe the CMake workflow.doc/release-notes-7481.mdannounces the switch.partial Merge bitcoin/bitcoin#30664removes Autotools:autogen.sh,configure.ac, everyMakefile.am, thesrc/Makefile.*.includefragments andbuild-aux/m4.share/genbuild.sh, depends'config.site.in,src/.bear-tidy-config,src/univalue/sources.mk.src/,src/qtandsrc/test, and the.gitignoreentries for in-tree build products..editorconfig.The vendored subtrees keep their own build files, and depends packages keep their upstream Autotools builds.
Not carried over:
vcpkg.json,CMakePresets.json(MSVC presets only) and the MSVC build guide. Dash Core does not build with MSVC, and CMake stops with an error there, so they would describe a build that cannot run. They can come back with MSVC support.libdashkernel(--with-experimental-kernel-lib --enable-shared), which the CMake build does not provide, so the nowallet CI job no longer builds it.--enable-suppress-external-warningshas no CMake equivalent; vendored headers are included as system headers instead.installinstead ofdeploy. The CI image has nomakensis, and Autotools skipped the installer silently, so CI never built it. Guix still does.How Has This Been Tested?
CI at the PR head covers every job: cross builds, sanitizers, fuzzing, multiprocess, tidy,
ctestand the full functional suite.Before the final history cleanup, on a tree that differs from this head only by the points listed below:
check-security,check-symbols,check-stdlib-paths, the macOS dSYM archives, the NSIS installer and the build-id debug split.PLATFORM_GUI=1, configured with-DWERROR=ON -DBUILD_BENCH=ON -DBUILD_FUZZ_BINARY=ON -DENABLE_EXTERNAL_SIGNER=ON -DWITH_ZMQ=ONand built everything, including dash-qt with the Platform SDK.ctestpassed 183/183, andbuild/test/functional/test_runner.pypassed wallet_basic, wallet_descriptor, tool_wallet, rpc_blockchain, mining_basic, feature_dip3_deterministicmns and feature_llmq_signing.Changed since then: the MSVC files were dropped, the minisketch test was added to
ctest, the fuzz and MSan jobs got their debug defines back, andgen-dash-conf.shandci/dash/build-docker.shdefault tobuild/. After #8136 merged, its newspan_tests.cppwas added to the unit-test list, and after #8072 merged, its newpackage_eval.cppwas added to the fuzz target list; those rebases were not built locally. After #7679 merged, the dashbls target picked up its new sources,dashbls_runtestgainedtest_secure.cpp, anddashbls_runbenchwas added; on macOS arm64 those three targets build with-Wall -Wextra -Werrorfromcmake/dashbls.cmakein a small wrapper project, anddashbls_runtestpasses. After #8165 merged (-Wundef, with the X11 headers fixed for it), the CMake build stopped downgrading-WundefunderWERROR, and the TSan job passes-D_LIBCPP_REMOVE_TRANSITIVE_INCLUDES; on macOS arm64 that head builds with-DWERROR=ON(no GUI), andcrypto_tests,hash_tests,netbase_tests,bls_testsanddashbls_runtestpass. On macOS arm64 the tree configures with CMake, and the newlibminisketch_testbuilds and passes.Breaking Changes
Building from source now requires CMake 3.22 or later:
cmake -B build && cmake --build build, with--toolchain depends/<host>/toolchain.cmakefor depends builds../autogen.shand./configureno longer exist. Functional tests are run asbuild/test/functional/test_runner.py. Release artifacts and executable names are unchanged.Checklist:
🤖 Generated with Claude Code