Starting with version 13.x, the mingw headers will define the value of
NTDDI_VERSION, based on the value of _WIN32_WINNT, if that version is <
Windows 10. Given that libevent was undefining our _WIN32_WINNT, and
redefining it to a value < Windows 10 (0x0501), NTDDI_VERSION was also
being defined to that value, leading to functions not being exposed in
the mingw-w64 headers; see here:
9c2668ef77/mingw-w64-headers/include/iphlpapi.h (L36-L41).
Imports a commit from usptream (a14ff91254f40cf36e0fee199e26fb11260fab49).
Fixes#32707.
Github-Pull: #32837
Rebased-From: f5647c6c5a
When using CMake policies 3.14 and below, the `export(PACKAGE)` command
by default populates the user package registry, which is stored outside
the build tree. Setting the `CMAKE_EXPORT_NO_PACKAGE_REGISTRY` variable
disables this side effect.
In CMake 3.15 and later, this behavior is disabled by default, and the
variable has no effect.
Github-Pull: #32943
Rebased-From: 44f3bae300
When building depends on FreeBSD/OpenBSD `aarch64`, the host compilers
default to `default_host_{CC,CXX}`, which resolves to `gcc`/`g++`. This
is incorrect on these systems, where Clang is the default system
compiler.
Github-Pull: #32716
Rebased-From: 4f10a57671
Nix patches cmake to remove the root directory `/` from
`CMAKE_SYSTEM_PREFIX_PATH`:
428b49b28e/pkgs/by-name/cm/cmake/001-search-path.diff (L10)
Without this, and when using the toolchain for depends builds, cmake's
`find_path()` and `find_package()` do not know where to find
dependencies, causing issues like:
https://github.com/bitcoin/bitcoin/issues/32428
Adding this path back via CMAKE_PREFIX_PATH is harmless on other
systems, and fixes the toolchain for Nix users.
We append the `/` dir a maximum of once, as the toolchain may be called
repeatedly during builds.
Co-authored-by: Russell Yanofsky <russ@yanofsky.org>
Co-authored-by: josibake <josibake@protonmail.com>
Github-Pull: #32798
Rebased-From: e27a94596f
On OpenBSD, the `sha256` command by default outputs hashsums on files in
"BSD" mode, looking like this:
$ sha256 ~/.vimrc
SHA256 (/home/thestack/.vimrc) = 6ba69d100e8c5ca0488ded6293d4e5f740a6a5d5ace96cbcf0599c18d27389e4
This is not compatible with our depends commands, which expect the
hashes to be on the first column (to be extracted via `cut -d" " -f1`).
Fix this by switching to GNU mode output, looking like this:
$ sha256 -r ~/.vimrc
6ba69d100e8c5ca0488ded6293d4e5f740a6a5d5ace96cbcf0599c18d27389e4 /home/thestack/.vimrc
Without this change, the multiprocess depends build fails with the following output:
$ gmake -C depends MULTIPROCESS=1 NO_BOOST=1 NO_LIBEVENT=1 NO_QT=1 NO_QR=1 NO_WALLET=1 NO_ZMQ=1 NO_USDT=1
[ ..... ]
Extracting native_libmultiprocess...
sha256: /home/thestack/bitcoin/depends/work/build/x86_64-unknown-openbsd7.7/native_libmultiprocess/-2bc902f4693/.src-ipc-libmultiprocess.tar.hash: no properly formatted checksum lines found
gmake: *** [funcs.mk:342: /home/thestack/bitcoin/depends/work/build/x86_64-unknown-openbsd7.7/native_libmultiprocess/-2bc902f4693/.stamp_extracted] Error 1
Github-Pull: #32690
Rebased-From: 8713e8060d
It looks like the mkdir detection in xproto is broken on Alpine. Ensure
we always use `mkdir -p`.
Fixes#32494.
Github-Pull: #32568
Rebased-From: df9ebbf659
According to the CMake docs, this is the correct way to setup a
toolchain file for cross-compilation using Clang. See
https://cmake.org/cmake/help/latest/manual/cmake-toolchains.7.html#cross-compiling-using-clang
Internally it looks like CMake will only take this variable into account
if it detects the compiler to be Clang, so this shouldn't effect other
builds, but in the case of our Apple cross builds, we'd end up with a
duplicated `--target=arm64-apple-darwin` on the compiler line, given we
are already setting `--target` for Darwin builds.
Would fix#31748.
Github-Pull: #31849
Rebased-From: 963355037f
From the GNU make 3.82 release announcement:
* The 'define' make directive now allows a variable assignment operator
after the variable name, to allow for simple, conditional, or appending
multi-line variable assignment.
macOS ships with 3.81. This caused the multiprocess config options
to be ignored.
Fixes#32068
Github-Pull: #32070
Rebased-From: 9157d9e449
Co-authored-by: Ryan Ofsky <ryan@ofsky.org>
Similar to #31840, currently our Linux toolchain file contains:
```bash
set(CMAKE_AR "aarch64-linux-gnu-ar")
set(CMAKE_RANLIB "aarch64-linux-gnu-ranlib")
set(CMAKE_STRIP "aarch64-linux-gnu-strip")
set(CMAKE_OBJCOPY "aarch64-linux-gnu-objcopy")
set(CMAKE_OBJDUMP "")
```
`objdump` is currently only used for the macOS cross build, where it's
`llvm-objdump`, but we should be consistent in producing a toolchain
file that points to actual tools, rather than leaving variables unset.
3edaf0b428 depends: add missing Darwin objcopy (fanquake)
Pull request description:
Our CMake toolchain for a Darwin cross build currently contains:
```bash
set(CMAKE_AR "/usr/bin/llvm-ar")
set(CMAKE_RANLIB "/usr/bin/llvm-ranlib")
set(CMAKE_STRIP "/usr/bin/llvm-strip")
set(CMAKE_OBJCOPY "arm64-apple-darwin-objcopy")
set(CMAKE_OBJDUMP "/usr/bin/llvm-objdump")
```
`objcopy` isn't currently used for the Darwin build (only for Linux and splitting the debug symbols), but we shouldn't be producing a toolchain file that refers to nonexistent tools.
ACKs for top commit:
laanwj:
Code review ACK 3edaf0b428
theuni:
utACK 3edaf0b428
Tree-SHA512: b74deb9f3f053c37d03505e698419d4a14131131f12a042dab698a81f2ad76b71fd55c1d1afd5f5822cc50fdaad5acdab15a8b20626c56f705179add1165449f
f89f16846e depends: Fix compiling `libevent` package on NetBSD (Hennadii Stepanov)
Pull request description:
Libevent [introduced](https://github.com/libevent/libevent/pull/909) the [`typeof`](https://gcc.gnu.org/onlinedocs/gcc/Typeof.html) C language extension in the NetBSD-specific code, which was pulled into our depends in https://github.com/bitcoin/bitcoin/pull/21991.
However, GCC [states](https://gcc.gnu.org/onlinedocs/gcc/Alternate-Keywords.html):
> the various `-std` options disable certain keywords.
Due to our use of b042c4f053/depends/hosts/netbsd.mk (L1)
the `typeof` keyword is disabled, resulting in a compilation error:
```
$ gmake -C depends libevent CC=/usr/pkg/gcc14/bin/gcc CXX=/usr/pkg/gcc14/bin/g++
<snip>
[ 37%] Building C object CMakeFiles/event_core_static.dir/kqueue.c.o
/home/hebasto/dev/bitcoin/depends/work/build/x86_64-unknown-netbsd10.0/libevent/2.1.12-stable-ca6b96ec97c/kqueue.c: In function 'kq_setup_kevent':
/home/hebasto/dev/bitcoin/depends/work/build/x86_64-unknown-netbsd10.0/libevent/2.1.12-stable-ca6b96ec97c/kqueue.c:56:27: error: implicit declaration of function 'typeof' [-Wimplicit-function-declaration]
56 | #define INT_TO_UDATA(x) ((typeof(((struct kevent *)0)->udata))(intptr_t)(x))
| ^~~~~~
/home/hebasto/dev/bitcoin/depends/work/build/x86_64-unknown-netbsd10.0/libevent/2.1.12-stable-ca6b96ec97c/kqueue.c:190:30: note: in expansion of macro 'INT_TO_UDATA'
190 | out->udata = INT_TO_UDATA(ADD_UDATA);
| ^~~~~~~~~~~~
/home/hebasto/dev/bitcoin/depends/work/build/x86_64-unknown-netbsd10.0/libevent/2.1.12-stable-ca6b96ec97c/kqueue.c:56:64: error: expected expression before 'intptr_t'
56 | #define INT_TO_UDATA(x) ((typeof(((struct kevent *)0)->udata))(intptr_t)(x))
| ^~~~~~~~
/home/hebasto/dev/bitcoin/depends/work/build/x86_64-unknown-netbsd10.0/libevent/2.1.12-stable-ca6b96ec97c/kqueue.c:190:30: note: in expansion of macro 'INT_TO_UDATA'
190 | out->udata = INT_TO_UDATA(ADD_UDATA);
| ^~~~~~~~~~~~
/home/hebasto/dev/bitcoin/depends/work/build/x86_64-unknown-netbsd10.0/libevent/2.1.12-stable-ca6b96ec97c/kqueue.c:56:27: error: called object is not a function or function pointer
56 | #define INT_TO_UDATA(x) ((typeof(((struct kevent *)0)->udata))(intptr_t)(x))
| ~^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
/home/hebasto/dev/bitcoin/depends/work/build/x86_64-unknown-netbsd10.0/libevent/2.1.12-stable-ca6b96ec97c/kqueue.c:190:30: note: in expansion of macro 'INT_TO_UDATA'
190 | out->udata = INT_TO_UDATA(ADD_UDATA);
| ^~~~~~~~~~~~
gmake[3]: *** [CMakeFiles/event_core_static.dir/build.make:328: CMakeFiles/event_core_static.dir/kqueue.c.o] Error 1
<snip>
```
This PR resolves this issue by following GCC's [recommendation](https://gcc.gnu.org/onlinedocs/gcc/Typeof.html):
> write `__typeof__` instead of `typeof`.
ACKs for top commit:
fanquake:
ACK f89f16846e
Tree-SHA512: c0d2e535408db120535781f8518c616b0f5a39b1c6babb2a74e8e0565348aaf00b0f5a93cac0af7cf6d6bf028d5d58763fe71b3969ed9c7059fa7c3dca9d084c
b28917be36 depends: Make default `host` and `build` comparable (Hennadii Stepanov)
Pull request description:
To detect cross-compiling, the host and build platforms are compared. The `build` variable is always an output of `config.sub`, but the `host` is not. This can lead to false results. For example, on OpenBSD:
- host=amd64-unknown-openbsd7.5
- build=x86_64-unknown-openbsd7.5
This PR sets the default value of the `host` variable to the value of `build`, ensuring cross-compiling won't be triggered when the `HOST` variable is not set.
This PR fixes needless triggering of cross-compiling for CMake-built packages in depends on OpenBSD due to this code:eb85cacd29/depends/funcs.mk (L193-L197)
No changes in Guix build.
ACKs for top commit:
laanwj:
Concept and code review ACK b28917be36
theuni:
utACK b28917be36.
Tree-SHA512: 8c5835cb8b739355b71f7cb161b350ad8b038a00e6b1def36354ba228cea3dcb9883df3c9a8e79d7d0143241f6f054129fe90772b1b2579702db51237f9d66ff
Our CMake toolchain for a Darwin cross build currently contains:
```bash
set(CMAKE_AR "/usr/bin/llvm-ar")
set(CMAKE_RANLIB "/usr/bin/llvm-ranlib")
set(CMAKE_STRIP "/usr/bin/llvm-strip")
set(CMAKE_OBJCOPY "arm64-apple-darwin-objcopy")
set(CMAKE_OBJDUMP "/usr/bin/llvm-objdump")
```
`objcopy` isn't currently used for the Darwin build (only for Linux and
splitting the debug symbols), but we shouldn't be producing a toolchain
file that refers to nonexistent tools.
The `-ffile-prefix-map` compiler option implicitly enables
`-fprofile-prefix-map` in GCC or `-fcoverage-prefix-map` in Clang, which
can cause issues with coverage builds.
This change ensures that only the options necessary for build
reproducibility are applied.
d9c8aacce3 depends, refactor: Avoid hardcoding `host_prefix` in toolchain file (Hennadii Stepanov)
Pull request description:
This PR allows the entire `depends/<host_prefix>` directory to be relocatable.
Only `libevent` package configuration files are non-relocatable for the version `2.1.12-stable` we use now. However, this issue has been fixed upstream in 1f1593ff27 and friends.
Fixes https://github.com/bitcoin/bitcoin/issues/31050.
ACKs for top commit:
theuni:
Neat. utACK d9c8aacce3
ryanofsky:
Code review ACK d9c8aacce3
Tree-SHA512: c4c340722e63fc1da36fba2b15f025a6e1d477da1991194d678f546f2f3ea9454e2f0ff054aae6ae6c332a0781a597c3ce63b9018b46b8c258033f0d19efbef3
The `libevent` package defaults to the "Release" build type, which
overrides our per-build-type optimization flags with `-O3`.
To prevent this behavior, set `CMAKE_BUILD_TYPE` to "None", consistent
with how other packages are handled.
01df180bfb depends: add mold & ld.lld to gen_id (fanquake)
d032ac8063 depends: add *FLAGS to gen_id (fanquake)
Pull request description:
The depends cache should be busted when flags change, the same as any other tooling change. I'd also like to start passing `*FLAGS` into depends inside the Guix env, which, without this change, doesn't bust the cache.
ACKs for top commit:
hebasto:
ACK 01df180bfb.
Tree-SHA512: 3809359fe763af9dde484e0c6bd3e262c4c09fcbe2f96ccf64194f5f9f840f5476b9c9929cf7bda7b8c14efeffd369cdb8c233625b79a944e1380df20698246f
31a0e5f090 depends: Qt 5.15.16 (fanquake)
Pull request description:
Contains a handful of miscellaneous bug fixes.
We can drop a few of our patches.
See https://github.com/qt/qtbase/compare/v5.15.14-lts-lgpl...v5.15.16-lts-lgpl.
ACKs for top commit:
hebasto:
ACK 31a0e5f090.
TheCharlatan:
ACK 31a0e5f090
Tree-SHA512: dd7b3332dd6ecb95189bc72364883425fb8869e03850791d2ee92555a37046c7abaaee16575a0396f1ce9674856b894563dbd36868c2cf46f9fee48028fd967b
According to the description for pkg-config, "pkgconf is a
replacement for pkg-config, providing additional functionality
while also maintaining compatibility. This package only provides
a dependency link to the pkgconf package to help with package
upgrades. It can be safely removed."
Thus several scripts and markdown files are updated.
We use `lld` when cross-compiling for macOS, and it's version should
be tied to LLVM. However someone compiling with GCC and `-fuse-ld=lld`
would not see a cache bust if the LLVM toolchain was updated.
We don't use `mold` directly, but I'm aware of it's usage in
infrastructure, along with depends, used to test the project.
The depends cache should be busted when flags change, the same as any
other tooling change. Id also like to start passing *FLAGS into depends
inside the Guix env, which, without this change, doesn't bust the cache.
On FreeBSD, the `shasum` utility is provided by the `perl5` port, which
is not part of the base system and must be installed separately.
Note that this requirement is currently not documented in
`depends/README.md`.
This change switches to using the `sha256sum` utility, which is included
in the base system.
This change resolves an issue where a missing space caused the value of
the `build_AR` variable to be concatenated with the "NM=" string. This
resulted in subsequent calls to `${AR}` and `${NM}` failing.
a10bb400e8 depends: Fix CXXFLAGS on NetBSD (Hennadii Stepanov)
Pull request description:
This PR corrects an issue where `CXXFLAGS` were mistakenly overridden by `CFLAGS`. This behaviour was introduced in 7e7b3e42fa (from https://github.com/bitcoin/bitcoin/pull/22380).
On the master branch:
```
$ gmake --no-print-directory -C depends print-x86_64_netbsd_CXXFLAGS
x86_64_netbsd_CXXFLAGS=-pipe -std=c11
```
With this PR:
```
$ gmake --no-print-directory -C depends print-x86_64_netbsd_CXXFLAGS
x86_64_netbsd_CXXFLAGS=-pipe -std=c++20
```
ACKs for top commit:
theuni:
utACK a10bb400e8
Tree-SHA512: 0c842db2965ebb0a58693394715922810235d9e5f2a7416fe258eb252dbd68ec04f90a0f7948abe938caf94a9194cca7deb53a08335c4404cce3a40c5cb44944
ee1128ead8 doc: update stack-clash-protection comment re mingw-w64 (fanquake)
bf47448f15 test: drop check for Windows < 10 (fanquake)
35b898c47f release: target Windows 10 or later (fanquake)
398754e70b depends: target Windows 10 when building for mingw-w64 (fanquake)
Pull request description:
Follows up to https://github.com/bitcoin/bitcoin/pull/31048#discussion_r1803165670.
We definitely cannot claim that Bitcoin Core is "supported and extensively tested on" on Windows 7.
Note that #30997 is also increasing the minimum required Windows version (for the GUI) to 10.
ACKs for top commit:
hodlinator:
cr-ACK ee1128ead8
davidgumberg:
ACK ee1128ead8
achow101:
ACK ee1128ead8
hebasto:
re-ACK ee1128ead8, only rebased, a commit message and a comment have been amended since my recent [review](https://github.com/bitcoin/bitcoin/pull/31172#pullrequestreview-2415452160).
TheCharlatan:
ACK ee1128ead8
Tree-SHA512: 245e0bac3d63414d919a1948661fef4ff79359faaacaf19d64abd91cc62e822797fb1cf3379e340bfdf9a85c0b88fd99a90eda450dd4218b6213ab78aefb1374
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
fanquake
Hennadii Stepanov
will
josibake
Sebastian Falbesoner
Sjors Provoost
Ryan Ofsky
Adlai Chandrasekhar
TheCharlatan
Ava Chow