The per-host flag variables hold platform-specific defaults that are ignored
when flag environment variables are set, so it was wrong for them to contain
-std options relied on by package definitions.
Additionally, these flags (-pipe and -std=xx) will no longer be passed into
the CMake build, meaning less duplication in the build summary.
Pulled out of #31920.
When using CMake policies 3.14 and below, the `export(PACKAGE)` command
by default populates the user package registry, which is stored outside
the build tree. Setting the `CMAKE_EXPORT_NO_PACKAGE_REGISTRY` variable
disables this side effect.
In CMake 3.15 and later, this behavior is disabled by default, and the
variable has no effect.
With newly introduced libmultiprocess subtree, there's no need for depends
system to download and track changes to the upstream repository.
Note that adding the libmultiprocess subtree does not allow dropping
libmultiprocess packages from the depends build, because libmultiprocess
includes a code generation tool called mpgen, and in cross-compiled builds,
bitcoin core's cmake build system doesn't have access to a native toolchain and
can't build mpgen itself, so the depends system (or the native environment if
not using depends) needs to supply it.
Move parts of the int_get_build_id into a new int_get_build_properties
function. There is no change in behavior. This just organizes assignments
better so some build properties can be used to help compute build ids in the
next commit.
Building packages in depends implies using GNU Make. However, this
assumption can be wrong in environments where the `CMAKE_GENERATOR`
variable is set.
This change explicitly makes CMake use the "Unix Makefiles" generator.
Rather than potentially having to set this per-package, set it globally,
as this should always be what we want. Without doing this, changes in
later commits will have to add this per-package.
Similar to https://github.com/bitcoin/bitcoin/pull/29488, which is the
Autotools equivalent.
Rather than setting this per package, set it globally, as this is always
what we want. Without doing this, later commit will have to add the same
doc + change to more packages.
We currently do this sporadically. Not only amongst packages, but across
OS's, i.e sometimes it's done for BSDs/Android, and sometimes not.
Configure with `--with-pic` globally instead. I think this generally
makes more sense, and should not have any downsides.
See related discussion in
https://github.com/bitcoin/bitcoin/pull/28846#discussion_r1399123100.
This will allow us to use the existing machinery for filtering by
arch, os, debug/release, etc.
For example, the following becomes possible for libevent:
$(package)_config_opts_release=-DEVENT__DISABLE_DEBUG_MODE
Now the define is only set when not building depends with DEBUG=1
To successfully call the `capnp_generate_cpp()` function, the
`libmultiprocess` build system must be provided with paths to the native
`capnp` and `capnpc-c++` tools.
e644591426 build, refactor: Drop useless `call` Make function (Hennadii Stepanov)
Pull request description:
Using the [`call`](https://www.gnu.org/software/make/manual/html_node/Call-Function.html) function with `$(package)_*_cmds` is effectively noop because the latter, which could be found in `<package>.mk` files, do not use temporary `$(1)` variable at all.
This PR removes useless calls of the `call` function, and makes code more readable and easier to reason about.
No change in resulted dependency binaries could be easy verified with bitcoin/bitcoin/#21995.
ACKs for top commit:
laanwj:
Code review ACK e644591426
shaavan:
Code review ACK e644591426
Tree-SHA512: 8481fa0dc5bbf7dd6a180f7fae5a2ccc07f85b50c7a966bceb2d7e010e07e5f211ee3f74f8ac79bc5acfde5f0764264d599d959ff3ebb8511b1b4a33f79509bd
8494dcae0e Replace $(AT) with .SILENCE. (Dmitry Goncharov)
Pull request description:
This reduces the amount of syntax noise in the makefiles.
Setting V=1 still enables verbose logging.
The only noticeable difference in behavior is that, unless V=1 is specified, make won't print its own messages like
make: Nothing to be done for 'all', make: 'all' is up to date, or touch <file>, if -t is specified.
ACKs for top commit:
laanwj:
Tested ACK 8494dcae0e
Tree-SHA512: 66b9111229995aa54a9e87f4571648727d89b8529caec651063cdfe5c00a64341371b648701d192b2334df0614617a00c28eaa56c7f08ee9c00127cada0293ab
We now use a script named gen_id to generate the base build_id/host_id.
This solves 2 problems:
1. GNU Make special-casing exit code 127 (command not found) meant that
warnings about missing tools would propagate to the user's terminal
and broke our opportunistic build_id construction.
2. This change ensures that we don't have arbitrary characters in our
make variables that would be misinterpreted by Make.
See comments in depends/Makefile and depends/gen_id for more
information.
After reading
https://www.gnu.org/savannah-checkouts/gnu/autoconf/manual/autoconf-2.70/autoconf.html#Specifying-Target-Triplets,
my understanding is that this change should mostly be a no-op, as
--build defaults to the output of config.guess, however, this may be
slightly more correct
> For historical reasons, whenever you specify --host, be sure to
> specify --build too; this will be fixed in the future.
and will quell some warnings in depends (#16354). If anything, this
also explicitly enables cross-compilation mode when `--host` differs
from `--build`.
As for "fixed in the future", this is the case for Autoconf 2.70+.
Prior to this commit, when int_vars was called for packages, it would
immediately expand the "single-dollar variables", which may be defined
in terms of variables which are not yet determined (e.g. variables
defined in package/*.mk, which are included after int_vars is called).
This is required for the next commit as after that commit, for darwin
cross-builds:
0. int_vars is defined in terms of $(1)_cc
1. $(1)_cc is defined in terms of darwin_CC
2. ... which is defined in terms of clang_resource_dir
3. ... which is defined in terms of native_cctools_clang_version
4. which is undetermined at the time when int_vars is being expanded and evaluated
After #19685 started setting LDFLAGS, the INSTALL_RPATH_USE_LINK_PATH cmake
option used in the libmultiprocess build no longer works, so it is neccessary
to set CMAKE_INSTALL_RPATH as a fallback.
It's unclear currently whether the bad interaction between
INSTALL_RPATH_USE_LINK_PATH and LDFLAGS is a bug, but the issue is reported:
https://github.com/bitcoin/bitcoin/issues/19981#issuecomment-696680877https://discourse.cmake.org/t/install-rpath-use-link-path-not-working-when-cmake-exe-linker-flags-ldflags-is-set/1892
Commands useful for building / testing this change
make -C depends MULTIPROCESS=1 print-libmultiprocess_cmake
make -C depends MULTIPROCESS=1 print-native_libmultiprocess_cmake
make -C depends MULTIPROCESS=1 HOST=x86_64-apple-darwin16 print-libmultiprocess_cmake
rm -rvf depends/x86_64-pc-linux-gnu/native depends/work/staging depends/work/build
make -C depends MULTIPROCESS=1 V=1 native_libmultiprocess_staged
for f in `find -name mpgen`; do echo == $f ==; readelf -d $f | grep -i path; done
make -C depends MULTIPROCESS=1 V=1 native_libmultiprocess_built
find -name CMakeCache.txt
Fixes#19981
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.
Hennadii Stepanov
Cory Fields
Ryan Ofsky
fanquake
Max Edwards
Mark Friedenbach
laanwj
W. J. van der Laan
Dmitry Goncharov
Carl Dong
Russell Yanofsky
MarcoFalke