060a2a64d40d75fecb60b7d2b9946a67e46aa6fc ci: remove boost thread installation (fanquake)
06e1d7d81d5a56d136c6fc88f09a2b0654a164f9 build: don't build or use Boost Thread (fanquake)
7097add83c8596f81be9edd66971ffd2486357eb refactor: replace Boost shared_mutex with std shared_mutex in sigcache (fanquake)
8e55981ef834490c438436719f95cbaf888c4914 refactor: replace Boost shared_mutex with std shared_mutex in cuckoocache tests (fanquake)
Pull request description:
This replaces `boost::shared_mutex` and `boost::unique_lock` with [`std::shared_mutex`](https://en.cppreference.com/w/cpp/thread/shared_mutex) & [`std::unique_lock`](https://en.cppreference.com/w/cpp/thread/unique_lock).
Even though [some concerns were raised](https://github.com/bitcoin/bitcoin/issues/16684#issuecomment-726214696) in #16684 with regard to `std::shared_mutex` being unsafe to use across some glibc versions, I still think this change is an improvement. As I mentioned in #21022, I also think trying to restrict standard library feature usage based on bugs in glibc is not only hard to do, but it's not currently clear exactly how we do that in practice (does it also extend to patching out use in our dependencies, should we be implementing more runtime checks for features we are using, when do we consider an affected glibc "old enough" not to worry about? etc). If you take a look through the [glibc bug tracker](https://sourceware.org/bugzilla/describecomponents.cgi?product=glibc) you'll no doubt find plenty of (active) bug reports for standard library code we already using. Obviously not to say we shouldn't try and avoid buggy code where possible.
Two other points:
[Cory mentioned in #21022](https://github.com/bitcoin/bitcoin/pull/21022#issuecomment-769274179):
> It also seems reasonable to me to worry that boost hits the same underlying glibc bug, and we've just not happened to trigger the right conditions yet.
Moving away from Boost to the standard library also removes the potential for differences related to Boosts configuration. Boost has multiple versions of `shared_mutex`, and what you end up using, and what it's backed by depends on:
* The version of Boost.
* The platform you're building for.
* Which version of `BOOST_THREAD_VERSION` is defined: (2,3,4 or 5) default=2. (see [here](https://www.boost.org/doc/libs/1_70_0/doc/html/thread/build.html#thread.build.configuration) for some of the differences).
* Is `BOOST_THREAD_V2_SHARED_MUTEX` defined? (not by default). If so, you might get the ["less performant, but more robust"](https://github.com/boostorg/thread/issues/230#issuecomment-475937761) version of `shared_mutex`.
A lot of these factors are eliminated by our use of depends, but users will have varying configurations. It's also not inconceivable to think that a distro, or some package manager might start defining something like `BOOST_THREAD_VERSION=3`. Boost tried to change the default from 2 to 3 at one point.
With this change, we no longer use Boost Thread, so this PR also removes it from depends, the build system, CI etc.
Previous similar PRs were #19183 & #20922. The authors are included in the commits here.
Also related to #21022 - pthread sanity checking.
ACKs for top commit:
laanwj:
Code review ACK 060a2a64d40d75fecb60b7d2b9946a67e46aa6fc
vasild:
ACK 060a2a64d40d75fecb60b7d2b9946a67e46aa6fc
Tree-SHA512: 572d14d8c9de20bc434511f20d3f431836393ff915b2fe9de5a47a02dca76805ad5c3fc4cceecb4cd43f3ba939a0508178c4e60e62abdbaaa6b3e8db20b75b03
we're currently using boost 1.77, this replacement was dropped in its
patch form in bitcoin#19764 (backported as 6a367fc1 in dash#4356), the
comment was dropped in bitcoin#19761 but the change is in this commit
instead due to direct relevance.
edc9a6afdc6926b40de2fed897b9b866d58f28d2 build, refactor: Reuse expat package version in its download path (Hennadii Stepanov)
4bb7821ab243325467a89ff8ffc1bed290f2cb58 build, refactor: Use conventional version notation for boost package (Hennadii Stepanov)
Pull request description:
`boost` package:
- `.` is used as a separator in versions of other depends packages.
`expat` package:
- reuse package version in its download path
---
The straightforward way to verify this PR:
```
$ cd depends
$ make clean-all
$ make boost_fetched
$ make expat_fetched
```
ACKs for top commit:
prusnak:
ACK edc9a6a
shaavan:
ACK edc9a6afdc6926b40de2fed897b9b866d58f28d2
Tree-SHA512: c15d672fe34ac59850425d3d6a6eee5f720e16d227aad1332a563b218465879b7ee6fb865dd1bac06aedf356f9bb1c67112d9d88da8f877f04838b50a9dc97be
22c5a986e95d2bd14273465ca0e15fbe3772252d depends: Consistent use of package variable (Peter Bushnell)
Pull request description:
All other mk files use the package variable consistently except for the two instances here, which have always been here, since depends was introduced in 0.10.
ACKs for top commit:
fanquake:
ACK 22c5a986e95d2bd14273465ca0e15fbe3772252d - tested a `make boost -C depends/ -j8`.
Tree-SHA512: 41766a328603db2ebb1f23ea0c5b2936de043587dd86396eaba73524d2f5bdeff25447040e33d61de2ef612a920281cd81c6fac097913270287f344beb839c5d
f447a0a7079619f0d650084df192781cca9fd826 Remove program options from build system (Chun Kuan Lee)
11588c639e8912f1b28e981c1a2a0e4306dbd093 Replace boost program_options (Chun Kuan Lee)
Pull request description:
Concept from #12744, but without parsing negated options.
Tree-SHA512: 7f418744bb8934e313d77a5f162633746ef5d043de802b9c9cd9f7c1842e7e566eb5f171cd9e2cc13317281b2449c6fbd553fa4f09b837e6af2f5d2b2aabdca2
50037e97d11356218c4b36767232e47b74742b0b depends: fix boost mac cross build with clang 9+ (Cory Fields)
Pull request description:
The ancient "darwin-4.9.1" profile has long been used to match against clang, which prior to version 9, reported 4.9.1 as its version when invoking "clang++ -dumpversion". Presumably this was a historical compatibility quirk related to Apple's switch from gcc to clang.
This was "fixed" in clang 9.0, so that -dumpversion reports the real version. Unfortunately that had the side-effect of breaking the (brittle) boost compiler detection.
Move to the seemingly more-correct "clang-darwin" profile, which passes the checks and builds correctly.
Also switch to using ar rather than libtool for archiving, as it's what the clang-darwin profile expects to be using.
Note that because this is using a different profile, some of the final command-line arguments end up changing. Those changes look sane at a glance.
ACKs for top commit:
fanquake:
ACK 50037e97d11356218c4b36767232e47b74742b0b - tested on on macOS, will wait for the gitian build.
Tree-SHA512: eac1f353513a445add6fbece7fc78dd3dbdde5e2219bfb7739b82f40bb14de449667a94d2e303d43c67d9b38e7ceb0ba5f0d8fe20b40be2017b1ca0875467c2c
2620e24b83d16bf0f2bfe360dee1e98b4be59ca5 [depends] boost: update to 1.70 (Sjors Provoost)
Pull request description:
Version [1.70](https://www.boost.org/users/history/version_1_70_0.html) is most recent.
Versions needed for:
* 1.66: #12557: fixes the single arm64 configuration ([06ee5b5](06ee5b54ef))
ACKs for commit 2620e2:
Tree-SHA512: 6e0174f1d92c2c24314c0689d4809e048914f8f42d17aa73799f5ee232169e0dd0ed71f5f973903c44c08309f2837c629c493f15e5c31ec6c7bd1daae5f3b25f
* build: Enable C++11 build, require C++11 compiler
Implements #6211.
* depends: use c++11
* build: update ax_cxx_compile_stdcxx to serial 4
* build: force a c++ standard to be specified
Newer compilers may switch to newer standards by default. For example, gcc6
uses std=gnu++14 by default.
* c++11: fix libbdb build against libc++ in c++11 mode
atomic_init clashes with
* c++11: CAccountingEntry must be defined before use in a list
c++11ism. This fixes builds against libc++.
Boost assumes variadic templates are always available in GCC 4.4+, but
they aren't since we don't build with -std=c++11.
This applies the patch that fixed the issue in boost 1.57:
eec8085549
See also: https://svn.boost.org/trac/boost/ticket/10500
tl;dr: This solves boost visibility problems for default/release build configs
on non-Linux platforms.
When Bitcoin builds against boost's header-only classes, it ends up with
objects containing symbols that the upstream boost libs also have. Since
Bitcoin builds by default with hidden symbol visibility, it can end up trying
to link against a copy of the same symbols with default visibility.
This is not a problem on Linux because 3rd party static libs are un-exported
by default (--exclude-libs,ALL), but that is not available for MinGW and OSX.
Those platforms (and maybe others?) end up confused about which version to use.
The OSX linker spews hundreds of: "ld: warning: direct access in <foo> to
global weak symbol guard variable for <bar> means the weak symbol cannot be
overridden at runtime. This was likely caused by different translation units
being compiled with different visibility settings."
MinGW's linker complains similarly.
Since the default symbol visibility for Bitcoin is hidden and releases are
built that way as well, build Boost with hidden visibility. Linux builds Boost
this way also, but only for the sake of continuity.
This means that the linker confusion logic is reversed, so the problem will
will now be encountered if Bitcoin is built with --disable-reduce-exports, but
that's better than the current situation.
