This has been around since the introduction of autotools. However at
this point I'm not sure we'd every want to suppress all warnings when
performing a build, and given that CXX FLAGS will have been overriden
when cross-compiling for Windows (using depends), this would rarely,
if-ever be used anyways.
From https://gcc.gnu.org/onlinedocs/gcc/Warning-Options.html:
-w
Inhibit all warning messages.
As is, this sanity check doesn't seem to be testing fdelt_chk, because
passing a value of "0" to FD_SET wont cause the compiler to insert any
calls to fdelt_chk().
The documentation is a little misleading. If we actually triggered fdelt_chk
at runtime, bitcoind would abort. I think this check would be better replaced
(if possible) by additional checks in security-check.py.
The compiler may insert a call to fdelt_warn() (aliased with fdelt_chk
in glibc) at compile time if it can determine that an invalid value is
being passed to FD_SET.
These checks are essentially; value < 0 or value >= FD_SETSIZE along
with a check for wether the value is a compile time constant.
If the compiler can determine an invalid value is being passed, a call
to fdelt_warn will be inserted. Passing 0 should never cause a call to
be inserted.
You can check this after compiling:
```bash
objdump -dC bitcoind | grep sanity_fdelt
...
0000000000399d20 <sanity_test_fdelt()>:
399d20: 48 81 ec 98 00 00 00 sub $0x98,%rsp
399d27: b9 10 00 00 00 mov $0x10,%ecx
399d2c: 64 48 8b 04 25 28 00 mov %fs:0x28,%rax
399d33: 00 00
399d35: 48 89 84 24 88 00 00 mov %rax,0x88(%rsp)
399d3c: 00
399d3d: 31 c0 xor %eax,%eax
399d3f: 48 89 e7 mov %rsp,%rdi
399d42: fc cld
399d43: f3 48 ab rep stos %rax,%es:(%rdi)
399d46: 48 8b 84 24 88 00 00 mov 0x88(%rsp),%rax
399d4d: 00
399d4e: 64 48 33 04 25 28 00 xor %fs:0x28,%rax
399d55: 00 00
399d57: 75 0d jne 399d66 <sanity_test_fdelt()+0x46>
399d59: b8 01 00 00 00 mov $0x1,%eax
399d5e: 48 81 c4 98 00 00 00 add $0x98,%rsp
399d65: c3 retq
399d66: e8 85 df c8 ff callq 27cf0 <__stack_chk_fail@plt>
399d6b: 0f 1f 44 00 00 nopl 0x0(%rax,%rax,1)
```
To test, you could modify this test to pass -1 to FD_SET, and check
that a call to fdelt_warn() is inserted, and that running bitcoind
fails. i.e:
```bash
0000000000399d20 <sanity_test_fdelt()>:
399d20: 48 81 ec 98 00 00 00 sub $0x98,%rsp
399d27: b9 10 00 00 00 mov $0x10,%ecx
399d2c: 64 48 8b 04 25 28 00 mov %fs:0x28,%rax
399d33: 00 00
399d35: 48 89 84 24 88 00 00 mov %rax,0x88(%rsp)
399d3c: 00
399d3d: 31 c0 xor %eax,%eax
399d3f: 48 89 e7 mov %rsp,%rdi
399d42: fc cld
399d43: f3 48 ab rep stos %rax,%es:(%rdi)
399d46: 48 c7 c7 ff ff ff ff mov $0xffffffffffffffff,%rdi
399d4d: e8 3e ff ff ff callq 399c90 <__fdelt_warn>
399d52: 0f b6 04 24 movzbl (%rsp),%eax
399d56: 83 e0 01 and $0x1,%eax
399d59: 48 8b 94 24 88 00 00 mov 0x88(%rsp),%rdx
399d60: 00
399d61: 64 48 33 14 25 28 00 xor %fs:0x28,%rdx
399d68: 00 00
399d6a: 75 08 jne 399d74 <sanity_test_fdelt()+0x54>
399d6c: 48 81 c4 98 00 00 00 add $0x98,%rsp
399d73: c3 retq
399d74: e8 77 df c8 ff callq 27cf0 <__stack_chk_fail@plt>
399d79: 0f 1f 80 00 00 00 00 nopl 0x0(%rax)
```
```bash
./src/bitcoind
*** buffer overflow detected ***: src/bitcoind terminated
Aborted
```
a029805f57 build: remove -Qunused-arguments workaround for clang + ccache (fanquake)
Pull request description:
This was added in 386efb7695 to address spammy Clang warnings when building with ccache.
The issue was addressed in [ccache 3.2](https://bugzilla.samba.org/show_bug.cgi?id=8118), and from a look at most major distros, it's only Debian Jessie that has a version of ccache older than that ([3.1](https://packages.debian.org/jessie/ccache)).
Therefore I think it's acceptable to drop this workaround, and re-enable warnings for unused driver arguments (when compiling using Clang and ccache).
ACKs for top commit:
hebasto:
ACK a029805f57.
vasild:
utACK a029805f57
Tree-SHA512: f887b9bd12f9c1c8d209943b86e8dafe33cfd1572912f2cafabe08ffe403973e48f0f7289280a8c6db9263c57aad43fbd4bb72f42db762eb090f3b1ef0538f43
03da4c7781 build: make linker checks more robust (Cory Fields)
Pull request description:
Check for a flag to turn linker warnings into errors. When flags are passed to
linkers via the compiler driver using a -Wl,-foo flag, linker warnings may be
swallowed rather than bubbling up.
This is one of [Corys commits](b9acd3d33e) that I've modified to also add `-Wl,-fatal_warnings`
for darwin.
ACKs for top commit:
vasild:
re-ACK 03da4c778
Tree-SHA512: 212031d619ed88e52aaae30cf3b711681d72c4d670884406403605d1d86c784c84cb07e2e0d6c30926e659db8f14f8dabd5af3de5291637f8080d6dfee358248
71f183a49b build: warn on potentially uninitialized reads (Vasil Dimov)
Pull request description:
* Enable `conditional-uninitialized` warning class to show potentially uninitialized
reads.
* Fix the sole such warning in Bitcoin Core in `GetRdRand()`: `r1` would be
set to `0` on `rdrand` failure, so initializing it to `0` is a non-functional
change.
ACKs for top commit:
practicalswift:
ACK 71f183a49b
laanwj:
ACK 71f183a49b
Tree-SHA512: 2c1d8caacd86424b16a9d92e5df19e0bedb51ae111eecad7e3bfa46447bc88e5fff1f32dacf6c4a28257ebb3d87e79f80f074ce2c523ce08b1a0c0a67ab44204
Check for a flag to turn linker warnings into errors. When flags are passed to
linkers via the compiler driver using a -Wl,-foo flag, linker warnings may be
swallowed rather than bubbling up.
Co-authored-by: fanquake <fanquake@gmail.com>
Stop the build if a warning is emitted due to `-Wgnu` and
`--enable-werror` has been used. As usual - this would help notice such
a warning that is about to be introduced in new code.
This is a followup to
https://github.com/bitcoin/bitcoin/pull/18088
build: ensure we aren't using GNU extensions
Enable -Wconditional-uninitialized to warn on potentially uninitialized
reads.
Fix the sole such warning in Bitcoin Core in GetRdRand(): r1 would be
set to 0 on rdrand failure, so initializing it to 0 is a non-functional
change.
From "Intel 64 and IA-32 ArchitecturesSoftware Developer's Manual" [1],
page 1711: "CF=1 indicates that the data in the destination is valid.
Otherwise CF=0 and the data in the destination operand will be returned
as zeros for the specified width."
[1] https://software.intel.com/sites/default/files/managed/39/c5/325462-sdm-vol-1-2abcd-3abcd.pdf
c31cbe7cfe Add C++17 test to Travis (Pieter Wuille)
7829685e27 Add configure option for c++17 (Pieter Wuille)
0fbde488b2 Support conversion between Spans of compatible types (Pieter Wuille)
7cbfebbf3d Update ax_cxx_compile_stdcxx.m4 (Pieter Wuille)
Pull request description:
This adds a `--enable-c++17` option to the configure script, fixes the only C++17 incompatibility (with a commit taken from #18468), and adds a Travis test for it.
This is all off by default, and release builds remain C++11.
It implements the first step of the plan in https://github.com/bitcoin/bitcoin/issues/16684.
ACKs for top commit:
elichai:
tACK c31cbe7cfe
practicalswift:
Tested ACK c31cbe7cfe
hebasto:
ACK c31cbe7cfe, tested on Linux Mint 19.3 both C++11 and C++17 modes. Compiled and passed tests locally.
Tree-SHA512: a4b00776dbceef9c12abbb404c6bcd48f7916ce24c8c7a14116355f64e817578b7fcddbedd5ce435322319d1e4de43429b68553f4d96d970c308fe3e3e59b9d1
182dbdf0f4 util: Detect posix_fallocate() instead of assuming (Vasil Dimov)
Pull request description:
Don't assume that `posix_fallocate()` is available on Linux and not
available on other operating systems. At least FreeBSD has it and we
are not using it.
Properly check whether `posix_fallocate()` is present and use it if it
is.
ACKs for top commit:
laanwj:
ACK 182dbdf0f4
Tree-SHA512: f9ed4bd661f33ff6b2b1150591e860b3c1f44e12b87c35e870d06a7013c4e841ed2bf17b41ad6b18fe471b0b23a4b5e42cf1400637180888e0bc56c254fe0766
b155fcda51 doc: fix typo in configure.ac (fanquake)
20a30922fb doc: note why we can't use thread_local with glibc back compat (fanquake)
Pull request description:
Given that we went through a [gitian build](https://github.com/bitcoin/bitcoin/pull/18681) to remember why this is the case, we might as well make a note of it in configure.ac.
[From #18681](https://github.com/bitcoin/bitcoin/pull/18681#issuecomment-615526634):
Looking at the Linux build log, this has failed with:
```bash
Checking glibc back compat...
bitcoind: symbol __cxa_thread_atexit_impl from unsupported version GLIBC_2.18
bitcoind: failed IMPORTED_SYMBOLS
bitcoin-cli: symbol __cxa_thread_atexit_impl from unsupported version GLIBC_2.18
bitcoin-cli: failed IMPORTED_SYMBOLS
bitcoin-tx: symbol __cxa_thread_atexit_impl from unsupported version GLIBC_2.18
bitcoin-tx: failed IMPORTED_SYMBOLS
bitcoin-wallet: symbol __cxa_thread_atexit_impl from unsupported version GLIBC_2.18
bitcoin-wallet: failed IMPORTED_SYMBOLS
test/test_bitcoin: symbol __cxa_thread_atexit_impl from unsupported version GLIBC_2.18
test/test_bitcoin: failed IMPORTED_SYMBOLS
bench/bench_bitcoin: symbol __cxa_thread_atexit_impl from unsupported version GLIBC_2.18
bench/bench_bitcoin: failed IMPORTED_SYMBOLS
qt/bitcoin-qt: symbol __cxa_thread_atexit_impl from unsupported version GLIBC_2.18
```
`__cxa_thread_atexit_impl` is used for [thread_local variable destruction](https://sourceware.org/glibc/wiki/Destructor%20support%20for%20thread_local%20variables):
> To implement this support, glibc defines __cxa_thread_atexit_impl exclusively for use by libstdc++ (which has the __cxa_thread_atexit to wrap around it), that registers destructors for thread_local variables in a list. Upon thread or process exit, the destructors are called in reverse order in which they were added.
As suggested, this only became available in glibc 2.18. From the [2.18 release notes](https://sourceware.org/legacy-ml/libc-alpha/2013-08/msg00160.html):
> * Add support for calling C++11 thread_local object destructors on thread
and program exit. This needs compiler support for offloading C++11
destructor calls to glibc.
ACKs for top commit:
hebasto:
ACK b155fcda51
Tree-SHA512: 5b9567e4a70598a4b0b91956f44ae0d93091db17c84cbf9817dac6cfa992c97d3438a8b1bb66644c74891f2149e44984daed445d22de93ca8858c5b0eabefb40
Don't assume that `posix_fallocate()` is available on Linux and not
available on other operating systems. At least FreeBSD has it and we
are not using it.
Properly check whether `posix_fallocate()` is present and use it if it
is.
b919efadff depends: Use default macos clang compiler (Russell Yanofsky)
d54f64c6c7 Add multiprocess travis configuration (Russell Yanofsky)
787f40668d Set LD_LIBRARY_PATH consistently in travis tests (Russell Yanofsky)
d630646662 libmultiprocess depends build (Russell Yanofsky)
e6e44eedd5 Multiprocess build changes (Russell Yanofsky)
Pull request description:
This PR is part of the [process separation project](https://github.com/bitcoin/bitcoin/projects/10).
This splits autotools, depends build, and travis changes out of #10102, so code changes and build system changes can be reviewed separately.
ACKs for top commit:
hebasto:
re-ACK b919efadff, rebased only since my [previous](https://github.com/bitcoin/bitcoin/pull/16367#issuecomment-605514556) review.
Tree-SHA512: ebc5e403cc99a0d9629ed7fe1595e01d57e6d1255cbf03968a3196ff6f528f734c78060fdc065724ee1f923bcc5aa2b29470fcb36a7f15957eb57c76d58178a4
01a3392b1b Drop bitcoin-wallet dependency on libevent (Russell Yanofsky)
0660119ac3 Drop unintended bitcoin-tx dependency on libevent (Russell Yanofsky)
Pull request description:
This fixes compile errors trying to build bitcoin-tx and bitcoin-wallet without libevent, which were reported by Luke Dashjr in https://github.com/bitcoin/bitcoin/issues/18465
The fix avoiding `bitcoin-tx` dependency on libevent just adds a conditional build rule. This is implemented in the first commit (more details in commit description).
The fix avoiding `bitcoin-wallet` dependency on libevent requires minor code changes, because `bitcoin-wallet` (unlike `bitcoin-tx`) links against code that calls `urlDecode` / `evhttp_uridecode`. This fix is implemented in the second commit (again details in the commit description).
ACKs for top commit:
jonasschnelli:
utACK 01a3392b1b.
Tree-SHA512: d2245e912ab494cccceeb427a1eca8e55b01a0006ff93eebcfb5461ae7cecd1083ac2de443d9db036b18bdc6f0fb615546caaa20c585046f66d234937f74870a
Now that 0.20 branch has been split off, master is 0.20.99 (pre-0.21).
Also clean out release notes.
Tree-SHA512: bba6133ae9708f75206c8934901b9f9909a233330f4dfefb3c24175bf8e11631cdc89a5d24a22421a73083f7eb743e977db8020b87dfbd3c1e6043929a19a285
This was added in 386efb7695 to
address spammy Clang warnings when building with ccache.
The issue was addressed in ccache 3.2
(https://bugzilla.samba.org/show_bug.cgi?id=8118, Nov 2014),
and from a look at all major distros, it's only Debian Jessie
that has a version of ccache older than that (3.1).
Therefore I think it's acceptable to drop this workaround, and
re-enable warnings for unused driver arguments (when compiling
using Clang).
autotools and automake changes to support multiprocess execution.
This adds a new --enable-multiprocess flag, and build configuration code to
detect libraries needed for multiprocess support. The --enable-multiprocess
flag builds new bitcoin-node and bitcoin-gui executables, which are updated in
https://github.com/bitcoin/bitcoin/pull/10102 to communicate across processes.
But for now they are functionally equivalent to existing bitcoind and
bitcoin-qt executables.
Don't include util/url.cpp to libbitcoin_util.a when libevent isn't available.
This fixes a compile error trying to build bitcoin-tx without libevent reported
by Luke Dashjr in https://github.com/bitcoin/bitcoin/issues/18465Fixes#18465
faf7d4fa86 build: Add cov_fuzz target (MarcoFalke)
fac71e364e build: link fuzz/test_runner.py for out-of-tree builds (MarcoFalke)
faf2c5aca0 build: Remove unused USE_COVERAGE (MarcoFalke)
Pull request description:
Only libFuzzer is supported right now, so clang is required. Thus, this needs a workaround such as https://github.com/bitcoin/bitcoin/issues/12602#issuecomment-562788247
Can be tested with:
```
mkdir build && cd build
../configure --enable-fuzz --with-sanitizers=fuzzer --enable-lcov --enable-lcov-branch-coverage CC=clang CXX=clang++
make $MAKEJOBS
make cov_fuzz
ACKs for top commit:
practicalswift:
ACK faf7d4fa86
Tree-SHA512: 6828f8f81d95f6781713d0b09d7eba2ffdb50217e09ca839db61791a4ed70024859c7a0cb01d9eede79166d574dd57ece01f9d9fe2610d4a72a4ca4a4ce0b838
sysctl() on *BSD takes a "const int *name", whereas sysctl() on macOS
it takes an "int *name". So our configure check and sysctl() detection on
macOS currently fails:
```bash
/usr/include/sys/sysctl.h:759:9: note: candidate function not viable:
no known conversion from 'const int [2]' to 'int *' for 1st argument
int sysctl(int *, u_int, void *, size_t *, void *, size_t);
```
This change removes the name argument from the sysctl() detection check,
meaning we will detect correctly on macOS and *BSD.
For consistency we also switch to using the more generic, non-const
version of the name parameter in the rest of our usage.
This change adds to the BITCOIN_QT_CONFIGURE script ability to use
pkg-config for MinGW. All of the non-pkg-config paths are removed as
needless.
If depends is built with DEBUG=1 the configure script fails to pickup
Qt:
- for macOS host (similar, but not the same as issue 16391)
- for Windows host (regression)
fae86c38bc util: Remove unused MilliSleep (MarcoFalke)
fa9af06d91 scripted-diff: Replace MilliSleep with UninterruptibleSleep (MarcoFalke)
fa4620be78 util: Add UnintrruptibleSleep (MarcoFalke)
Pull request description:
We don't use the interruptible feature of boost's sleep anywhere, so replace it with the sleep in `std::thread`
ACKs for top commit:
ajtowns:
ACK fae86c38bc quick code review
practicalswift:
ACK fae86c38bc -- patch looks correct
sipa:
Concept and code review ACK fae86c38bc
fanquake:
ACK fae86c38bc - note that an instance of `DHAVE_WORKING_BOOST_SLEEP_FOR` was missed in the [linter](https://github.com/bitcoin/bitcoin/blob/master/test/lint/extended-lint-cppcheck.sh#L69), but that can be cleaned up later.
Tree-SHA512: 7c0f8eb197664b9f7d9fe6c472c77d384f11c797c913afc31de4b532e3b4fd9ea6dd174f92062ff9d1ec39b25e0900ca7c597435add87f0f2477d9557204848c
d76894987d logging: enable thread_local usage on macOS (fanquake)
Pull request description:
Now that we're building against a newer SDK (`10.14`), we should be able to enable `thread_local` usage on macOS. Have tested building and running locally, as well as cross-compiling and running the binaries on a macOS 10.14 system.
#### master 8a56f79d49
```bash
src/bitcoind -logthreadnames=1
2020-02-06T04:38:33Z [] Bitcoin Core version v0.19.99.0-8a56f79d4 (release build)
2020-02-06T04:38:33Z [] Assuming ancestors of block 00000000000000000005f8920febd3925f8272a6a71237563d78c2edfdd09ddf have valid signatures.
2020-02-06T04:38:33Z [] Setting nMinimumChainWork=000000000000000000000000000000000000000008ea3cf107ae0dec57f03fe8
2020-02-06T04:38:33Z [] Using the 'sse4(1way),sse41(4way),avx2(8way)' SHA256 implementation
2020-02-06T04:38:33Z [] Using RdSeed as additional entropy source
```
#### this PR d76894987d
```bash
checking for thread_local support... yes
...
src/bitcoind -logthreadnames=1
2020-02-06T04:17:49Z [net] net thread start
2020-02-06T04:17:49Z [opencon] opencon thread start
2020-02-06T04:17:49Z [dnsseed] dnsseed thread start
2020-02-06T04:17:49Z [init] init message: Done loading
2020-02-06T04:17:49Z [msghand] msghand thread start
2020-02-06T04:17:49Z [addcon] addcon thread start
...
2020-02-06T04:17:54Z [init] tor: Thread interrupt
2020-02-06T04:17:54Z [init] Shutdown: In progress...
```
From the [Xcode 8 release notes](https://developer.apple.com/library/archive/releasenotes/DeveloperTools/RN-Xcode/Chapters/Introduction.html#//apple_ref/doc/uid/TP40001051-CH1-SW78)
> C++ now supports the thread_local keyword, which declares thread-local storage (TLS) and supports C++ classes with non-trivial constructors and destructors. (9001553)
ACKs for top commit:
jonasschnelli:
Tested ACK d76894987d
nijynot:
ACK d768949
hebasto:
ACK d76894987d
Tree-SHA512: 48f3e4104b80bd7b6aedcef10bb1957b073530130f33af7c5cb59e876ac3f5480e53d7af1c0b226d809fe9eef1add3d6c3fb6de4af174966202c6030060ea823
677fb8e923 test: Add ubsan surpression for crc32c (Wladimir J. van der Laan)
8e68bb1dde build: Disable msvc warning 4722 for leveldb build (Aaron Clauson)
be23949765 build: MSVC changes for leveldb update (Aaron Clauson)
9ebdf04757 build: CRC32C build system integration (Wladimir J. van der Laan)
402252a808 build: Add LCOV exception for crc32c (Wladimir J. van der Laan)
3a037d0067 test: Add crc32c exception to various linters and generation scripts (Wladimir J. van der Laan)
84ff1b2076 test: Add crc32c to subtree check linter (Wladimir J. van der Laan)
7cf13a5134 doc: Add crc32c subtree to developer notes (Wladimir J. van der Laan)
24d02a9ac0 build: Update build system for new leveldb (Wladimir J. van der Laan)
2e1819311a Squashed 'src/crc32c/' content from commit 224988680f7673cd7c769963d4035cb315aa3388 (Wladimir J. van der Laan)
66480821b3 Squashed 'src/leveldb/' changes from f545dfabff4c2e9836efed094dba99a34fbc6b88..f8ae182c1e5176d12e816fb2217ae33a5472fdd7 (Wladimir J. van der Laan)
Pull request description:
This updates leveldb to currently newest upstream commit 0c40829872:
- CRC32C hardware acceleration is now an external library [crc32c](https://github.com/google/crc32c). This adds acceleration on ARM, and should be faster on x86 because of using prefetch. It also makes it easy to support similar instruction sets on other platforms in the future.
- Thread handling uses C++11, instead of platform specific code.
- Native windows environment was added. No need to maintain our own hacky one, anymore.
- Upstream now builds using CMake. This doesn't mean we need to use that (phew), but internal configuration changed to a a series of checks, instead of OS profiles. This means the blanket error "Cannot build leveldb for $host. Please file a bug report' is removed.
All changes: a53934a3ae...0c40829872
Pretty much all our changes have been subsumed by upstream, so we figured it was cleaner to start over with a new branch from upstream with the still-relevant patches applied: https://github.com/bitcoin-core/leveldb/tree/bitcoin-fork-new
There's quite some testing to be done (see below). See https://github.com/bitcoin-core/leveldb/issues/25 and https://github.com/bitcoin-core/leveldb/pull/26 for more history and context.
TODO:
- [x] Subtree `crc32c`
- [x] Make linters happy about crc32 subtree
- [x] Integrate `crc32c` library into build system
- [x] MSVC build system
ACKs for top commit:
sipa:
ACK 677fb8e923
Tree-SHA512: 37ee92a750e053e924bc4626b12bb3fd81faa9f8c5ebaa343931fee810c45ba05aa6051fdea82535fa351bf2be7297801b98af9469865fc5ead771650a5d6240
This flag was used when building 32-bit Windows executables, which we no-longer
do, and is not accepted by the linker for any of the hosts we currently build
for. i.e:
```bash
checking whether the linker accepts -Wl,--large-address-aware... no
```
--large-address-aware
If given, the appropriate bit in the "Characteristics" field of the COFF
header is set to indicate that this executable supports virtual addresses
greater than 2 gigabytes. This should be used in conjunction with the /3GB
or /USERVA=value megabytes switch in the "[operating systems]" section of
the BOOT .INI. Otherwise, this bit has no effect. [This option is specific
to PE targeted ports of the linker]
You can check that the appropriate bit in the COFF header of our current
Windows binaries is still be set using dumpbin. i.e:
```powershell
dumpbin /headers .\bitcoind.exe
FILE HEADER VALUES
<snip>
26 characteristics
Executable
Line numbers stripped
Application can handle large (>2GB) addresses
```
