From what I can see the only platform this drops support for is CentOS
7. CentOS 7 reached the end of it's "full update" support at the end of
2020. It does receive maintenance updates until 2024, however I don't
think supporting glibc 2.17 until 2024 is realistic. Note that anyone
wanting to self-compile and target a glibc 2.17 runtime could build with
--disable-threadlocal.
glibc 2.18 was released in August 2013.
https://sourceware.org/legacy-ml/libc-alpha/2013-08/msg00160.html
29173d6c6c ubsan: add minisketch exceptions (Cory Fields)
54b5e1aeab Add thin Minisketch wrapper to pick best implementation (Pieter Wuille)
ee9dc71c1b Add basic minisketch tests (Pieter Wuille)
0659f12b13 Add minisketch dependency (Gleb Naumenko)
0eb7928ab8 Add MSVC build configuration for libminisketch (Pieter Wuille)
8bc166d5b1 build: add minisketch build file and include it (Cory Fields)
b2904ceb85 build: add configure checks for minisketch (Cory Fields)
b6487dc4ef Squashed 'src/minisketch/' content from commit 89629eb2c7 (fanquake)
Pull request description:
This takes over #21859, which has [recently switched](https://github.com/bitcoin/bitcoin/pull/21859#issuecomment-921899200) to my integration branch. A few more build issues came up (and have been fixed) since, and after discussing with sipa it was decided I would open a PR to shepherd any final changes through.
> This adds a `src/minisketch` subtree, taken from the master branch of https://github.com/sipa/minisketch, to prepare for Erlay implementation (see #21515). It gets configured for just supporting 32-bit fields (the only ones we're interested in in the context of Erlay), and some code on top is added:
> * A very basic unit test (just to make sure compilation & running works; actual correctness checking is done through minisketch's own tests).
> * A wrapper in `minisketchwrapper.{cpp,h}` that runs a benchmark to determine which field implementation to use.
Only changes since my last update to the branch in the previous PR have been rebasing on master and fixing an issue with a header in an introduced file.
ACKs for top commit:
naumenkogs:
ACK 29173d6c6c
Tree-SHA512: 1217d3228db1dd0de12c2919314e1c3626c18a416cf6291fec99d37e34fb6eec8e28d9e9fb935f8590273b8836cbadac313a15f05b4fd9f9d3024c8ce2c80d02
AC_DEFINE'd values won't be passed down to minisketch because it does not
use bitcoin-config.h. Thus we need a way to know if we should manually add
defines for minisketch files.
17ae2601c7 build: remove build stubs for external leveldb (Cory Fields)
Pull request description:
Presumably these stubs indicate to packagers that external leveldb is meant to be supported in some way. It is not. Remove the stubs to avoid sending any mixed messages.
For context, this was reported on IRC:
> \<Talkless> bitcoind fails to start with undefined symbol: _ZTIN7leveldb6LoggerE in Debian Sid after leveldb upgraded from 1.22 to 1.23: https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=996486
ACKs for top commit:
fanquake:
ACK 17ae2601c7
hebasto:
ACK 17ae2601c7. I have reviewed the code and it looks OK, I agree it can be merged.
Tree-SHA512: 2f1ac2cb30dac64791933a245a2b66ce237bde3955e6f4a6b7ec181248f77a9b1b10597d865d3e2c2b6def696af70de40e905ec274e4ae7cccd1daf461473957
These tests are failing when run against OpenSSL 3, and have been
removed upstream, https://github.com/bitcoin-core/secp256k1/pull/983, so
disabled them for now to avoid `make check` failures.
Note that this will also remove warning output from our build, due to
the use of deprecated OpenSSL API functions. See #23048.
0f95247246 Integrate univalue into our buildsystem (Cory Fields)
9b49ed656f Squashed 'src/univalue/' changes from 98fadc0909..a44caf65fe (fanquake)
Pull request description:
This PR more tightly integrates building Univalue into our build system. This follows the same approach we use for [LevelDB](https://github.com/bitcoin-core/leveldb/), ([`Makefile.leveldb.include`](https://github.com/bitcoin/bitcoin/blob/master/src/Makefile.leveldb.include)), and [CRC32C](https://github.com/bitcoin-core/crc32c) ([`Makefile.crc32c.include`](https://github.com/bitcoin/bitcoin/blob/master/src/Makefile.crc32c.include)), and will be the same approach we use for [minisketch](https://github.com/sipa/minisketch); see #23114.
This approach yields a number of benefits, including:
* Faster configuration due to one less subconfigure being run during `./configure` i.e 22s with this PR vs 26s
* Faster autoconf i.e 13s with this PR vs 17s
* Improved caching
* No more issues with compiler flags i.e https://github.com/bitcoin/bitcoin/pull/12467
* More direct control means we can build exactly the objects we want
There might be one argument against making this change, which is that builders should have the option to use "proper shared/system libraries". However, I think that falls down for a few reasons. The first being that we already don't support building with a number of system libraries (secp256k1, leveldb, crc32c); some for good reason. Univalue is really the odd one out at the moment.
Note that the only fork of Core I'm aware of, that actively patches in support for using system libs, also explicitly marks them as ["DANGEROUS"](a886811721/configure.ac (L1430)) and ["NOT SUPPORTED"](a886811721/configure.ac (L1312)). So it would seem they exist more to satisfy a distro requirement, as opposed to something that anyone should, or would actually use in practice.
PRs like #22412 highlight the "issue" with us operating with our own Univalue fork, where we actively fix bugs, and make improvements, when upstream (https://github.com/jgarzik/univalue) may not be taking those improvements, and by all accounts, is not currently actively maintained. Bitcoin Core should not be hamstrung into not being able to fix bugs in a library, and/or have to litter our source with "workarounds", i.e #22412, for bugs we've already fixed, based on the fact that an upstream project is not actively being maintained. Allowing builders to use system libs is really only exacerbating this problem, with little benefit to our project. Bitcoin Core is not quite like your average piece of distro packaged software.
There is the potential for us to give the same treatment to libsecp256k1, however it seems doing that is currently less straightforward.
ACKs for top commit:
dongcarl:
ACK 0f95247246 less my comment above, always nice to have an include-able `sources.mk` which makes integration easier.
theuni:
ACK 0f95247246. Thanks fanquake for keeping this going.
Tree-SHA512: a7f2e41ee7cba06ae72388638e86b264eca1b9a8b81c15d1d7b45df960c88c3b91578b4ade020f8cc61d75cf8d16914575f9a78fa4cef9c12be63504ed804b99
Presumably these stubs indicate to packagers that external leveldb is meant to
be supported in some way. It is not. Remove the stubs to avoid sending any
mixed messages.
aa69fd6caf build: Drop -Wno-unused-local-typedef (Hennadii Stepanov)
672e8c5d07 build: remove -Wunused-variable (fanquake)
5239af0574 build: remove -Wswitch (fanquake)
0375906e0a build: use loop-analysis over range-loop-analysis (fanquake)
12712fa2c4 build: remove -Wsign-compare (fanquake)
Pull request description:
This remove the addition of flags that are already part of other options, such as `-Wall` or `-Wextra`; see each commit message for details. All of the flags being removed here already exist as part of `-Wall` as of GCC 8, or, for Clang, all exist in `-Wmost` (included in `-Wall)`, or as part of `-Wextra` as of Clang 7. Both of which are our minimum required compilers.
Also cherry-picks one change from #21458.
To give an example of how GCCs `-Wall` has changed over the last few releases:
### 11.x to trunk (12.x)
Added:
```bash
-Wzero-length-bounds
-Wmismatched-dealloc
-Wmismatched-new-delete (only for C/C++)
```
### 10.x to 11.x
Added:
```bash
-Warray-parameter=2 (C and Objective-C only)
-Wrange-loop-construct (only for C++)
-Wsizeof-array-div
-Wvla-parameter (C and Objective-C only)
```
Removed:
```bash
-Wenum-conversion in C/ObjC;
```
### 9.x to 10.x
Added:
```bash
-Wenum-conversion in C/ObjC;
-Wformat-overflow
-Wformat-truncation
-Wzero-length-bounds
```
### 8.x to 9.x
Added:
```bash
-Wpessimizing-move
```
Removed:
```bash
-Wstringop-truncation
```
### 7.x to 8.x
Added:
```bash
-Wcatch-value (C++ and Objective-C++ only)
-Wmissing-attributes
-Wmultistatement-macros
-Wrestrict
-Wsizeof-pointer-div
-Wstringop-truncation
```
[Clang Warning Options](https://clang.llvm.org/docs/DiagnosticsReference.html)
[GCC Warning Options](https://gcc.gnu.org/onlinedocs/gcc/Warning-Options.html)
ACKs for top commit:
meshcollider:
utACK aa69fd6caf
Tree-SHA512: 34dde6bd773c864202c151eaa35f902d03fb531c27fe5e1ef659225da03acade2efe5df56df3efb4df5bbded3d395348ce03c25b837fce83be53af3352f0f2bc
ce69e18947 scripts: remove pixie.py (fanquake)
00b85d0b13 scripts: only parse the binary once in security-check.py (fanquake)
cad40a5b16 scripts: use LIEF for ELF checks in security-check.py (fanquake)
8242ae230e scripts: only parse the binary once in symbol-check.py (fanquake)
309eac9019 scripts: use LIEF for ELF checks in symbol-check.py (fanquake)
610a8a8e39 test-*-check: Pass in *FLAGS and compile with them (Carl Dong)
Pull request description:
This finishes the transition to using LIEF for the ELF symbol and security checks.
Note that there's currently a work around used for identifying RISCV binaries (just checking the interpreter). I've sent a PR upstream, https://github.com/lief-project/LIEF/pull/562, and we should be able to drop that when using LIEF 0.12.0 and onwards.
ACKs for top commit:
dongcarl:
Code Review ACK ce69e18947
laanwj:
Code review ACK ce69e18947
Tree-SHA512: 911ba693cd9777ad1fc1f66dff6c4d3630a907351215380cbde5b14a4bbf5cf7eebf52eafa7e86b27deabd2d93d1b403f34aabd356b5ceaab3cc6e9941a01dd4
38fd709fa5 build: make --enable-werror just -Werror (fanquake)
Pull request description:
No longer special case a set of warnings, to make up our own -Werror,
just use -Werror outright. This shouldn't really have any effect on
existing builders, who were already using `--enable-werror`, and is more
inline with what they would expect `--enable-werror` to be, which is
erroring on any/all warnings.
We keep `-Wno-error=return-type` because we know that is broken when using
mingw-w64. It should only be applied when cross-compiling for Windows.
Similar to the change in #20544, but with (hopefully) less work-arounds,
and other bundled changes. A step towards some configure "cleanups".
ACKs for top commit:
hebasto:
ACK 38fd709fa5 (also see https://github.com/bitcoin/bitcoin/pull/23149#issuecomment-940420776), tested:
MarcoFalke:
Concept ACK 38fd709fa5
Tree-SHA512: 37f1857d9408442cab63e40f9280427b73e09cdf03146b19c1339d7e44abd78e93df7f270ca1da0e83b79343cd3ea915f7b9e4e347488b5bc5ceaaa7540e5926
This addresses issues like the one in #12467, where some of our compiler flags
end up being dropped during the subconfigure of Univalue. Specifically, we're
still using the compiler-default c++ version rather than forcing c++17.
We can drop the need subconfigure completely in favor of a tighter build
integration, where the sources are listed separately from the build recipes,
so that they may be included directly by upstream projects. This is
similar to the way leveldb build integration works in Core.
Core benefits of this approach include:
- Better caching (for ex. ccache and autoconf)
- No need for a slow subconfigure
- Faster autoconf
- No more missing compile flags
- Compile only the objects needed
There are no benefits to Univalue itself that I can think of. These changes
should be a no-op there, and to downstreams as well until they take advantage
of the new sources.mk.
This also removes the option to use an external univalue to avoid similar ABI
issues with mystery binaries.
Co-authored-by: fanquake <fanquake@gmail.com>
0bc666b053 doc: add info for debugging with relative paths (S3RK)
a8b515c317 configure: keep relative paths in debug info (S3RK)
Pull request description:
This is a follow-up for #20353 that fixes#21885
It also adds a small section to assist debugging without absolute paths in debug info.
ACKs for top commit:
kallewoof:
Tested ACK 0bc666b053
Zero-1729:
Light crACK 0bc666b053
Tree-SHA512: d4b75183c3d3a0f59fe786841fb230581de87f6fe04cf7224e4b89c520d45513ba729d4ad8c0e62dd1dbaaa7a25741f04d036bc047f92842e76c9cc31ea47fb2
4747da3a5b Add syscall sandboxing (seccomp-bpf) (practicalswift)
Pull request description:
Add experimental syscall sandboxing using seccomp-bpf (Linux secure computing mode).
Enable filtering of system calls using seccomp-bpf: allow only explicitly allowlisted (expected) syscalls to be called.
The syscall sandboxing implemented in this PR is an experimental feature currently available only under Linux x86-64.
To enable the experimental syscall sandbox the `-sandbox=<mode>` option must be passed to `bitcoind`:
```
-sandbox=<mode>
Use the experimental syscall sandbox in the specified mode
(-sandbox=log-and-abort or -sandbox=abort). Allow only expected
syscalls to be used by bitcoind. Note that this is an
experimental new feature that may cause bitcoind to exit or crash
unexpectedly: use with caution. In the "log-and-abort" mode the
invocation of an unexpected syscall results in a debug handler
being invoked which will log the incident and terminate the
program (without executing the unexpected syscall). In the
"abort" mode the invocation of an unexpected syscall results in
the entire process being killed immediately by the kernel without
executing the unexpected syscall.
```
The allowed syscalls are defined on a per thread basis.
I've used this feature since summer 2020 and I find it to be a helpful testing/debugging addition which makes it much easier to reason about the actual capabilities required of each type of thread in Bitcoin Core.
---
Quick start guide:
```
$ ./configure
$ src/bitcoind -regtest -debug=util -sandbox=log-and-abort
…
2021-06-09T12:34:56Z Experimental syscall sandbox enabled (-sandbox=log-and-abort): bitcoind will terminate if an unexpected (not allowlisted) syscall is invoked.
…
2021-06-09T12:34:56Z Syscall filter installed for thread "addcon"
2021-06-09T12:34:56Z Syscall filter installed for thread "dnsseed"
2021-06-09T12:34:56Z Syscall filter installed for thread "net"
2021-06-09T12:34:56Z Syscall filter installed for thread "msghand"
2021-06-09T12:34:56Z Syscall filter installed for thread "opencon"
2021-06-09T12:34:56Z Syscall filter installed for thread "init"
…
# A simulated execve call to show the sandbox in action:
2021-06-09T12:34:56Z ERROR: The syscall "execve" (syscall number 59) is not allowed by the syscall sandbox in thread "msghand". Please report.
…
Aborted (core dumped)
$
```
---
[About seccomp and seccomp-bpf](https://en.wikipedia.org/wiki/Seccomp):
> In computer security, seccomp (short for secure computing mode) is a facility in the Linux kernel. seccomp allows a process to make a one-way transition into a "secure" state where it cannot make any system calls except exit(), sigreturn(), and read() and write() to already-open file descriptors. Should it attempt any other system calls, the kernel will terminate the process with SIGKILL or SIGSYS. In this sense, it does not virtualize the system's resources but isolates the process from them entirely.
>
> […]
>
> seccomp-bpf is an extension to seccomp that allows filtering of system calls using a configurable policy implemented using Berkeley Packet Filter rules. It is used by OpenSSH and vsftpd as well as the Google Chrome/Chromium web browsers on Chrome OS and Linux. (In this regard seccomp-bpf achieves similar functionality, but with more flexibility and higher performance, to the older systrace—which seems to be no longer supported for Linux.)
ACKs for top commit:
laanwj:
Code review and lightly tested ACK 4747da3a5b
Tree-SHA512: e1c28e323eb4409a46157b7cc0fc29a057ba58d1ee2de268962e2ade28ebd4421b5c2536c64a3af6e9bd3f54016600fec88d016adb49864b63edea51ad838e17
No longer special case a set of warnings, to make up our own -Werror,
just use -Werror outright. This shouldn't really have any effect on
existing builders, who were already using --enable-werror, and is more
inline with what they would expect --enable-werror to be, which is
erroring on any/all warnings.
We keep -Wno-error=return-type because we know that is broken when using
mingw-w64. It should only be applied when cross-compiling for Windows.
Co-authored-by: Hennadii Stepanov <32963518+hebasto@users.noreply.github.com>
It was [pointed out in #23030](https://github.com/bitcoin/bitcoin/pull/23030#issuecomment-922893367) that we might be able to get rid of our weak linking of [`getauxval()`](https://man7.org/linux/man-pages/man3/getauxval.3.html) (`HAVE_WEAK_GETAUXVAL`) entirely, with only Android being a potential holdout:
> I wonder if it's time to get rid of HAVE_WEAK_GETAUXVAL. I think it's confusing. Either we build against a C library that has this functionality, or not. We don't do this weak linking thing for any other symbols and recently got rid of the other glibc backwards compatibility stuff.
> Unless there is still a current platform that really needs it (Android?), I'd prefer to remove it from the build system, it has caused enough issues.
After looking at Android further, it would seem that given we are moving to using `std::filesystem`, which [requires NDK version 22 and later](https://github.com/android/ndk/wiki/Changelog-r22), and `getauxval` has been available in the since [API version 18](https://developer.android.com/ndk/guides/cpu-features#features_using_libcs_getauxval3), that shouldn't really be an issue. Support for API levels < 19 will be dropped with the NDK 24 release, and according to [one website](https://apilevels.com/), supporting API level 18+ will cover ~99% of devices. Note that in the CI we currently build with NDK version 22 and API level 28.
The other change in this PR is removing the include of headers for ARM intrinsics, from the check for strong `getauxval()` support in configure, as they shouldn't be needed. Including these headers also meant that the check would basically only succeed when building for ARM. This would be an issue if we remove weak linking, as we wouldn't detect `getauxval()` as supported on other platforms. Note that we also use `getauxval()` in our RNG when it's available.
I've checked that with these changes we detect support for strong `getauxval()` on Alpine (muslibc). On Linux, previously we'd be detecting support for weak getauxval(), now we detect strong support. Note that we already require glibc 2.17, and `getauxval()` was introduced in `2.16`.
This is an alternative / supersedes #23030.
`crc32c`'s hardware accelerated code doesn't handle ARM 32-bit at all.
Make the check in `configure.ac` check for this architecture explicitly.
For the release binaries, the current `configure.ac` check happens
to work: it enables it on aarch64 but disables it for armhf. However
some combination of compiler version and settings might ostensibly cause
this check to succeed on armhf (as reported on IRC). So make the 64-bit
platform requirement explicit.
3ec633ef1a build: improve check for ::(w)system (fanquake)
Pull request description:
`AC_DEFINE()` takes `HAVE_STD__SYSTEM || HAVE_WSYSTEM` literally, meaning you
end up with the following in bitcoin-config.h:
```cpp
/* std::system or ::wsystem */
#define HAVE_SYSTEM HAVE_STD__SYSTEM || HAVE_WSYSTEM
```
This works for the preprocessor, because `HAVE_SYSTEM`, is defined, just unusually. Remove this in favor of setting `have_any_system` in either case, given we don't actually use `HAVE_STD__SYSTEM` or `HAVE_WSYSTEM`, and defining `HAVE_SYSTEM` to 1 thereafter.
ACKs for top commit:
laanwj:
Code review ACK 3ec633ef1a
Tree-SHA512: 02c39ba3179136ec1dc28df026b7fa5d732914c85622298ba7ec880f1ae9324208d322a47be451a5c2ff2e165ad1d446bae92e7018db8e517e7ac38fca25a0a3
At this point, or minimum required glibc is implicitly 2.18, due to
thread_local support being enabled by default. However, users can
disable thread_local support to maintain 2.17 ccompat for now, which is
currently done in the Guix build.
`AC_DEFINE()` takes `HAVE_STD__SYSTEM || HAVE_WSYSTEM` literally, meaning you
end up with the following in bitcoin-config.h:
```cpp
/* std::system or ::wsystem */
#define HAVE_SYSTEM HAVE_STD__SYSTEM || HAVE_WSYSTEM
```
This works for the preprocessor, because `HAVE_SYSTEM`, is defined, just unusually.
Remove this in favor of defining `HAVE_SYSTEM` to 1 in either case, given we
don't actually use `HAVE_STD__SYSTEM` or `HAVE_WSYSTEM`. We just use ::system if
we aren't building for Windows.
2445df4eb3 build: Fix macOS Apple Silicon build with miniupnpc and libnatpmp (Hennadii Stepanov)
Pull request description:
On master (7a49fdc581) the `configure` script does not pick up Homebrew's `miniupnpc` and `libnatpmp` packages on macOS Apple Silicon:
```
% ./configure --with-miniupnpc
...
checking for miniupnpc/miniupnpc.h... no
checking for miniupnpc/upnpcommands.h... no
checking for miniupnpc/upnperrors.h... no
...
checking whether to build with support for UPnP... configure: error: "UPnP requested but cannot be built. Use --without-miniupnpc."
```
```
% ./configure --with-natpmp
...
checking for natpmp.h... no
...
checking whether to build with support for NAT-PMP... configure: error: NAT-PMP requested but cannot be built. Use --without-natpmp
```
The preferred Homebrew [prefix for Apple Silicon](https://docs.brew.sh/Installation) is `/opt/homebrew`. Therefore, if we do not use `pkg-config` to detect packages, we should set the `CPPFLAGS` and `LDFLAGS` variables for them explicitly.
ACKs for top commit:
Zero-1729:
re-tACK 2445df4eb3 (re-tested on an M1 Machine running macOS 11.4).
jarolrod:
re-ACK 2445df4eb3
Tree-SHA512: d623d26492f463812bf66ca519847ff4b23d517466b6c51c3caf3642a582d02e5f03ce57915742b29f01bf9bceb731a3978ef9a5fdc82e568bcb62548eda758a
b367745cfe ci: Make Cirrus CI Windows build with --enable-werror (Hennadii Stepanov)
c713bb2b24 Fix Windows build with --enable-werror on Ubuntu Focal (Hennadii Stepanov)
Pull request description:
This PR makes possible to cross-compile Windows build with `--enable-werror --enable-suppress-external-warnings`.
Some problems are fixed, others are silenced.
Also `--enable-werror` is enabled for Cirrus CI Windows build (the last one on Cirrus CI without `--enable-werror`).
ACKs for top commit:
practicalswift:
cr ACK b367745cfe19f6de3f44b3adc90fa08e36e44bb6: patch looks correct
laanwj:
Code review ACK b367745cfe
vasild:
ACK b367745cfe
jarolrod:
ACK b367745cfe
Tree-SHA512: 64f5c99b7dad4c0efce80cd45d7074f275bd8411235dc9e0841287bdab64b812c6f8f9d632c35531d0b8210148531f53aaaac77be7699b29d2d6aaae304dbee0
The preferred Homebrew prefix for Apple Silicon is /opt/homebrew.
Therefore, if we do not use pkg-config to detect packages, we should set
the CPPFLAGS and LDFLAGS variables for them explicitly.
e4c8bb62e4 build: Fix undefined reference to __mulodi4 (Hennadii Stepanov)
Pull request description:
When compiling with clang on 32-bit systems the `__mulodi4` symbol is defined in compiler-rt only.
Fixes#21294.
See more:
- https://bugs.llvm.org/show_bug.cgi?id=16404
- https://bugs.llvm.org/show_bug.cgi?id=28629
ACKs for top commit:
MarcoFalke:
tested-only ACK e4c8bb62e4
luke-jr:
utACK e4c8bb62e4
fanquake:
ACK e4c8bb62e4 - it's a bit of an awkward workaround to carry, but at-least it's contained to the fuzzers.
Tree-SHA512: 93edb4ed568027702b1b9aba953ad50889b834ef97fde3cb99d1ce70076d9c00aa13f95c86b12d6f59b24fa90108d93742f920e15119901a2848fb337ab859a1
fanquake