2c4b229c90 cmake: Introduce `FUZZ_LIBS` (Hennadii Stepanov)
ea929c0848 scripted-diff: Rename CMake helper module (Hennadii Stepanov)
8d238c1dfd cmake: Delete `check_cxx_source_links*` macros (Hennadii Stepanov)
71bf8294a9 cmake: Convert `check_cxx_source_compiles_with_flags` to a function (Hennadii Stepanov)
88ee6800c9 cmake: Delete `check_cxx_source_links_with_flags` macro (Hennadii Stepanov)
09e8fd25b1 build: Don't override CMake's default try_compile target (Hennadii Stepanov)
Pull request description:
This was requested in https://github.com/bitcoin/bitcoin/pull/31359#issuecomment-2515287092.
From https://github.com/bitcoin/bitcoin/pull/31359#issuecomment-2511246212:
> (Almost?) every CMake check internally uses the [`try_compile()`](https://cmake.org/cmake/help/latest/command/try_compile.html) command, whose behaviour, in turn, depends on the [`CMAKE_TRY_COMPILE_TARGET_TYPE`](https://cmake.org/cmake/help/latest/variable/CMAKE_TRY_COMPILE_TARGET_TYPE.html) variable:
>
> 1. The default value, `EXECUTABLE`, enables both compiler and linker checks.
>
> 2. The `STATIC_LIBRARY` value enables only compiler checks.
>
>
> To mimic Autotools' behaviour, we [disabled](d3f42fa08f/cmake/module/CheckSourceCompilesAndLinks.cmake (L9-L10)) linker checks by setting `CMAKE_TRY_COMPILE_TARGET_TYPE` to `STATIC_LIBRARY` globally (perhaps not the best design). This effectively separates the entire CMake script into regions where `CMAKE_TRY_COMPILE_TARGET_TYPE` is:
>
> * unset
>
> * set to `STATIC_LIBRARY`
>
> * set to `EXECUTABLE`
From https://github.com/bitcoin/bitcoin/pull/31359#issuecomment-2515287092:
> > This seems very fragile and unintuitive, and the fact that this could silently break at any point is not documented in any way. I don't think other bad design decisions should lead to us having to write even more boilerplate code to fix things that should "just work" (minus the upstream bugs).
>
> Agreed. I forgot that we set `CMAKE_TRY_COMPILE_TARGET_TYPE` globally. And even worse, it's buried in a module. If that upsets CMake internal tests, I think we should undo that.
This PR ensures that `CMAKE_TRY_COMPILE_TARGET_TYPE` is modified only within local scopes.
Additionally, the `FUZZ_LIBS` variable has been introduced to handle additional libraries required for linking, rather than link options, in certain build environment, such as OSS-Fuzz.
ACKs for top commit:
TheCharlatan:
Re-ACK 2c4b229c90
theuni:
utACK 2c4b229c90
Tree-SHA512: f72ffa8f50f216fc1a2f8027ba8ddfd4acd42b94ff6c1cb2138f2da51eb8f945660e97d3c247d7f3f7ec8dfebbccec3ab84347d6ae2e3f8a40f3d7aa8b14cde9
CMake assumes the default value internally, so overriding this causes
problems. The minimal speedup of skipping the linker isn't worth the
complexity of setting it to static.
This check is only used in release builds, where hardening should always
be enabled. I can't think of a reason we'd want to silently skip these
checks if hardening was inadvertently disabled.
81c174e318 cmake: Refer to the configure log instead of printing PIE test error (Hennadii Stepanov)
65a0920ca6 cmake: Add `CheckLinkerSupportsPIE` module (Hennadii Stepanov)
Pull request description:
This new module is a wrapper around CMake's `CheckPIESupported` module that fixes an upstream bug.
See: https://gitlab.kitware.com/cmake/cmake/-/issues/26463.
Fixes https://github.com/bitcoin/bitcoin/issues/30771.
ACKs for top commit:
theuni:
utACK 81c174e318.
vasild:
ACK 81c174e318
Tree-SHA512: 77d7022238551a4e69c59d1fe6b78975bb552cbbed5339459853d7ebf0086813036081f464fed230be330b3bd7d6cf8590b536b064028d2f786d6ae40f342f95
9b033bebb1 cmake: rename Kernel component to bitcoinkernel for consistency (Cory Fields)
2e0c92558e cmake: add and use install_binary_component (Cory Fields)
0264c5d86c cmake: use per-target components for bitcoin-qt and bitcoin-gui (Cory Fields)
fb0546b1c5 ci: don't try to install for a fuzz build (Cory Fields)
Pull request description:
This makes it possible to build/install only the desired binaries regardless of the configuration.
For consistency, the component names match the binary names. `Kernel` and `GUI` have been renamed.
Additionally it fixes#31762 by installing only the manpages for the configured targets (and includes them in the component installs for each).
Also fixes#31745.
Alternative to #31765 which is (imo) more correct/thorough.
Can be tested using (for ex):
```bash
$ cmake -B build
$ cmake --build build -t bitcoind -t bitcoin-cli
$ cmake --install build --component bitcoind
$ cmake --install build --component bitcoin-cli
```
ACKs for top commit:
hebasto:
ACK 9b033bebb1.
TheCharlatan:
Re-ACK 9b033bebb1
stickies-v:
re-ACK 9b033bebb1
Tree-SHA512: fd4818e76f190dbeafbf0c246b466f829771902c9d6d7111ed917093b811c8a5536a4a45e20708f73e7f581d6cb77c8e61cfa69e065788dcf0886792f553a355
12fa9511b5 build: simplify dependency graph (Cory Fields)
c4e498300c build: avoid unnecessary dependencies on generated headers (Cory Fields)
Pull request description:
These changes speed up my build (default config/options/targets) by roughly 10%. I suspect the difference may be more significant in other build configs.
Before:
> $ time cmake --build build -j24
> real3m26.932s
After:
> $ time cmake --build build -j24
> real3m7.556s
Generally they allow for jobservers (either `make -jX` or `ninja`) to be better utilized. This can be verified using `top` while building and looking at the number of compiles running at any given time before/after these changes. Before, it's easy to observe periods of stalling when only one or two compiles are happening. After these changes, the compiler process count should mostly match the number of jobs given (`-jX`) until it falls off at the end.
---
The first commit sets [DEPENDS_EXPLICIT_ONLY](https://cmake.org/cmake/help/latest/command/add_custom_command.html#command:add_custom_command) for commands which generate our test header files. Without this option, `test_bitcoin`'s generated headers won't be built until all of its other dependencies have been built. This introduces a significant stall in the build, though currently only Ninja benefits from this being set, and only CMake >= 3.27 understands it.
Example from a generated `build.ninja`:
Before:
> \# Custom command for src/test/data/base58_encode_decode.json.h
>
> build src/test/data/base58_encode_decode.json.h | ${cmake_ninja_workdir}src/test/data/base58_encode_decode.json.h: CUSTOM_COMMAND /home/cory/dev/bitcoin/src/test/data/base58_encode_decode.json /home/cory/dev/bitcoin/cmake/script/GenerateHeaderFromJson.cmake || libcrc32c.a libcrc32c_sse42.a libleveldb.a libminisketch.a minisketch_clmul src/bitcoin_clientversion src/crypto/libbitcoin_crypto.a src/crypto/libbitcoin_crypto_avx2.a src/crypto/libbitcoin_crypto_sse41.a src/crypto/libbitcoin_crypto_x86_shani.a src/generate_build_info src/libbitcoin_cli.a src/libbitcoin_common.a src/libbitcoin_consensus.a src/libbitcoin_node.a src/secp256k1/src/libsecp256k1.a src/secp256k1/src/secp256k1_precomputed src/test/util/libtest_util.a src/univalue/libunivalue.a src/util/libbitcoin_util.a src/wallet/libbitcoin_wallet.a src/zmq/libbitcoin_zmq.a
After:
> \# Custom command for src/test/data/base58_encode_decode.json.h
>
> build src/test/data/base58_encode_decode.json.h | ${cmake_ninja_workdir}src/test/data/base58_encode_decode.json.h: CUSTOM_COMMAND /home/cory/dev/bitcoin/src/test/data/base58_encode_decode.json /home/cory/dev/bitcoin/cmake/script/GenerateHeaderFromJson.cmake
---
The second commit is more significant. It sets [CMAKE_OPTIMIZE_DEPENDENCIES](https://cmake.org/cmake/help/latest/prop_tgt/OPTIMIZE_DEPENDENCIES.html) globally, which allows the objects of static libs to be built in parallel when one lib depends on the other. This can be set as a per-lib property, ~but I don't see any need for that as we don't currently have any edge-cases where this wouldn't be ok. If those should arise, we could always disable on a per-lib basis~.
Edit: turns out this triggers an [upstream bug](https://gitlab.kitware.com/cmake/cmake/-/issues/24058), which I guess can be considered an edge-case until fixed in CMake. I've added 2 per-lib opt-outs as a result.
Example:
Before:
> \# Link the static library src/libbitcoin_cli.a
>
> build src/libbitcoin_cli.a: CXX_STATIC_LIBRARY_LINKER__bitcoin_cli_RelWithDebInfo src/CMakeFiles/bitcoin_cli.dir/compat/stdin.cpp.o src/CMakeFiles/bitcoin_cli.dir/rpc/client.cpp.o || src/univalue/libunivalue.a
After:
> \# Link the static library src/libbitcoin_cli.a
>
> build src/libbitcoin_cli.a: CXX_STATIC_LIBRARY_LINKER__bitcoin_cli_RelWithDebInfo src/CMakeFiles/bitcoin_cli.dir/compat/stdin.cpp.o src/CMakeFiles/bitcoin_cli.dir/rpc/client.cpp.o
>
ACKs for top commit:
l0rinc:
utACK 12fa9511b5
hebasto:
ACK 12fa9511b5.
Tree-SHA512: f85f507e70cdc06acd07542161d9f9b8edf9ba866f08c8ef17aaaed770fa11530a27521c4413456d863463a6e77d4d6983fa623a64e17bbd602c2bc70aacc112
56a9b847bb build: set build type and per-build-type flags as early as possible (Cory Fields)
f605f7a9c2 build: refactor: set debug definitions in main CMakeLists (Cory Fields)
Pull request description:
This ensures that most compiler tests are not run with the wrong build type's flags. The initial c++ checks are an exception to that because many internal CMake variables are unset until a language is selected, so it's problematic to change our build type before that.
The difference can be seen in `build/CMakeFiles/CMakeConfigureLog.yaml`. Before, `Debug` was used for many of the earlly checks. After this PR, it's only the first 2 checks.
ACKs for top commit:
hebasto:
ACK 56a9b847bb.
Tree-SHA512: 87947352d6d4fd08554515822cb13634ed3be33fcda2af817c22ef943b1d0856ceb39311ddc01b8221397528fdc09f630dc7e74fc92f5a4a073f09c4ae493596
With the exception of the first c++ checks, this ensures that compiler tests
are never run with the wrong build type's flags.
Co-Authored-By: Hennadii Stepanov <32963518+hebasto@users.noreply.github.com>
These scripts are becoming more of nuisance, than a value-add;
particularly since we've been building releases using Guix. Adding new
(release bin) tests can be harder, because it requires constructing a
failing test, which is becoming less easy e.g trying to disable a
feature or protection that has been built into the compiler/toolchain by
default.
In the pre-Guix days, these were valuable to sanity-check the environment,
because we were pulling that pre-built from Ubuntu, with little control.
At this point, it's less clear what these scripts are (sanity) checking.
Note that these also weren't completely ported to CMake (#31698), see
also #31715 which contains other fixes that would be needed for these
test-tests, to accomodate future changes.
This prevents the generation of these headers from also depending on the
dependencies of the libs/binaries which consume them.
Specifically, this prevents generated test headers (such as
test/data/base58_encode_decode.json.h) from depending on the
dependencies of test_bitcoin (libcrc32c.a libcrc32c_sse42.a libleveldb.a)
Note that this is currently only relevant for Ninja.
For more detail, see:
https://cmake.org/cmake/help/latest/command/add_custom_command.html
97a18c8545 cmake: Fix `IF_CHECK_PASSED` option handling (Hennadii Stepanov)
Pull request description:
`IF_CHECK_PASSED` is a multi-value keyword, resulting in a list value. Convert it to a string before applying any `string()` command.
Split from https://github.com/bitcoin/bitcoin/pull/30861.
No current CMake code is affected by this bug.
ACKs for top commit:
theuni:
utACK 97a18c8545
Tree-SHA512: d2556ca38c35a8992175e9f948c2028a789e71c2b2d5fdf365b31710c8ed3d5edf5d0363853c5d750d29abb58cfda3c78cdc2971a627e5b4c61aca4ec2a33356
5a96767e3f depends, libevent: Do not install *.pc files and remove patches for them (Hennadii Stepanov)
ffda355b5a cmake, refactor: Move `HAVE_EVHTTP_...` to `libevent` interface (Hennadii Stepanov)
b619bdc330 cmake: Revamp `FindLibevent` module (Hennadii Stepanov)
Pull request description:
This PR generalizes the use of `find_package` / `pkg_check_modules`, prioritizing the former.
Addresses https://github.com/bitcoin/bitcoin/pull/30903#issuecomment-2444700876:
> We should also follow up with refactoring the libevent module, to more generically use CMake/pkg-config, rather than restricting the CMake usage to `vcpkg`. At that point, we'd likely be able to dump pkg-config for the depends path entirely.
Similar to https://github.com/bitcoin/bitcoin/pull/30903.
ACKs for top commit:
fanquake:
ACK 5a96767e3f
Tree-SHA512: 181020c16ccd2821e718c73f264badcdc5e62980c4a8d9691e759efe2ea00da2326e26308d1dcfdeac01e9e27930428ecace9f36941deee951b751b138d7266c
0de3e96e33 tracing: use bitcoind pid in bcc tracing examples (0xb10c)
411c6cfc6c tracing: only prepare tracepoint args if attached (0xb10c)
d524c1ec06 tracing: dedup TRACE macros & rename to TRACEPOINT (0xb10c)
Pull request description:
Currently, if the tracepoints are compiled (e.g. in depends and release builds), we always prepare the tracepoint arguments regardless of the tracepoints being used or not. We made sure that the argument preparation is as cheap as possible, but we can almost completely eliminate any overhead for users not interested in the tracepoints (the vast majority), by gating the tracepoint argument preparation with an `if(something is attached to this tracepoint)`. To achieve this, we use the optional semaphore feature provided by SystemTap.
The first commit simplifies and deduplicates our tracepoint macros from 13 TRACEx macros to a single TRACEPOINT macro. This makes them easier to use and also avoids more duplicate macro definitions in the second commit.
The Linux tracing tools I'm aware of (bcc, bpftrace, libbpf, and systemtap) all support the semaphore gating feature. Thus, all existing tracepoints and their argument preparation is gated in the second commit. For details, please refer to the commit messages and the updated documentation in `doc/tracing.md`.
Also adding unit tests that include all tracepoint macros to make sure there are no compiler problems with them (e.g. some varadiac extension not supported).
Reviewers might want to check:
- Do the tracepoints still work for you? Do the examples in `contrib/tracing/` run on your system (as bpftrace frequently breaks on every new version, please test master too if it should't work for you)? Do the CI interface tests still pass?
- Is the new documentation clear?
- The `TRACEPOINT_SEMAPHORE(event, context)` macros places global variables in our global namespace. Is this something we strictly want to avoid or maybe move to all `TRACEPOINT_SEMAPHORE`s to a separate .cpp file or even namespace? I like having the `TRACEPOINT_SEMAPHORE()` in same file as the `TRACEPOINT()`, but open for suggestion on alternative approaches.
- Are newly added tracepoints in the unit tests visible when using `readelf -n build/src/test/test_bitcoin`? You can run the new unit tests with `./build/src/test/test_bitcoin --run_test=util_trace_tests* --log_level=all`.
<details><summary>Two of the added unit tests demonstrate that we are only processing the tracepoint arguments when attached by having a test-failure condition in the tracepoint argument preparation. The following bpftrace script can be used to demonstrate that the tests do indeed fail when attached to the tracepoints.</summary>
`fail_tests.bt`:
```c
#!/usr/bin/env bpftrace
usdt:./build/src/test/test_bitcoin:test:check_if_attached {
printf("the 'check_if_attached' test should have failed\n");
}
usdt:./build/src/test/test_bitcoin:test:expensive_section {
printf("the 'expensive_section' test should have failed\n");
}
```
Run the unit tests with `./build/src/test/test_bitcoin` and start `bpftrace fail_tests.bt -p $(pidof test_bitcoin)` in a separate terminal. The unit tests should fail with:
```
Running 594 test cases...
test/util_trace_tests.cpp(31): error: in "util_trace_tests/test_tracepoint_check_if_attached": check false has failed
test/util_trace_tests.cpp(51): error: in "util_trace_tests/test_tracepoint_manual_tracepoint_active_check": check false has failed
*** 2 failures are detected in the test module "Bitcoin Core Test Suite"
```
</details>
These links might provide more contextual information for reviewers:
- [How SystemTap Userspace Probes Work by eklitzke](https://eklitzke.org/how-sytemtap-userspace-probes-work) (actually an example on Bitcoin Core; mentions that with semaphores "the overhead for an untraced process is effectively zero.")
- [libbpf comment on USDT semaphore handling](1596a09b5d/src/usdt.c (L83-L92)) (can recommend the whole comment for background on how the tracepoints and tracing tools work together)
- https://sourceware.org/systemtap/wiki/UserSpaceProbeImplementation#Semaphore_Handling
ACKs for top commit:
willcl-ark:
utACK 0de3e96e33
laanwj:
re-ACK 0de3e96e33
jb55:
utACK 0de3e96e33
vasild:
ACK 0de3e96e33
Tree-SHA512: 0e5e0dc5e0353beaf5c446e4be03d447e64228b1be71ee9972fde1d6fac3fac71a9d73c6ce4fa68975f87db2b2bf6eee2009921a2a145e24d83a475d007a559b
