8f7704d032 build: improve detection of eBPF support (fanquake)
Pull request description:
Just checking for the `sys/sdt.h` header isn't enough, as systems like macOS have the header, but it doesn't actually have the `DTRACE_PROBE*` probes, which leads to [compile failures](https://github.com/bitcoin/bitcoin/pull/22006#issuecomment-859559004). The contents of `sys/sdt.h` in the macOS SDK is:
```bash
#ifndef _SYS_SDT_H
#define _SYS_SDT_H
/*
* This is a wrapper header that wraps the mach visible sdt.h header so that
* the header file ends up visible where software expects it to be. We also
* do the C/C++ symbol wrapping here, since Mach headers are technically C
* interfaces.
*
* Note: The process of adding USDT probes to code is slightly different
* than documented in the "Solaris Dynamic Tracing Guide".
* The DTRACE_PROBE*() macros are not supported on Mac OS X -- instead see
* "BUILDING CODE CONTAINING USDT PROBES" in the dtrace(1) manpage
*
*/
#include <sys/cdefs.h>
__BEGIN_DECLS
#include <mach/sdt.h>
__END_DECLS
#endif /* _SYS_SDT_H */
```
The `BUILDING CODE CONTAINING USDT PROBES` section from the dtrace manpage is available [here](https://gist.github.com/fanquake/e56c9866d53b326646d04ab43a8df9e2), and outlines the more involved process of using USDT probes on macOS.
ACKs for top commit:
jb55:
utACK 8f7704d032
practicalswift:
cr ACK 8f7704d032
hebasto:
ACK 8f7704d032, tested on macOS Big Sur 11.4 (20F71) and on Linux Mint 20.1 (x86_64) with depends.
Tree-SHA512: 5f1351d0ac2e655fccb22a5454f415906404fdaa336fd89b54ef49ca50a442c44ab92d063cba3f161cb8ea0679c92ae3cd6cfbbcb19728cac21116247a017df5
2f5bdcbc31 gui: misc external signer fixes and translation hints (Sjors Provoost)
d672404466 refactor: make ExternalSigner NetworkArg() and m_chain private (Sjors Provoost)
4455145e26 refactor: reduce #ifdef ENABLE_EXTERNAL_SIGNER usage (Sjors Provoost)
5be90c907e build: enable external signer by default (Sjors Provoost)
7d9453041b refactor: clean up external_signer.h includes (Sjors Provoost)
fc0eca31b3 fuzz: fix fuzz binary linking order (Sjors Provoost)
Pull request description:
This follows the introduction of GUI support in https://github.com/bitcoin-core/gui/pull/4
I don't think we should expect GUI users to self compile. This also enables external signer support by default for RPC users.
In addition this PR reduces the number of `#ifdef ENABLE_EXTERNAL_SIGNER`, which also fixes#21919. When compiled with `--disable-external-signer` such wallets can't be created in RPC or GUI, but they can be loaded. Attempting any action that calls HWI will trigger an error.
Side-note: this PR may or may not (currently) break CI for the GUI repository, as explained here: https://github.com/bitcoin-core/gui/pull/4#issuecomment-769859001
ACKs for top commit:
achow101:
ACK 2f5bdcbc31
hebasto:
re-ACK 2f5bdcbc31
Tree-SHA512: 1b71c5a8bea2be077ee9fa33a01130c957a0cf90951d4b7b04d3d0ef826bb77e474c3963abddfef2e2c1ea99d9c72cd2302d1eb9b5fcb7ba0bd2a625f006aa05
Just checking for the `sys/sdt.h` header isn't enough, as systems like
macOS have the header, but it doesn't actually have the dtrace probes,
which leads to compile failures.
5c7ee1b2da libsecp256k1 no longer has --with-bignum= configure option (Pieter Wuille)
bdca9bcb6c Squashed 'src/secp256k1/' changes from 3967d96bf1..efad3506a8 (Pieter Wuille)
cabb566123 Disable certain false positive warnings for libsecp256k1 msvc build (Pieter Wuille)
Pull request description:
This updates our src/secp256k1 subtree to the latest upstream master. The changes include:
* The introduction of safegcd-based modular inverses, reducing ECDSA signing time by 25%-30% and ECDSA verification time by 15%-17%.
* [Original paper](https://gcd.cr.yp.to/papers.html) by Daniel J. Bernstein and Bo-Yin Yang
* [Implementation](https://github.com/bitcoin-core/secp256k1/pull/767) by Peter Dettman; [final](https://github.com/bitcoin-core/secp256k1/pull/831) version
* [Explanation](https://github.com/bitcoin-core/secp256k1/blob/master/doc/safegcd_implementation.md) of the algorithm using Python snippets
* [Analysis](https://github.com/sipa/safegcd-bounds) of the maximum number of iterations the algorithm needs
* [Formal proof in Coq](https://medium.com/blockstream/a-formal-proof-of-safegcd-bounds-695e1735a348) by Russell O'Connor, for a high-level equivalent algorithm
* Removal of libgmp as an (optional) dependency (which wasn't used in the Bitcoin Core build)
* CI changes (Travis -> Cirrus)
* Build system improvements
ACKs for top commit:
laanwj:
Tested ACK 5c7ee1b2da
Tree-SHA512: ad8ac3746264d279556a4aa7efdde3733e114fdba8856dd53218588521f04d83950366f5c1ea8fd56329b4c7fe08eedf8e206f8f26dbe3f0f81852e138655431
e9f948c727 build: Convert warnings into errors when testing for -fstack-clash-protection (Hennadii Stepanov)
Pull request description:
Apple clang version 12.0.5 (clang-1205.0.22.9) that is a part of Xcode 12.5, and is based on LLVM clang 11.1.0, fires spammy warnings:
```
clang: warning: argument unused during compilation: '-fstack-clash-protection' [-Wunused-command-line-argument]
```
From the https://github.com/apple/llvm-project:
```
$ git log --oneline | grep 'stack-clash-protection'
00065d5cbd02 Revert "-fstack-clash-protection: Return an actual error when used on unsupported OS"
4d59c8fdb955 -fstack-clash-protection: Return an actual error when used on unsupported OS
df3bfaa39071 [Driver] Change -fnostack-clash-protection to -fno-stack-clash-protection
68e07da3e5d5 [clang][PowerPC] Enable -fstack-clash-protection option for ppc64
515bfc66eace [SystemZ] Implement -fstack-clash-protection
e67cbac81211 Support -fstack-clash-protection for x86
454621160066 Revert "Support -fstack-clash-protection for x86"
0fd51a4554f5 Support -fstack-clash-protection for x86
658495e6ecd4 Revert "Support -fstack-clash-protection for x86"
e229017732bc Support -fstack-clash-protection for x86
b03c3d8c6209 Revert "Support -fstack-clash-protection for x86"
4a1a0690ad68 Support -fstack-clash-protection for x86
f6d98429fcdb Revert "Support -fstack-clash-protection for x86"
39f50da2a357 Support -fstack-clash-protection for x86
```
I suppose, that Apple clang-1205.0.22.9 ends with on of the "Revert..." commits.
This PR prevents using of the `-fstack-clash-protection` flag if it causes warnings.
---
System: macOS Big Sur 11.3 (20E232).
ACKs for top commit:
jarolrod:
re-ACK e9f948c727
Sjors:
tACK e9f948c727 on macOS 11.3.1
Tree-SHA512: 30186da67f9b0f34418014860c766c2e7f622405520f1cbbc1095d4aa4038b0a86014d76076f318a4b1b09170a96d8167c21d7f53a760e26017f486e1a7d39d4
When building with Clang, if `-fstack-clash-protection` is used with an
unsupported target, it may result in hundreds of
`-Wunused-command-line-argument` warnings at compile time. This is
currently the case when building for at least Darwin using Apple or LLVM
Clang.
Unsupported targets may also include *BSD, however that is changing; see
further discussion in https://reviews.llvm.org/D92245 and
https://reviews.freebsd.org/D27366.
Note that this option is already skipped for Windows.
a5491882a0 build: fix configuring when building depends with NO_BDB=1 (fanquake)
Pull request description:
Currently, if you build depends using `NO_BDB=1` (only sqlite wallets), `./configure` will fail as it still tries to find bdb. i.e:
```bash
make -C depends/ NO_QT=1 NO_BDB=1 NO_UPNP=1 NO_ZMQ=1 NO_NATPMP=1 -j8
...
copying packages: native_b2 boost libevent sqlite
./autogen.sh
./configure --prefix=/home/ubuntu/bitcoin/depends/x86_64-pc-linux-gnu
...
checking for Berkeley DB C++ headers... default
configure: error: Found Berkeley DB other than 4.8, required for portable BDB wallets (--with-incompatible-bdb to ignore or --without-bdb to disable BDB wallet support)
```
This PR fixes the build such that you can build depends, opting out of bdb, without opting out of wallets entirely, and still configure successfully. I think I've tested across most potential configurations. i.e:
```bash
./configure (bdb and sqlite on system)
bdb & sqlite are both are available
./configure --without-bdb (bdb and sqlite on system)
only sqlite
./configure --without-sqlite (bdb and sqlite on system)
only bdb
./configure --disable-wallet (bdb and sqlite on system)
neither bdb or sqlite
depends NO_WALLET=1
./configure --prefix=/bitcoin/depends/x86_64-apple-darwin19.6.0
neither bdb or sqlite
depends NO_BDB=1
./configure --prefix=/bitcoin/depends/x86_64-apple-darwin19.6.0
only sqlite
depends NO_SQLITE=1
./configure --prefix=/bitcoin/depends/x86_64-apple-darwin19.6.0
only bdb
depends
./configure --prefix=/bitcoin/depends/x86_64-apple-darwin19.6.0
bdb and sqlite
```
ACKs for top commit:
laanwj:
Code review ACK a5491882a0
jarolrod:
ACK a5491882a0
Tree-SHA512: baf7d2543a401db0d846095415ff449c04ecfb4a74c734dc51e79453702f9051210daeef686970f11fcffd32cdfadbc58acd54f0706aceecfb3edb0ff17310d7
7abac98d3e configure: Support -f{debug,macro}-prefix-map (Anthony Towns)
Pull request description:
When bitcoin is checked out in two directories (eg via git worktree) object files between the two will differ due to the full path being included in the debug section. `-fdebug-prefix-map` is used to replace this with "." to avoid this unnecessary difference and allow ccache to share objects between worktrees (provided the source and compile options are the same).
Also provide `-fmacro-prefix-map` if supported so that the working dir is not encoded in `__FILE__` macros.
ACKs for top commit:
practicalswift:
cr ACK 7abac98d3e3c1bc8ad66cb5c05184b9c5cc674d5: patch looks correct
fanquake:
ACK 7abac98d3e
Tree-SHA512: b6a37c1728ec3b2e552f244da0e66db113c1e7662c7ac502e12ff466f3dbfbfefae12695ca135137c50dbb1c4c5d84059116c0cd09b391a17466dc77b8726679
Currently, if you build depends using `NO_BDB=1` (only sqlite wallets),
./configure will fail as it still tries to find bdb. i.e:
```bash
checking for Berkeley DB C++ headers... default
configure: error: Found Berkeley DB other than 4.8, required for portable BDB wallets (--with-incompatible-bdb to ignore or --without-bdb to disable BDB wallet support)
```
This PR fixes the build such that you can build depends, opting out of
bdb without opting out of wallets entirely, and still configure
successfully.
a4e970adb6 build: enable -Wdocumentation if suppressing external warnings (fanquake)
3b0078f958 doc: fixup -Wdocumentation issues (fanquake)
c6edcf1c71 build: suppress libevent warnings if supressing external warnings (fanquake)
Pull request description:
Enable `-Wdocumentation` by taking advantage of our `--enable-suppress-external-warnings` flag. Most of the CIs are using this flag now, so any regressions should be caught.
This also required modifying libevents flags when suppressing warnings, as depending on the version being built against, that could generate a large number of warnings. i.e:
```bash
In file included from httpserver.cpp:34:
In file included from ./support/events.h:12:
/usr/local/Cellar/libevent/2.1.12/include/event2/http.h:464:11: warning: parameter 'req' not found in the function declaration [-Wdocumentation]
@param req a request object
^~~
/usr/local/Cellar/libevent/2.1.12/include/event2/http.h:465:11: warning: parameter 'databuf' not found in the function declaration [-Wdocumentation]
@param databuf the data chunk to send as part of the reply.
^~~~~~~
/usr/local/Cellar/libevent/2.1.12/include/event2/http.h:467:11: warning: parameter 'call' not found in the function declaration [-Wdocumentation]
@param call back's argument.
^~~~
/usr/local/Cellar/libevent/2.1.12/include/event2/http.h:939:4: warning: declaration is marked with '@deprecated' command but does not have a deprecation attribute [-Wdocumentation-deprecated-sync]
@deprecated This function is deprecated; you probably want to use
~^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
/usr/local/Cellar/libevent/2.1.12/include/event2/http.h:946:1: note: add a deprecation attribute to the declaration to silence this warning
char *evhttp_decode_uri(const char *uri);
^
__AVAILABILITY_INTERNAL_DEPRECATED
/usr/local/Cellar/libevent/2.1.12/include/event2/http.h:979:5: warning: declaration is marked with '@deprecated' command but does not have a deprecation attribute [-Wdocumentation-deprecated-sync]
@deprecated This function is deprecated as of Libevent 2.0.9. Use
~^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
/usr/local/Cellar/libevent/2.1.12/include/event2/http.h:987:1: note: add a deprecation attribute to the declaration to silence this warning
int evhttp_parse_query(const char *uri, struct evkeyvalq *headers);
^
__AVAILABILITY_INTERNAL_DEPRECATED
/usr/local/Cellar/libevent/2.1.12/include/event2/http.h:1002:11: warning: parameter 'query_parse' not found in the function declaration [-Wdocumentation]
@param query_parse the query portion of the URI
^~~~~~~~~~~
/usr/local/Cellar/libevent/2.1.12/include/event2/http.h:1002:11: note: did you mean 'uri'?
@param query_parse the query portion of the URI
^~~~~~~~~~~
uri
69 warnings generated.
```
Note that a lot of these have already been fixed upstream.
ACKs for top commit:
laanwj:
Concept and code review ACK a4e970adb6
practicalswift:
cr ACK a4e970adb6de8425025ae3f62fb89d9e27a8ab1f: automatic compiler feedback comes sooner and is more reliable than manual reviewer feedback
jonatack:
Light ACK a4e970adb6 skimmed the changes, clang 11 build is clean with the change, verified -Wdocumentation build warnings with this change when a doc fix was reverted
Tree-SHA512: 57a1e30cffcc8bcceee72d85f58ebe29eae525861c70acb237541bd480c51ede89875c033042c0af376fdbb49fb7f588ef9282a47c6e78f9d4501c41f1b21eb6
The register keyword was deprecated in C++11, and removed in C++17. Now
that we require C++17, we shouldn't have to supress warnings for a
non-existant feature.
0eabb2abed build: Remove unused header from the build system (Hennadii Stepanov)
Pull request description:
The only `#include <miniupnpc/miniwget.h>` was removed in #16659.
ACKs for top commit:
practicalswift:
cr ACK 0eabb2abed
fanquake:
ACK 0eabb2abed
Tree-SHA512: 630da03875c851e80286561eae0f966c89624cbb17b90f70e2bec9a69146e79d088fc176e07a4906915770ac1cdb11341a7a431ea7cf6a59d2816e927486f335
246774e264 depends: fix Qt precompiled headers bug (Igor Cota)
8e7ad4146d depends: disable Qt Vulkan support on Android (Igor Cota)
ba46adaa1a CI: add Android APK build to cirrus (Igor Cota)
7563720e30 CI: add Android APK build script (Igor Cota)
ebfb10cb75 Qt: add Android packaging support (Igor Cota)
Pull request description:
This PR is the third and final piece of the basic Android support puzzle - it depends on https://github.com/bitcoin/bitcoin/pull/16110 and is related to https://github.com/bitcoin/bitcoin/pull/16883. It introduces an `android` directory under `qt` and a simple way to build an Android package of `bitcoin-qt`:
1. Build depends for Android as described in the [README](https://github.com/bitcoin/bitcoin/blob/master/depends/README.md)
2. Configure with one of the resulting prefixes
3. Run `make && make apk` in `src/qt`
The resulting APK files will be in `android/build/outputs/apk`. You can install them manually or with [adb](https://developer.android.com/studio/command-line/adb). One can also open the `android` directory in Android Studio for that integrated development and debugging experience. `BitcoinQtActivity` is your starting point.
Under the hood makefile `apk` target:
1. Renames the `bitcoin-qt` binary to `libbitcoin-qt.so` and copies it over to a folder under `android/libs` depending on which prefix and corresponding [ABI](https://developer.android.com/ndk/guides/abis.html#sa) `bitcoin-qt` was built for
2. Takes `libc++_shared.so` from the Android NDK and puts in the same place. It [must be included](https://developer.android.com/ndk/guides/cpp-support) in the APK
3. Extracts Qt for Android Java support files from the `qtbase` archive in `depends/sources` to `android/src`
There is also just a tiny bit of `ifdef`'d code to make the Qt Widgets menus usable. It's not pretty but it works and is a stepping stone towards https://github.com/bitcoin/bitcoin/pull/16883.
ACKs for top commit:
MarcoFalke:
cr ACK 246774e264
laanwj:
Code review ACK 246774e264
Tree-SHA512: ba30a746576a167545223c35a51ae60bb0838818779fc152c210f5af1413961b2a6ab6af520ff92cbc8dcd5dcb663e81ca960f021218430c1f76397ed4cead6c
Introduce an android directory under qt and allow one to package bitcoin-qt for Android by running make apk.
Add bitcoin-qt Android build instructions.
This has never worked with any of the mingw-w64 compilers we use, and
the -O0 is causing issues for builders applying spectre mitigations.
Recent discussion on https://gcc.gnu.org/bugzilla/show_bug.cgi?id=90458
also indicates that this should just not be used on Windows.
e017a913d0 bitcoind: Add -daemonwait option to wait for initialization (Wladimir J. van der Laan)
c3e6fdee6d shutdown: Use RAII TokenPipe in shutdown (Wladimir J. van der Laan)
612f746a8f util: Add RAII TokenPipe (Wladimir J. van der Laan)
Pull request description:
This adds a `-daemonwait` flag that does the same as `-daemon` except that it, from a user perspective, backgrounds the process only after initialization is complete. This is similar to the behaviour of some other software such as c-lightning.
This can be useful when the process launching bitcoind wants to guarantee that either the RPC server is running, or that initialization failed, before continuing. The exit code indicates the initialization result.
The use of the libc function `daemon()` is replaced by a custom implementation which is inspired by the [glibc implementation](https://github.com/lattera/glibc/blob/master/misc/daemon.c#L44), but which also creates a pipe from the child to the parent process for communication.
An additional advantage of having our own `daemon()` implementation is that no MACOS-specific pragmas are needed anymore to silence a deprecation warning.
TODO:
- [x] Factor out `token_read` and `token_write` to an utility, and use them in `shutdown.cpp` as well—this is exactly the same kind of communication mechanism.
- [x] RAII-ify pipe endpoints.
- [x] Improve granularity of the `configure.ac` checks. This currently still checks for the function `daemon()` which makes no sense as it's not used. It should check for individual functions such as
`fork()` and `setsid()` etc—the former being required, the second optional.
- [-] ~~Signal propagation during initialization: if say, pressing Ctrl-C during `-daemonwait` it would be good to pass this SIGINT on to the child process instead of detaching the parent process and letting the child run free.~~ This is not necessary, see https://github.com/bitcoin/bitcoin/pull/21007#issuecomment-769007341.
Future:
- Consider if it makes sense to use this in the RPC tests (there would be no more need for "is RPC ready" polling loops). I think this is out of scope for this PR.
ACKs for top commit:
jonatack:
Tested ACK e017a913d0 checked change since previous review is move-only
Tree-SHA512: 53369b8ca2247e4cf3af8cb2cfd5b3399e8e0e3296423d64be987004758162a7ddc1287b01a92d7692328edcb2da4cf05d279b1b4ef61a665b71440ab6a6dbe2
This adds a `-daemonwait` flag that does the same as `-daemon` except
it, from a user perspective, backgrounds the process only after
initialization is complete.
This can be useful when the process launching bitcoind wants to
guarantee that either the RPC server is running, or that initialization
failed, before continuing. The exit code indicates the initialization
result.
This replaces the use of the libc function `daemon()` by a custom
implementation which is inspired by the glibc implementation, but also
creates a pipe from the child to the parent process for communication.
An additional advantage of having our own `daemon()` implementation is
that no MACOS-specific pragmas are needed anymore to silence a
deprecation warning.
faa06ecc9c build: Bump minimum Qt version to 5.9.5 (Hennadii Stepanov)
Pull request description:
Close#20104.
ACKs for top commit:
laanwj:
Code review ACK faa06ecc9c
jarolrod:
ACK faa06ecc9c
fanquake:
ACK faa06ecc9c - this should be ok to do now.
Tree-SHA512: 7295472b5fd37ffb30f044e88c39d375a5a5187d3f2d44d4e73d0eb0c7fd923cf9949c2ddab6cddd8c5da7e375fff38112b6ea9779da4fecce6f024d05ba9c08
fae216a73d scripted-diff: Rename MakeFuzzingContext to MakeNoLogFileContext (MarcoFalke)
fa4fbec03e scripted-diff: Rename PROVIDE_MAIN_FUNCTION -> PROVIDE_FUZZ_MAIN_FUNCTION (MarcoFalke)
Pull request description:
Split out two renames from #21003:
* `PROVIDE_FUZZ_MAIN_FUNCTION`. *Reason*: This in only used by fuzzing, so the name should indicate that.
* `MakeNoLogFileContext`. *Reason*: Better reflects what the helper does. Also, prepares it to be used in non-fuzz tests in the future.
ACKs for top commit:
practicalswift:
cr ACK fae216a73d71441432cf07212313d18771ce7deb: scripted-diff looks correct
Tree-SHA512: e5d347746f5da72b0c86fd4f07ac2e4b3016e88e8c97a830c73bd79d0af6d0245fe7712487fc20344d6cc25958941716c1678124a123930407e3a437265b71df
9bac71350d build: make HAVE_O_CLOEXEC available outside LevelDB (bugfix) (Sebastian Falbesoner)
584fd91d2d init: only use pipe2 if availabile, check in configure (Sebastian Falbesoner)
Pull request description:
The result of the O_CLOEXEC availability check is currently only set in the Makefile and passed to LevelDB (see `LEVELDB_CPPFLAGS_INT` in `src/Makefile.leveldb.include`), but not defined to be used in our codebase. This means that code within the preprocessor conditional `#if HAVE_O_CLOEXEC` was actually never compiled. On the master branch this is currently used for pipe creation in `src/shutdown.cpp`, PR #21007 moves this part to a new module (I found the issue while testing that PR).
The fix is similar to the one in #19803, which solved the same problem for HAVE_FDATASYNC.
In the course of working on the PR it turned out that pipe2 is not available an all platforms, hence a configure check and a corresponding define HAVE_PIPE2 is introduced and used.
The PR can be tested by anyone with a system that has pipe2 and O_CLOEXEC available by putting gibberish into the HAVE_O_CLOEXEC block: on master, everything should compile fine, on PR, the compiler should abort with an error. At least that's my naive way of testing preprocessor logic, happy to hear more sophisticated ways :-)
ACKs for top commit:
laanwj:
Code review ACK 9bac71350d
Tree-SHA512: aec89faf6ba52b6f014c610ebef7b725d9e967207d58b42a4a71afc9f1268fcb673ecc85b33a2a3debba8105a304dd7edaba4208c5373fcef2ab83e48a170051
This option replaces --with-boost-process
This prepares external signer support to be disabled by default.
It adds a configure option to enable this feature and to check
if Boost::Process is present.
This also exposes ENABLE_EXTERNAL_SIGNER to the test suite via test/config.ini
c5da2749e2 build: actually stop configure if Boost isn't available (fanquake)
cad8b527ea build: explicitly install libboost-dev package (fanquake)
Pull request description:
If Boost is not found via AX_BOOST_BASE, we don't actually stop
configuring, only a warning is emitted:
```bash
checking for boostlib >= 1.58.0 (105800)... configure: We could not detect the boost libraries (version MINIMUM_REQUIRED_BOOST or higher). If you have a staged boost library (still not installed) please specify $BOOST_ROOT in your environment and do not give a PATH to --with-boost option. If you are sure you have boost installed, then check your version number looking in <boost/version.hpp>. See http://randspringer.de/boost for more documentation.
```
Instead we usually fail when one of the other AX_BOOST_* macros fails to find a library. These macros are slowly being
removed, and in any case, it makes more sense to fail earlier if Boost is missing.
If Boost is unavailable, the failure now looks like:
```bash
checking for boostlib >= 1.58.0 (105800)... configure: We could not detect the boost libraries (version 1.58.0 or higher). If you have a staged boost library (still not installed) please specify $BOOST_ROOT in your environment and do not give a PATH to --with-boost option. If you are sure you have boost installed, then check your version number looking in <boost/version.hpp>. See http://randspringer.de/boost for more documentation.
configure: error: Boost is not available!
```
Note that we now just pass the version into AX_BOOST_BASE, which fixes it's display in the output (rather than showing `MINIMUM_REQUIRED_BOOST`).
This PR also has a commit that adds `libboost-dev` to our install instructions and CI. This package is currently installed as a side-effect of installing our other libboost-*-dev packages. However as those continue to disappear, it makes sense to install boost-dev explicitly.
ACKs for top commit:
laanwj:
Code review ACK c5da2749e2
MarcoFalke:
Concept ACK c5da2749e2
Tree-SHA512: f866062f9d7d3a2316b6c887f17c664b9cfff41fdc0cb99ca79d641240fb01a5ae0d34140e515bc465219e1b43d5ca84f7c55f48b9c5b45a80ff2795dafd072b
If Boost is not found via AX_BOOST_BASE, we don't actually stop
configuring, only a warning is emitted:
```bash
checking for boostlib >= 1.58.0 (105800)... configure: We could not detect the boost libraries (version MINIMUM_REQUIRED_BOOST or higher). If you have a staged boost library (still not installed) please specify $BOOST_ROOT in your environment and do not give a PATH to --with-boost option. If you are sure you have boost installed, then check your version number looking in <boost/version.hpp>. See http://randspringer.de/boost for more documentation.
```
Instead we would usually fail when one of the other
AX_BOOST_* macros fails to find a library. These macros are slowly being
removed, and in any case, it makes more sense to fail earlier if Boost
is missing.
If Boost is unavailable, the failure now looks like:
```bash
checking for boostlib >= 1.58.0 (105800)... configure: We could not detect the boost libraries (version 1.58.0 or higher). If you have a staged boost library (still not installed) please specify $BOOST_ROOT in your environment and do not give a PATH to --with-boost option. If you are sure you have boost installed, then check your version number looking in <boost/version.hpp>. See http://randspringer.de/boost for more documentation.
configure: error: Boost is not available!
```
Note that we now just pass the version into AX_BOOST_BASE, which fixes
it's display in the output (rather than MINIMUM_REQUIRED_BOOST).
Performing a series of link checks for a Boost component that is
header-only doesn't make much sense, and currently means we just have
another confusing Boost macro in our tree. I'm not sure why this was
originally done this way; maybe Sjors or luke-jr can elaborate
(#15382 (929cda5470))?
The macro also has the side-effect of producing confusing error
messages. i.e in #20744, the CI is currently failing with:
```bash
checking for boostlib >= 1.58.0 (105800) lib path in "/tmp/cirrus-ci-build/depends/x86_64-pc-linux-gnu/lib"... yes
checking for boostlib >= 1.58.0 (105800)... yes
checking whether the Boost::Process library is available... yes
configure: error: Could not find a version of the Boost::Process library!
```
This isn't useful, given there is no such thing as a `Boost::Process`
library.
This PR just removes the macro entirely, but maintains a `--with-boost-process`
(defaulting to off), flag to configure. Hopefully this will also be
removed, in favour of `--enable-disable-external-signer` if/when #16546
is merged.
060a2a64d4 ci: remove boost thread installation (fanquake)
06e1d7d81d build: don't build or use Boost Thread (fanquake)
7097add83c refactor: replace Boost shared_mutex with std shared_mutex in sigcache (fanquake)
8e55981ef8 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 060a2a64d4
vasild:
ACK 060a2a64d4
Tree-SHA512: 572d14d8c9de20bc434511f20d3f431836393ff915b2fe9de5a47a02dca76805ad5c3fc4cceecb4cd43f3ba939a0508178c4e60e62abdbaaa6b3e8db20b75b03
We are already testing for this, and our test works correctly with a Darwin
target, where the macro does not. Darwin targets do not support "protected"
visibility.
32cbb06676 build: build fuzz tests by default. (Dan Benjamin)
Pull request description:
This fixes issue #19388. The changes are as follows:
- Add a new flag to configure, --enable-fuzz-binary, which allows building test/fuzz/fuzz regardless of whether we are building to do actual fuzzing
- Set -DPROVIDE_MAIN_FUNCTION whenever --enable-fuzz is no
- Add the following libraries to FUZZ_SUITE_LD_COMMON:
- LIBBITCOIN_WALLET
- SQLLITE_LIBS
- BDB_LIBS
- if necessary, some or all of:
- NATPMP_LIBS
- MINIUPNPC_LIBS
- LIBBITCOIN_ZMQ / ZMQ_LIBS
Fixes #19388
ACKs for top commit:
MarcoFalke:
review ACK 32cbb06676📭
Tree-SHA512: c91d713ffe54a3d055daaec02c4317d7e13eed6688821ddc10d894224950b18e276fbdd4acc758c7103b50f34a132b1882b68bc8b60409f97438e0759ced77e1
fanquake
