567cec9a05 doc: add release notes and help text for unix sockets (Matthew Zipkin)
bfe5192891 test: cover UNIX sockets in feature_proxy.py (Matthew Zipkin)
c65c0d0163 init: allow UNIX socket path for -proxy and -onion (Matthew Zipkin)
c3bd43142e gui: accomodate unix socket Proxy in updateDefaultProxyNets() (Matthew Zipkin)
a88bf9dedd i2p: construct Session with Proxy instead of CService (Matthew Zipkin)
d9318a37ec net: split ConnectToSocket() from ConnectDirectly() for unix sockets (Matthew Zipkin)
ac2ecf3182 proxy: rename randomize_credentials to m_randomize_credentials (Matthew Zipkin)
a89c3f59dc netbase: extend Proxy class to wrap UNIX socket as well as TCP (Matthew Zipkin)
3a7d6548ef net: move CreateSock() calls from ConnectNode() to netbase methods (Matthew Zipkin)
74f568cb6f netbase: allow CreateSock() to create UNIX sockets if supported (Matthew Zipkin)
bae86c8d31 netbase: refactor CreateSock() to accept sa_family_t (Matthew Zipkin)
adb3a3e51d configure: test for unix domain sockets (Matthew Zipkin)
Pull request description:
Closes https://github.com/bitcoin/bitcoin/issues/27252
UNIX domain sockets are a mechanism for inter-process communication that are faster than local TCP ports (because there is no need for TCP overhead) and potentially more secure because access is managed by the filesystem instead of serving an open port on the system.
There has been work on [unix domain sockets before](https://github.com/bitcoin/bitcoin/pull/9979) but for now I just wanted to start on this single use-case which is enabling unix sockets from the client side, specifically connecting to a local Tor proxy (Tor can listen on unix sockets and even enforces strict curent-user-only access permission before binding) configured by `-onion=` or `-proxy=`
I copied the prefix `unix:` usage from Tor. With this patch built locally you can test with your own filesystem path (example):
`tor --SocksPort unix:/Users/matthewzipkin/torsocket/x`
`bitcoind -proxy=unix:/Users/matthewzipkin/torsocket/x`
Prep work for this feature includes:
- Moving where and how we create `sockaddr` and `Sock` to accommodate `AF_UNIX` without disturbing `CService`
- Expanding `Proxy` class to represent either a `CService` or a UNIX socket (by its file path)
Future work:
- Enable UNIX sockets for ZMQ (https://github.com/bitcoin/bitcoin/pull/27679)
- Enable UNIX sockets for I2P SAM proxy (some code is included in this PR but not tested or exposed to user options yet)
- Enable UNIX sockets on windows where supported
- Update Network Proxies dialog in GUI to support UNIX sockets
ACKs for top commit:
Sjors:
re-ACK 567cec9a05
tdb3:
re ACK for 567cec9a05.
achow101:
ACK 567cec9a05
vasild:
ACK 567cec9a05
Tree-SHA512: de81860e56d5de83217a18df4c35297732b4ad491e293a0153d2d02a0bde1d022700a1131279b187ef219651487537354b9d06d10fde56225500c7e257df92c1
These come from GUI code, and haven't/aren't being fixed, see discussion
in https://github.com/bitcoin-core/gui/issues/112. For now, just ignore
them entirely. Note that this only applies to ObjCXX code, so will not
hide any relevant warnings coming from C or CXX code (and they would be
unlikely in any case).
Alternative to #29362, which disables all compiler warnings, for macOS
builds in the CI.
Relevant output:
```bash
qt/macnotificationhandler.mm:27:9: warning: 'NSUserNotification' is deprecated: first deprecated in macOS 11.0 - All NSUserNotifications API should be replaced with UserNotifications.frameworks API [-Wdeprecated-declarations]
NSUserNotification* userNotification = [[NSUserNotification alloc] init];
^
/Library/Developer/CommandLineTools/SDKs/MacOSX.sdk/System/Library/Frameworks/Foundation.framework/Headers/NSUserNotification.h:24:12: note: 'NSUserNotification' has been explicitly marked deprecated here
@interface NSUserNotification : NSObject <NSCopying> {
^
qt/macnotificationhandler.mm:27:50: warning: 'NSUserNotification' is deprecated: first deprecated in macOS 11.0 - All NSUserNotifications API should be replaced with UserNotifications.frameworks API [-Wdeprecated-declarations]
NSUserNotification* userNotification = [[NSUserNotification alloc] init];
^
/Library/Developer/CommandLineTools/SDKs/MacOSX.sdk/System/Library/Frameworks/Foundation.framework/Headers/NSUserNotification.h:24:12: note: 'NSUserNotification' has been explicitly marked deprecated here
@interface NSUserNotification : NSObject <NSCopying> {
^
qt/macnotificationhandler.mm:30:11: warning: 'NSUserNotificationCenter' is deprecated: first deprecated in macOS 11.0 - All NSUserNotifications API should be replaced with UserNotifications.frameworks API [-Wdeprecated-declarations]
[[NSUserNotificationCenter defaultUserNotificationCenter] deliverNotification: userNotification];
^
/Library/Developer/CommandLineTools/SDKs/MacOSX.sdk/System/Library/Frameworks/Foundation.framework/Headers/NSUserNotification.h:118:12: note: 'NSUserNotificationCenter' has been explicitly marked deprecated here
@interface NSUserNotificationCenter : NSObject {
^
3 warnings generated.
```
86b7f28d6c serialization: use internal endian conversion functions (Cory Fields)
432b18ca8d serialization: detect byteswap builtins without autoconf tests (Cory Fields)
297367b3bb crypto: replace CountBits with std::bit_width (Cory Fields)
52f9bba889 crypto: replace non-standard CLZ builtins with c++20's bit_width (Cory Fields)
Pull request description:
This replaces #28674, #29036, and #29057. Now ready for testing and review.
Replaces platform-specific endian and byteswap functions. This is especially useful for kernel, as it means that our deep serialization code no longer requires bitcoin-config.h.
I apologize for the size of the last commit, but it's hard to avoid making those changes at once.
All platforms now use our internal functions rather than libc or platform-specific ones, with the exception of MSVC.
Sadly, benchmarking showed that not all compilers are capable of detecting and optimizing byteswap functions, so compiler builtins are instead used where possible. However, they're now detected via macros rather than autoconf checks.
This[ matches how libc++ implements std::byteswap for c++23](https://github.com/llvm/llvm-project/blob/main/libcxx/include/__bit/byteswap.h#L26).
I suggest we move/rename `compat/endian.h`, but I left that out of this PR to avoid bikeshedding.
#29057 pointed out some irregularities in benchmarks. After messing with various compilers and configs for a few weeks with these changes, I'm of the opinion that we can't win on every platform every time, so we should take the code that makes sense going forward. That said, if any real-world slowdowns are caused here, we should obviously investigate.
ACKs for top commit:
maflcko:
ACK 86b7f28d6c📘
fanquake:
ACK 86b7f28d6c - we can finish pruning out the __builtin_clz* checks/usage once the minisketch code has been updated. This is more good cleanup pre-CMake & for the kernal.
Tree-SHA512: 715a32ec190c70505ffbce70bfe81fc7b6aa33e376b60292e801f60cf17025aabfcab4e8c53ebb2e28ffc5cf4c20b74fe3dd8548371ad772085c13aec8b7970e
1. It didn't actually disable asm usage in our code. Regardless of the setting,
asm is used in random.cpp and support/cleanse.cpp.
2. The value wasn't forwarded to libsecp as a user might have reasonably
expected.
3. We now have the DISABLE_OPTIMIZED_SHA256 define which is what disable-asm
actually did in practice.
If there is any desire, we can hook DISABLE_OPTIMIZED_SHA256 up to a new
configure option that actually does what it says.
ad7584d8b6 serialization: replace char-is-int8_t autoconf detection with c++20 concept (Cory Fields)
Pull request description:
Doesn't depend on #29263, but it's really only relevant after that one's merged.
This removes the only remaining autoconf macro in our serialization code (after #29263), so it can now be used trivially and safely out-of-tree.
~Our code does not currently contain any concepts, but couldn't find any discussion or docs about avoiding them. I guess we'll see if this blows up our c-i.~
Edit: Ignore this. ajtowns pointed out that we're already using a few concepts.
This was introduced in #13580. Please check my logic on this as I'm unable to test on a SmartOS system. Even better would be a confirmation from someone who can build there.
ACKs for top commit:
Empact:
Code review ACK ad7584d8b6
Tree-SHA512: 1faf65c900700efb1cf3092c607a2230321b393cb2f029fbfb94bc8e50df1dabd7a9e4b91e3b34f0d2f3471aaf18ee7e56d91869db5c5f4bae84da95443e1120
These replace our platform-specific mess in favor of c++20 endian detection
via std::endian and internal byteswap functions when necessary.
They no longer rely on autoconf detection.
Rather than a complicated set of tests to decide which bswap functions to
use, always prefer the compiler built-ins when available.
These builtins and fallbacks can all be removed once we're using c++23, which
adds std::byteswap.
This avoids cases of missing -O2, when *FLAGS has been overriden.
Removes the need for duplicate code to clear autoconf defaults.
Also, move CORE_CXXFLAGS before DEBUG_CXXFLAGS, so that -O2 is always
overriden if debugging etc.
2d1b1c7dae build: remove --enable-lto (fanquake)
Pull request description:
This has outlived its usefulness, doesn't gel well with newer compilers & `-flto` related options, i.e thin vs full, or `=auto`, and having `-flto` as the only option means that sometimes this just needs to be worked around, i.e in oss-fuzz:
https://github.com/google/oss-fuzz/blob/master/projects/bitcoin-core/build.sh.
While it was convenient when `-flto` was newer, support for `-flto` is now in all compilers we use, and there's also no-longer any real need for us to treat `-flto` different to any other optimization option.
Remove it, to remove build complexity, and so there's no need to port a similar option to CMake.
Note that the LTO option remains in depends, because we still a way to build packages that have LTO specific patches/options.
ACKs for top commit:
TheCharlatan:
ACK 2d1b1c7dae
hebasto:
ACK 2d1b1c7dae.
Tree-SHA512: 91812de7da35346f51850714a188fcffbac478bc8b348bf756c2555fcbde86ba622ac2fb77d294dea0378c741d3656f06121ef3a795aeed63fd170fc31bfa5af
aaaace2fd1 fuzz: Assume presence of __builtin_*_overflow, without checks (MarcoFalke)
fa223ba5eb Revert "build: Fix undefined reference to __mulodi4" (MarcoFalke)
fa7c751bd9 build: Bump clang minimum supported version to 14 (MarcoFalke)
Pull request description:
Most supported operating systems ship with clang-14 (or later), so bump the minimum to that and allow new code to drop workarounds for previous clang bugs.
For reference:
* https://packages.debian.org/bookworm/clang (`clang-14`)
* https://packages.ubuntu.com/jammy/clang (`clang-14`)
* CentOS-like 8/9 Stream: All Clang versions from 15 to 17
* FreeBSD 12/13: All Clang versions from 15 to 16
* OpenSuse Tumbleweed ships with https://software.opensuse.org/package/clang (`clang17`); No idea about OpenSuse Leap
On operating systems where the clang version is not shipped by default, the user would have to use GCC, or install clang in a different way. For example:
* https://packages.debian.org/bullseye/g++ (g++-10)
* https://packages.ubuntu.com/focal/g++-10
* https://apt.llvm.org/, or nix, or guix, or compile clang from source, ...
ACKs for top commit:
fanquake:
ACK aaaace2fd1
Tree-SHA512: 81d066b14cc568d27312f1cc814b09540b038a10a0a8e9d71fc9745b024fb6c32a959af673e6819b817ea7cef98da4abfa63dff16cffb7821b40083016b0291f
This has outlived its usefulness, doesn't gel well with
newer compilers & `-flto` related options, i.e thin vs full, or `=auto`,
and having `-flto` as the only option means that sometimes this just
needs to be worked around, i.e in oss-fuzz:
https://github.com/google/oss-fuzz/blob/master/projects/bitcoin-core/build.sh.
While it was convenient when `-flto` was newer, support for `-flto` is now
in all compilers we use, and there's also no-longer any real need
for us to treat `-flto` different to any other optimization option.
Remove it, to remove build complexity, and so there's no need
to port a similar option to CMake.
Note that the LTO option remains in depends, because we still a way to
build packages that have LTO specific patches/options.
If we decide to merge this, I'll follow up downstream in oss-fuzz first,
to make sure we don't break the build.
49a90915aa build: Bump minimum required Boost to 1.73.0 to support C++20 (Hennadii Stepanov)
Pull request description:
Boost versions <1.73 have C++20-specific bugs that were fixed in the following commits:
- 15fcf21356
- 495c095dc0
I tested [`libboost1.71-dev`](https://packages.ubuntu.com/focal/libboost1.71-dev) in Ubuntu 20.04 and Boost 1.71, 1.72, 1.73 in our depends build system.
Closes https://github.com/bitcoin/bitcoin/issues/29063.
ACKs for top commit:
fanquake:
ACK 49a90915aa
Tree-SHA512: b8ebc08af85abfa3fda70961bd1136ee9e5149dd76a3f901e43acba624d231971873cba5cbf30837f9e5ab58790b8330f241a76cb76d8cf5dce5ad0cca33fba8
308aec3e56 build: disable external-signer for Windows (fanquake)
35537318a1 ci: remove --enable-external-signer from win64 job (fanquake)
Pull request description:
It's come to light that Boost ASIO (a Boost Process sub dep) has in some
instances, been quietly initialising our network stack on Windows (see
PR https://github.com/bitcoin/bitcoin/pull/28486 and discussion in https://github.com/bitcoin/bitcoin/issues/28940).
This has been shielding a bug in our own code, but the larger issue
is that Boost Process/ASIO is running code before main, and doing things
like setting up networking. This undermines our own assumptions about
how our binary works, happens before we run any sanity checks,
and before we call our own code to setup networking. Note that ASIO also
calls WSAStartup with version `2.0`, whereas we call with `2.2`.
It's also not clear why a feature like external signer would have a
dependency that would be doing anything network/socket related,
given it only exists to spawn a local process.
See also the discussion in https://github.com/bitcoin/bitcoin/issues/24907. Note that the maintaince of Boost Process in general,
has not really improved. For example, rather than fixing bugs like https://github.com/boostorg/process/issues/111,
i.e, https://github.com/boostorg/process/pull/317, the maintainer chooses to just wrap exception causing overflows
in try-catch blocks: 0c42a58eac. These changes get merged in large,
unreviewed PRs, i.e https://github.com/boostorg/process/pull/319.
This PR disables external-signer on Windows for now. If, in future, someone
changes how Boost Process works, or replaces it entirely with some
properly reviewed and maintained code, we could reenable this feature on
Windows.
ACKs for top commit:
hebasto:
re-ACK 308aec3e56.
TheCharlatan:
ACK 308aec3e56
Tree-SHA512: 7405f7fc9833eeaacd6836c4e5b1c1a7845a40c1fdd55c1060152f8d8189e4777464fde650e11eb1539556a75dddf49667105987078b1457493ee772945da66e
fa8adbe7c1 build: Enable -Wunreachable-code (MarcoFalke)
Pull request description:
It seems a bit confusing to write code after a `return`. This can even lead to bugs, or incorrect code, such as https://github.com/bitcoin/bitcoin/pull/28830/files#r1415372320 . (Edit: The linked instance is not found by clang's `-Wunreachable-code`).
Fix all issues by enabling `-Wunreachable-code`.
This flag also enables `-Wunreachable-code-loop-increment`, according to https://clang.llvm.org/docs/DiagnosticsReference.html#wunreachable-code, so remove that.
ACKs for top commit:
ajtowns:
> ACK [fa8adbe](fa8adbe7c1)
stickies-v:
ACK fa8adbe7c1
jonatack:
ACK fa8adbe7c1 tested with arm64 clang 17.0.6
Tree-SHA512: 12a2f74b69ae002e62ae08038f7458837090a12051a4c154d05ae4bb26fb19fc1fa76c63aedf2b3fbb36f048c593ca3b8c0efe03fe93cf07a0fd114fc84ce1e7
It's come to light that Boost ASIO (a Boost Process sub dep) has in some
instances, been queitly initialising our network stack on Windows (see
PR #28486 and discussion in #28940).
This has been shielding a bug in our own code, but the larger issue
is that Boost Process/ASIO is running code before main, and doing things
like setting up networking. This undermines our own assumptions about
how our binary works, happens before we get to run any sanity checks,
and also runs before we call our own code to setup networking.
It's also not clear why a feature like external signer would have a
dependency that would be doing anything network/socket related, given it
only exists to spawn a local process.
228d6a2969 build: Fix regression in "ARMv8 CRC32 intrinsics" test (Hennadii Stepanov)
Pull request description:
In the master branch, the `aarch64` binaries lack support for CRC32 intrinsics.
The `vmull_p64` is a part of the Crypto extensions from the ACLE. They are optional extensions, so they get enabled with a `+crypto` for architecture flags.
The regression was introduced in https://github.com/bitcoin/bitcoin/pull/26183 (v25.0).
The `./configure` script log excerpts:
- the master branch @ d752349029:
```
checking whether C++ compiler accepts -march=armv8-a+crc... yes
checking whether C++ compiler accepts -march=armv8-a+crypto... yes
checking for ARMv8 CRC32 intrinsics... no
checking for ARMv8 SHA-NI intrinsics... yes
```
- this PR:
```
checking whether C++ compiler accepts -march=armv8-a+crc+crypto... yes
checking whether C++ compiler accepts -march=armv8-a+crypto... yes
checking for ARMv8 CRC32 intrinsics... yes
checking for ARMv8 SHA-NI intrinsics... yes
```
Guix build:
```
x86_64
2afd81f540c6d3b36ff305e88bafe935e4272cd3efef3130aa69d49a0522541b guix-build-228d6a2969e4/output/aarch64-linux-gnu/SHA256SUMS.part
6c704d6d30d495adb3fb86befdb500eb389a02c1167163f14ab5c3c3e630e6b3 guix-build-228d6a2969e4/output/aarch64-linux-gnu/bitcoin-228d6a2969e4-aarch64-linux-gnu-debug.tar.gz
e4419963c9c0d99adc4e38538900b648f2c14f793b60c8ee2e6f5acc9d3fadd3 guix-build-228d6a2969e4/output/aarch64-linux-gnu/bitcoin-228d6a2969e4-aarch64-linux-gnu.tar.gz
7d11052b6bd28cdf26d5f2a4987f02d32c93a061907bcd048fb6d161a0466ca9 guix-build-228d6a2969e4/output/dist-archive/bitcoin-228d6a2969e4.tar.gz
```
ACKs for top commit:
TheCharlatan:
ACK 228d6a2969
Tree-SHA512: 4c27ca8acb953bf56e972d907a282ee19e3f30f7a4bf8a9822395fe0e28977cd6233e8b65b4a25cc1d3d5ff6a796d7af07653e18531c44ee3efaff1563d96d32
f95af98128 guix: default ssp for Windows GCC (fanquake)
95d55b96c2 guix: remove ssp workaround from Windows GCC (fanquake)
8f43302a0a build: remove explicit libssp linking from Windows build (fanquake)
Pull request description:
I was expecting this to fail to compile somewhere, maybe in the CI, but that doesn't seem to be the case?
Seems workable given the SSP related changes in the newer mingw-w64 headers (which are in Guix):
> Implement some of the stack protector functions/variables so -lssp is now optional when _FORTIFY_SOURCE or -fstack-protector-strong is used.
However I think this would still be broken in some older environments, so we might have to wait for a compiler bump, or similar. The optional -lssp also seems to work when using older headers, which doesn't make sense.
Would fix#28104.
ACKs for top commit:
hebasto:
ACK f95af98128, I've verified binaries from `bitcoin-f95af98128f1-win64.zip` on Windows 11 Pro 23H2.
TheCharlatan:
ACK f95af98128
Tree-SHA512: 71169ec513cfe692dfa7741d2bf37b45da05627c0af1cbd50cf8c3c04cc21c4bf88f3284532bddc1e3e648391ec78dbaca5170987a13c21ac204a7bcaf27f349
The `vmull_p64` is a part of the Crypto extensions from the ACLE. They
are optional extensions, so they get enabled with a `+crypto` for
architecture flags.
This is deprecated on macOS:
```bash
ld: warning: -bind_at_load is deprecated on macOS
```
and likely redundant anyways, given the behaviour of dyld3.
Unfortunately libtool is still injecting a `-bind_at_load`:
```bash
# Don't allow lazy linking, it breaks C++ global constructors
# But is supposedly fixed on 10.4 or later (yay!).
if test CXX = "$tagname"; then
case ${MACOSX_DEPLOYMENT_TARGET-10.0} in
10.[0123])
func_append compile_command " $wl-bind_at_load"
func_append finalize_command " $wl-bind_at_load"
;;
esac
fi
```
so this doesn't remove all the warnings, but removes us as a potential
source of them.
Note that anywhere the ld64 warnings are being emitted, we are already
not adding this flag to our hardened ldflags, because of `-Wl,-fatal_warnings`.
Having the link check in the header check loop means we get `-lminiupnpc
-lminiupnpc -lminiupnpc` on the link line. This is unnecessary, and
results in warnings, i.e:
```bash
ld: warning: ignoring duplicate libraries: '-levent', '-lminiupnpc'
ld: warning: ignoring duplicate libraries: '-levent', '-lminiupnpc'
ld: warning: ignoring duplicate libraries: '-levent', '-lminiupnpc'
```
These warnings have been occurring since the new linker released with
Xcode 15, and also came up in https://github.com/hebasto/bitcoin/pull/34.
fa25e8b0a1 doc: Recommend lint image build on every call (MarcoFalke)
faf70c1f33 Bump python minimum version to 3.9 (MarcoFalke)
fa8996b930 ci: Bump i686_multiprocess.sh to latest Ubuntu LTS (MarcoFalke)
Pull request description:
All supported operating systems ship with python 3.9 (or later), so bumping the minimum should not cause any issues. A bump will allow new code to use new python 3.9 features.
For reference:
* https://packages.debian.org/bullseye/python3
* https://packages.ubuntu.com/focal/python3.9
* FreeBSD 12/13 also ships with 3.9
* CentOS-like 8/9 also ships with 3.9 (and 3.11)
* OpenSuse Leap also ships with 3.9 (and 3.11) https://software.opensuse.org/package/python311-base
This is for Bitcoin Core 27.0 in 2024 (next year), not the soon upcoming 26.0 next month.
ACKs for top commit:
Sjors:
ACK fa25e8b0a1
jamesob:
ACK fa25e8b0a1 ([`jamesob/ackr/28211.1.MarcoFalke.bump_python_minimum_supp`](https://github.com/jamesob/bitcoin/tree/ackr/28211.1.MarcoFalke.bump_python_minimum_supp))
Tree-SHA512: 86c9f6ac4b5ba94a62ee6a6062dd48a8295d8611a39cdb5829f4f0dbc77aaa1a51edccc7a99275bf699143ad3a6fe826de426d413e5a465e3b0e82b86d10c32e
This is not a hardening specific flag, it should be used at all times,
regardless of if hardening is enabled or not. Note that this was
still the case here, but having this exist in the hardening flags is
confusing, and may lead someone to move it inside one of the `use_hardening`
blocks, where it would become unused, with `--disable-hardening`.
This is a simpler (less hardening) version of #24123.
Scoped to aarch64 to avoid unused command line option warnings when
building on x86_64.
Related to #19075.
We currently work around a longstanding GCC issue with aligned vector
instructions, in our release builds, by patching the behaviour we want
into GCC (see discussion in #24736).
A new option now exists in the binutils assembler,
`-muse-unaligned-vector-move`, which should also achieve the behaviour
we want (at least for our code). This was added in the 2.38 release,
see
https://sourceware.org/git/?p=binutils-gdb.git;a=commit;h=c8480b58e1968f209b6365af7422678f348222c2.
```bash
x86: Add -muse-unaligned-vector-move to assembler
Unaligned load/store instructions on aligned memory or register are as
fast as aligned load/store instructions on modern Intel processors. Add
a command-line option, -muse-unaligned-vector-move, to x86 assembler to
encode encode aligned vector load/store instructions as unaligned
vector load/store instructions.
```
Even if we introduce this option into our build system, we'll have to
maintain our GCC patching, as we want all code that ends up in the
binary, to avoid these instructions. However, there may be some value in
adding the option, as it could be an improvement for someone building
(bitcoind.exe) with an unpatched compiler.
08eb5f1b67 ci: document that -Wreturn-type has been fixed upstream (Windows) (fanquake)
Pull request description:
`noreturn` attributes have been added to the mingw-w64 headers, 1690994f51, meaning that [from 11.0.0 onwards](https://www.mingw-w64.org/changelog/), you'll no-longer see `-Wreturn-type` warnings when using `assert(false)`.
Add -Wno-return-type to the Windows CI, where is should have been all
along, and document why it's required. This can be dropped when we are
using the fixed version of the mingw-w64 headers there.
Drop the -Werror -Wno-return-type special case from our build system.
-Wreturn-type is on by default in Clang and GCC.
The new mingw-w64 header behaviour can be checked on Ubuntu mantic, [which ships with 11.0.0](https://packages.ubuntu.com/mantic/mingw-w64), using:
```cpp
#include <cassert>
int f(){ assert(false); }
int main() {
return 0;
}
```
On Mantic (with 11.0.0):
```bash
x86_64-w64-mingw32-g++ test.cpp -Wreturn-type
# nada
```
On Lunar ([with 10.0.0](https://packages.ubuntu.com/lunar/mingw-w64)):
```bash
x86_64-w64-mingw32-g++ test.cpp -Wreturn-type
test.cpp: In function 'int f()':
test.cpp:3:25: warning: no return statement in function returning non-void [-Wreturn-type]
3 | int f(){ assert(false); }
| ^
```
ACKs for top commit:
TheCharlatan:
ACK 08eb5f1b67
Tree-SHA512: 9cd4310a96abd87bf8ceb37949ad0259fe4adee3367c604f4c4ad521a0cf09bdcc5dd305db19a0f45ce74c85178b0d739e2fca5ad0fc841ac935523a23b28a7f
