- guard PKG_PROG_PKG_CONFIG with an m4_ifdef. If not building for windows,
require it
- add nops as necessary in case the ifdef reduces the if/then to nothing
- AC_SUBST some missing _LIBS. These were split out over time, but not all were
properly substituted. They continued to work if pkg-config is installed
because it does the AC_SUBST itself
- create a script to handle split debug. This will also eventually need to check
targets, and use dsymutil for osx.
- update config.guess/config.sub for bdb for aarch64.
- temporarily disable symbol checks for arm/aarch64
- quit renaming to linux32/linux64 and use the host directly
This also adds a hack to work around an Ubuntu bug in the gcc-multilib package:
https://bugs.launchpad.net/ubuntu/+source/gcc-defaults-armhf-cross/+bug/1347820
The problem is that gcc-multilib conflicts with the aarch toolchain.
gcc-multilib installs a symlink that points
/usr/include/asm -> /usr/include/x86_64-linux-gnu/asm.
Without this link, gcc -m32 can't find asm/errno.h (and others), since
/usr/include/x86_64-linux-gnu isn't in its default include path. But
/usr/include/i386-linux-gnu is (though it doesn't exist on disk).
So work around the problem by linking
/usr/include/i386-linux-gnu/asm -> /usr/include/x86_64-linux-gnu/asm.
The symlink fix is actually quite reasonable, but echoing the password into
sudo is nasty, and should probably be addressed in gitian itself. It makes more
sense to enable passwordless sudo for the build user by default.
- Link pull-tester/rpc-tests.py to the build dir
- Add the build-dir's config to the python path so that tests can find it
- The tests themselves are in srcdir
- Clean up __pycache__ in 'make clean'
7df9224 doc: Add note about new build/test requirements to release notes (Wladimir J. van der Laan)
2aacc72 build: update ax_cxx_compile_stdcxx to serial 4 (Wladimir J. van der Laan)
a398549 depends: use c++11 (Cory Fields)
67969af build: Enable C++11 build, require C++11 compiler (Wladimir J. van der Laan)
Disabling warnings can be tricky, because doing so can cause a different
compiler to create new warnings about unsupported disable flags. Also, some
warnings don't surface until they're paired with another warning (gcc). For
example, adding "-Wno-foo" won't cause any trouble, but if there's a legitimate
warning emitted, the "unknown option -Wno-foo" will show up as well.
Work around this in 2 ways:
1. When checking to see if -Wno-foo is supported, check for "-Wfoo" instead.
2. Enable -Werror while checking 1.
If "-Werror -Wfoo" compiles, "-Wno-foo" is almost guaranteed to be supported.
-Werror itself is also checked. If that fails to compile by itself, it likely
means that the user added a flag that adds a warning. In that case, -Werror
won't be used while checking, and the build may be extra noisy. The user would
need to fix the bad input flag.
Also, silence 2 more additional warnings that can show up post-c++11.
leveldb's buildsystem causes us a few problems:
- breaks out-of-tree builds
- forces flags used for some tools
- limits cross builds
Rather than continuing to add wrappers around it, simply integrate it into our
build.
Unfortunately, the target namees defined at the Makefile.am level can't be used
for *.in substitution. So these new defines will have to stay synced up with
those targets.
Using the new variables for the deploy targets in the main Makefile.am will
ensure that they stay in sync, otherwise build tests will fail.
This removes the following executables from the binary gitian release:
- test_bitcoin-qt[.exe]
- bench_bitcoin[.exe]
@jonasschnelli and me discussed this on IRC a few days ago - unlike the
normal `bitcoin_tests` which is useful to see if it is safe to run
bitcoin on a certain OS/environment combination, there is no good reason
to include these. Better to leave them out to reduce the download
size.
Sizes from the 0.12 release:
```
2.4M bitcoin-0.12.0/bin/bench_bitcoin.exe
22M bitcoin-0.12.0/bin/test_bitcoin-qt.exe
```
Ubuntu 16.04 "xenial xerus" does not come with Python 2.x by default.
It is possible to install a python-2.7 package, but this has its own
problem: no `python` or `python2` symlink (see #7717).
This fixes the following scripts to work with python 3:
- `make check` (bctest,py, bitcoin-util-test.py)
- `make translate` (extract_strings_qt.py)
- `make symbols-check` (symbol-check.py)
- `make security-check` (security-check.py)
Explicitly call the python commands using $(PYTHON) instead
of relying on the interpreter line at the top of the scripts.
The old configure.ac did not work for a copyright holders string
containing commas due to insufficient quoting. The new one allows this.
While this is, of course, not of direct consequence to the current code
(where the string is "Bitcoin Core"), it should still be fixed now that
the string is actually factored out.
Due to include ordering, defining in one place was not enough to ensure correct
usage. Use global defines so that we don't have to worry abou this ordering.
Also add a comment in configure about the test.
This is ugly, but temporary. boost::filesystem will likely be dropped soon
after c++11 is enabled. Otherwise, we could simply roll our own copy_file. I've
fixed this at the buildsystem level for now in order to avoid mixing in
functional changes.
Explanation:
If boost (prior to 1.57) was built without c++11, it emulated scoped enums
using c++98 constructs. Unfortunately, this implementation detail leaked into
the abi. This was fixed in 1.57.
When building against that installed version using c++11, the headers pick up
on the native c++11 scoped enum support and enable it, however it will fail to
link. This can be worked around by disabling c++11 scoped enums if linking will
fail.
Add an autoconf test to determine incompatibility. At build-time, if native
enums are being used (a c++11 build), and force-disabling them causes a
successful link, we can be sure that there's an incompatibility and enable the
work-around.
Because Python is (going to be) used to run the RPC tests, when
gathering coverage data with lcov, it is explicitly checked, whether
Python is really available.
* Fixes#6679
* Tested with --disable-zmq
* Tested with and without pkgconfig
* Tested with and without zmq installed
Signed-off-by: Johnathan Corgan <johnathan@corganlabs.com>
1) created rpc-tests.py
2) deleted rpc-tests.sh
3) travis.yml points to rpc-tests.py
4) Modified Makefile.am
5) Updated README.md
6) Added tests_config.py and deleted tests-config.sh
7) Modified configure.ac with script to set correct path in tests_config.py
Prevent these warnings in clang 3.6:
./serialize.h:96:9: warning: explicitly assigning value of variable of type 'uint64_t' (aka 'unsigned long') to itself [-Wself-assign]
obj = (obj);
~~~ ^ ~~~
a3874c7 doc: no longer require use of openssl in OpenBSD build guide (Wladimir J. van der Laan)
5978388 build: remove libressl check (Wladimir J. van der Laan)
Now that BIP66 passed, OpenSSL is no longer directly part of the
consensus. What matters is that DER signatures are correctly parsed, and
secp256k1 crypto is implemented correctly (as well as the other
functions we use from OpenSSL, such as random number generation)
This means that effectively, using LibreSSL is not a larger risk than
using another version of OpenSSL.
Remove the specific check for LibreSSL.
Includes the still-relevant part of #6729: make sure CHECK_HEADER is
called using the right CXXFLAGS, not CFLAGS (as AC_LANG is c++).
Benchmarking framework, loosely based on google's micro-benchmarking
library (https://github.com/google/benchmark)
Wny not use the Google Benchmark framework? Because adding Even More Dependencies
isn't worth it. If we get a dozen or three benchmarks and need nanosecond-accurate
timings of threaded code then switching to the full-blown Google Benchmark library
should be considered.
The benchmark framework is hard-coded to run each benchmark for one wall-clock second,
and then spits out .csv-format timing information to stdout. It is left as an
exercise for later (or maybe never) to add command-line arguments to specify which
benchmark(s) to run, how long to run them for, how to format results, etc etc etc.
Again, see the Google Benchmark framework for where that might end up.
See src/bench/MilliSleep.cpp for a sanity-test benchmark that just benchmarks
'sleep 100 milliseconds.'
To compile and run benchmarks:
cd src; make bench
Sample output:
Benchmark,count,min,max,average
Sleep100ms,10,0.101854,0.105059,0.103881
Continues Johnathan Corgan's work.
Publishing multipart messages
Bugfix: Add missing zmq header includes
Bugfix: Adjust build system to link ZeroMQ code for Qt binaries
Checking libcrypto for a function after we've already found a (possibly
different) libcrypto is not what we want to do here.
pkg-config might've found a cross lib while AC_CHECK_LIB may find a different
or native one.
Run a link-test against the lib that's already been found instead.
Three changes to how configure --enable-debug behaves:
1. Preserve user-passed CXXFLAGS/CFLAGS
2. Compile with -DDEBUG_LOCKORDER
3. Add -DDEBUG -DDEBUG_LOCKORDER to CPPFLAGS (since they are preprocessor options)
Until secp256k1 is used for verification there is no reason for Bitcoin
Core's secp256k1 to link against gmp, even if available. Pass a flag to
configure to override the bignum implementation.
This fixes a crash at runtime on ppc64 reported by @gmaxwell.
- Detect endian instead of stopping configure on big-endian
- Add `byteswap.h` and `endian.h` header for compatibility with
Windows and other operating systems that don't come with them
- Update `crypto/common.h` functions to use compat
endian header
This was added a while ago for testing purposes, but was never intended to be
used. Remove it until upstream libsecp256k1 decides that verification is
stable/ready.
Backwards-compatibility for libstdc++ is not limited to straightforward abi
changes. Symbol visibility also needs to be taken into consideration, and
that really can't be addressed simply.
Instead, just static-link libstdc++ for backwards-compat.
This is really a packager's option. While it's helpful to encourage devs to
test this option for daily builds, it's not reliable in several real-world
use-cases. Some older libstdc++ runtimes (freebsd 9, debian wheezy, for
example) fail to properly catch exceptions due to mismatched type_info.
See https://gcc.gnu.org/bugzilla/show_bug.cgi?id=19664 for more info.
2ecd294 Bugfix: configure: Correctly detect "nothing to build" condition (Luke Dashjr)
b7a4ecc Bugfix: Only check for boost when building code that requires it (Luke Dashjr)
a19eeac Bugfix: configure: Check for openssl/ec.h (Luke Dashjr)
fe925e2 Use EXTRA_LIBRARIES instead of noinst_LIBRARIES so we can avoid building unused code (Cory Fields)
Similar to the INCLUDES changes in 6b099402b4, split out LIBS into individual
entries for more fine-grained control.
Also add MINIUPNPC_LIBS which was missing before, and hook it up to
executables.
Qt5 is bottled, so configure won't find it without some help. Use
brew to find out its prefix.
Also, qt5 added the host_bins variable to pkg-config, use it.
Windows needed a few fixups to get the tests running:
1. bitcoin-tx needs a file extension in Windows. Take this opportunity to
add an env file, which pulls variables out of our build config. This can
be extended as needed, for now it's very simple.
2. After #1, split the args out of the exec key in the test data.
3. Correct the line-endings from windows stdout
b144a74 depends: bump miniupnpc to 1.9.20140701. (Cory Fields)
f628127 depends: bump openssl to 1.0.1i (Cory Fields)
9f7f504 build: add -DMINIUPNP_STATICLIB for new version (Cory Fields)
It's only confusing people into thinking that they should mess with
boost versions, which should not be necessary to get bitcoind to work.
If there is a bug in the build system with autodetecting boost it needs
to be solved not worked around.
libminiupnpc changed their required static define to the much more sane
"MINIUPNP_STATICLIB". Sadly, they don't respect the old "STATICLIB" for
back-compat. Define them both since the old one didn't seem to be conflicting
anywhere.
Also go ahead and split out the cppflags so that they can be applied only where
they're needed. This will help us to build dll's from our libs without having
their import/export declspecs poisoned.
When using clang and ccache, builds spew lots of:
Clang: warning: argument unused during compilation
Upstream bug: https://bugzilla.samba.org/show_bug.cgi?id=8118
This is harmless, bug annoying. If ccache is being used and the
-Qunused-arguments flag is supported (clang), use it.
This was committed previously as 4975ae172 and reverted, because the flags were
applied even if the checks didn't pass. This is the same commit, fixed up to
actually disable the functionality when necessary.
Enabled automatically if boost >= 1.49.
See: https://svn.boost.org/trac/boost/ticket/2309
Also, check for a default visibility attribute, so that we can mark future
api functions correctly.
Enabled automatically if boost >= 1.49.
See: https://svn.boost.org/trac/boost/ticket/2309
Also, check for a default visibility attribute, so that we can mark future
api functions correctly.
While we're at it, reduce the use of LIBS as well. This makes dependencies
explicit.
Fixes building with (the not-yet-merged) libsecp256k1 as well.
Github-Pull: #4689
Rebased-By: Wladimir J. van der laan <laanwj@gmail.com>
Rebased-From: 909b347c0e5dda
If clock_gettime is implemented outside of libc (librt in this case), configure
would fail when testing boost. Since clock_gettime is not present on all OSs,
boost only uses it when it can. Check for it in librt and add it to LIBS if
found, but don't fail if it's not (since boost won't be expecting it in this
case).
Also, reverse the link order as necessary for static libs.
Note that it's possible that there are other similar cases for boost, which may
be handled the same way.
This is the first part of a huge effort to rework the handling of dependencies.
To start, this change allows all supported platforms to build against a static
Qt. 5.2.1 and 5.3 have been successfully tested against osx64, win32, win64,
linux32, and linux64.
It also makes a small change to the windows config, to allow linking against
qt builds with or without built-in libjpeg/libpng/libpcre/libz.
The actual build processes to take advantage of these changes (for gitian and
pull-tester) are coming soon. Until then, this should be a no-op.
Note: This is added to our existing automake targets rather than as a
libtool-style lib. The switch to libtool-style targets can come later if it
proves to not add any complications.
In the LookupIntern(), things changed are:
1. Call getaddrinfo_a() instead of getaddrinfo() if available, the former is a sync version of the latter;
2. Try using inet_pton()/inet_addr() to convert the input text to a network addr structure at first, if success the extra name resolving thread inside getaddrinfo_a() could be avoided;
3. An interruption point added in the waiting loop for return from getaddrinfo_a(), which completes the improve for thread responsiveness.
A easy way to see the effect is to kick off a 'bitcoind stop' immediately after 'bitcoind -daemon', before the change it would take several, or even tens of, minutes on a bad network situation to wait for the running bitcoind to exit, now it costs only seconds.
Signed-off-by: Huang Le <4tarhl@gmail.com>
Build logic moves from individual Makefile.am's to include files, which
the main src/Makefile.am includes. This avoids having to manage a gigantic
single Makefile.
TODO: Move the rules from the old Makefile.include to where they actually
belong and nuke the old file.
Log the name of the error as well as the error code if a network problem
happens. This makes network troubleshooting more convenient.
Use thread-safe strerror_r and the WIN32 equivalent FormatMessage.
The year is 2014. All supported operating systems have IPv6 support,
most certainly at build time (this doesn't mean that IPv6 is configured,
of course).
If noone is exercising the functionality to disable it, that means it
doesn't get tested, and IMO it's better to get rid of it.
(it's also not used consistently in RPC/boost and Net code...)
Luke Dashjr