We need to append-to rather than set CXXFLAGS, otherwise we loose -O2 &
-pipe. Currently this results in zeromq being built without optimizations
at all (or whatever the compiler would default too, essentially always -O0).
Bdb is the same, for the CXX portion of its code. C code has been built
with -O2.
Boost has actually been uneffected because it receives -O3 from it's own
build flags.
After two reports on IRC of issues building depends on an Apple M1
machine, it turns out that this option can't be used when targeting
arm-apple-darwin. For now, just use it for x86_64-apple-darwin.
```bash
Apple clang version 12.0.5 (clang-1205.0.22.9)
Target: x86_64-apple-darwin20.4.0
error: option 'cf-protection=return' cannot be specified on this target
error: option 'cf-protection=branch' cannot be specified on this target
2 errors generated.
```
The LLVM Clang we use for cross-compilation supports this option, and it's expected
that any builders on macOS will also be using an Apple Clang that supports it.
2f5dfe4a7f depends: build qt in c++17 mode (fanquake)
104e859c97 builds: don't pass -silent to qt when building in debug mode (fanquake)
e2c500636c depends: build zeromq with -std=c++17 (fanquake)
2374f2fbef depends: build Boost with -std=c++17 (fanquake)
2dde55702d depends: build bdb with -std=c++17 (fanquake)
Pull request description:
In packages where we are passing `-std=c++11` switch to `-std=c++17`, or, `-std=c++1z` in the case of Qt.
This PR also contains a [commit](104e859c97) that improves debug output when building Qt for debugging (`DEBUG=1`).
Now we'll get output like this:
```bash
g++ -c -pipe -ffunction-sections -O2 -fPIC -std=c++11 -fno-exceptions <lots more> ../../corelib/kernel/qcoreapplication.cpp
```
rather than just:
```bash
compiling ../../corelib/kernel/qcoreapplication.cpp
```
Note that when you look at the DEBUG output for these changes when building Qt, you'll see objects being compiled with a mix of C++11 and C++17. The breakdown is roughly:
1. `qmake` built with `-std=c++11`:
```bash
Creating qmake...
make[1]: Entering directory '<trim>/qt/5.9.8-4110fa99945/qtbase/qmake'
g++ -c -o project.o -std=c++11 -ffunction-sections -O2 -g <trim> <trim>/qt/5.9.8-4110fa99945/qtbase/qmake/project.cpp
# when qmake, Qt also builds some of it's corelib, such as corelib/global/qmalloc.cpp
g++ -c -o qmalloc.o -std=c++11 -ffunction-sections -O2 -g <trim> <trim>/qt/5.9.8-4110fa99945/qtbase/src/corelib/global/qmalloc.cpp
```
2. `qmake` is run, and passed our build options, including `-c++std`:
```bash
make[1]: Entering directory '<trim>/qt/5.9.8-4110fa99945/qtbase'
<trim>qt/5.9.8-4110fa99945/qtbase/bin/qmake -o Makefile qtbase.pro -- -bindir <trim>/native/bin -c++std c++1z -confirm-license <trim>
```
3. After some cleaning and configuring, we actually start to build Qt, as well as it's tools and internal libs:
```bash
Building qt...
make[1]: Entering directory '<trim>/qt/5.9.8-4110fa99945/qtbase/src'
# build libpng, zlib etc
gcc -c -m64 -pipe -pipe -O1 <trim> -o .obj/png.o png.c
# build libQt5Bootstrap, using C++11, which again compiles qmalloc.cpp
make[2]: Entering directory '<trim>/qt/5.9.8-4110fa99945/qtbase/src/tools/bootstrap'
g++ -c -pipe -ffunction-sections -O2 -fPIC -std=c++11 <trim> -o .obj/qmalloc.o ../../corelib/global/qmalloc.cpp
# build a bunch of tools like moc, rcc, uic, qfloat16-tables, qdbuscpp2xml, using C++11
g++ -c -pipe -O2 -std=c++11 -fno-exceptions -Wall -W <trim> -o .obj/rcc.o rcc.cpp
# from here, Qt is compiled with -std=c++1z, including qmalloc.cpp, for the third and final time:
g++ -c -include .pch/Qt5Core <trim> -g -Og -fPIC -std=c++1z -fvisibility=hidden <trim> -o .obj/qmalloc.o global/qmalloc.cpp
```
4. Finally, build tools like `lrelease`, `lupdate`, etc, but back to using -std=c++11
```bash
make[1]: Entering directory '<trim>/qt/5.9.8-4110fa99945/qttools/src/linguist/lrelease'
g++ -c -pipe -O2 -std=c++11 -fno-exceptions -Wall -W <trim> -o .obj/translator.o ../shared/translator.cpp
```
If you dump the debug info from the built Qt libs, they should also tell you that they were compiled with `C++17`:
```bash
objdump -g bitcoin/depends/x86_64-pc-linux-gnu/lib/libQt5Core.a
GNU C++17 9.3.0 -m64 -mtune=generic -march=x86-64 -g -O1 -Og -std=c++17 -fPIC -fvisibility=hidden -fvisibility-inlines-hidden -fasynchronous-unwind-tables -fstack-protector-strong -fstack-clash-protection -fcf-protection
```
ACKs for top commit:
laanwj:
Code review ACK 2f5dfe4a7f
practicalswift:
cr ACK 2f5dfe4a7ff12b6b57427374142cdf7e266b73bc: patch looks correct
fjahr:
Code review ACK 2f5dfe4a7f
hebasto:
ACK 2f5dfe4a7f, I have reviewed the code and it looks OK, I agree it can be merged.
Tree-SHA512: fc5e9d7c7518c68349c8228fb1aead829850373efc960c9b8c079096a83d1dad19c62a9730fce5802322bf07e320960fd47851420d429eda0a87c307f4e8b03a
Previously, we specified the target-os in the toolset (and sometimes
used the wrong command line flags), now we have a clear separation,
which is favored by ./bootstrap.sh and ./b2.
This means that all supported OSes will specify the correct target-os=
and toolset= on the command line.
b2 will pickup our user-config.jam just fine, however, bootstrap.sh has
its own toolset autodetect mechanism, which doesn't GAF about our
user-config.jam
This reverts the changes made in merge commit
1b307613604883daea4913a65da30ae073c9dc4d:
This reverts commit b919efadff.
This reverts commit d54f64c6c7.
This reverts commit 787f40668d.
This reverts commit d630646662.
This reverts commit e6e44eedd5.
All other mk files use the package variable consistently except for the two instances here, which have always been here, since depends was introduced in 0.10.
f9af3ced1c Android: add all arch support (Block Mechanic)
d419ca7e32 depends: export dynamic JNI symbols from static qtforandroid.a (Igor Cota)
ed30684d03 Qt: patch androidjnimain.cpp to make sure JNI is initialised when statically compiled (Igor Cota)
e4c319e8a1 builds: remove superfluous config_opts_aarch64_android (Igor Cota)
24ffef0c27 Patch libevent when building for Android (fix arc4random_addrandom) (Igor Cota)
f1e40b3e71 Update bitcoin_qt.m4 (BlockMechanic)
b4057d8261 Define TARGET_OS when host is android (Igor Cota)
80b475f159 Fix Android zlib cross compilation issue (https://stackoverflow.com/questions/21396988/zlib-build-not-configuring-properly-with-cross-compiler-ignores-ar) (Igor Cota)
45f8219015 Add full Android build example command and instructions on getting SDK/NDK (Igor Cota)
b68f2a68c2 Add config opts and patch for aarch64_android build of Qt (Igor Cota)
9c4cb0166e Add ranlib to android.mk hosts file (fix OSX Android NDK build) (Igor Cota)
c2a749c9c1 Add example Android host-platform-triplet and options (Igor Cota)
0b0cff3c61 Add support for building Android dependencies (Igor Cota)
Pull request description:
This allows one to build the dependencies with the Android SDK and goes towards fixing #11844. It has been tested to work with:
`make HOST=aarch64-linux-android ANDROID_API_LEVEL=28 ANDROID_TOOLCHAIN_BIN=/home/user/Android/Sdk/ndk-bundle/toolchains/llvm/prebuilt/linux-x86_64/bin NO_QT=1 NO_WALLET=1`
ACKs for top commit:
Sjors:
ACK f9af3ce. I'm OK with merging and then improving later.
Tree-SHA512: cb805115ebe5c9e33db2bf3eab8628808fe3f50052053d8877d8b8e4406d6fea1ed9e5c4dff85d777fb99c81be6ffb9d95a0e6d32344e728e5e0da6c653e2ce7
The ancient "darwin-4.9.1" profile has long been used to match against
clang, which prior to version 9, reported 4.9.1 as its version when
invoking "clang++ -dumpversion". Presumably this was a historical
compatibility quirk related to Apple's switch from gcc to clang.
This was "fixed" in clang 9.0, so that -dumpversion reports the real
version. Unfortunately that had the side-effect of breaking the
(brittle) boost compiler detection.
Move to the seemingly more-correct "clang-darwin" profile, which passes
the checks and builds correctly.
Also switch to using ar rather than libtool for archiving, as it's what
the clang-darwin profile expects to be using.
Note that because this is using a different profile, some of the final
command-line arguments end up changing. The changes look sane at a
glance.
Boost assumes variadic templates are always available in GCC 4.4+, but
they aren't since we don't build with -std=c++11.
This applies the patch that fixed the issue in boost 1.57:
eec8085549
See also: https://svn.boost.org/trac/boost/ticket/10500
tl;dr: This solves boost visibility problems for default/release build configs
on non-Linux platforms.
When Bitcoin builds against boost's header-only classes, it ends up with
objects containing symbols that the upstream boost libs also have. Since
Bitcoin builds by default with hidden symbol visibility, it can end up trying
to link against a copy of the same symbols with default visibility.
This is not a problem on Linux because 3rd party static libs are un-exported
by default (--exclude-libs,ALL), but that is not available for MinGW and OSX.
Those platforms (and maybe others?) end up confused about which version to use.
The OSX linker spews hundreds of: "ld: warning: direct access in <foo> to
global weak symbol guard variable for <bar> means the weak symbol cannot be
overridden at runtime. This was likely caused by different translation units
being compiled with different visibility settings."
MinGW's linker complains similarly.
Since the default symbol visibility for Bitcoin is hidden and releases are
built that way as well, build Boost with hidden visibility. Linux builds Boost
this way also, but only for the sake of continuity.
This means that the linker confusion logic is reversed, so the problem will
will now be encountered if Bitcoin is built with --disable-reduce-exports, but
that's better than the current situation.
fanquake