NOT_FOR_ARCHS only works if _REASON can be applied to all affected
architectures. If any other architecture doesn't share it one'd have to
add BROKEN_${ARCH}, anyway.
$ make -V '${MAKE_ARGS:M--toolset*}'
/bin/sh: clang39: not found
make: "/usr/ports/Mk/Uses/compiler.mk" line 69: warning: "clang39 --version" returned non-zero status
make: "/usr/ports/Mk/Uses/compiler.mk" line 120: warning: "clang++39 -### /dev/null 2>&1" returned non-zero status
--toolset=gcc
PR: 213867
MFH: 2016Q4 (r430247 followup)
- aarch64 requires Clang with new enough integrated assembler
- sparc64 is no longer supported since we don't split context library
into a separate package (unlike Debian)
PR: 213867 214942
MFH: 2016Q4
$ make WITH=ASIO -C www/nghttp2
[...]
In file included from asio_common.cc:25:
In file included from ./asio_common.h:32:
In file included from ../src/includes/nghttp2/asio_http2.h:37:
In file included from /usr/local/include/boost/asio/ssl.hpp:19:
In file included from /usr/local/include/boost/asio/ssl/context.hpp:786:
/usr/local/include/boost/asio/ssl/impl/context.ipp:232:27: error: no member named 'SSL_CTX_get_default_passwd_cb_userdata' in the global namespace; did you mean 'SSL_CTX_set_default_passwd_cb_userdata'?
void* cb_userdata = ::SSL_CTX_get_default_passwd_cb_userdata(handle_);
~~^
/usr/local/include/openssl/ssl.h:1640:6: note: 'SSL_CTX_set_default_passwd_cb_userdata' declared here
void SSL_CTX_set_default_passwd_cb_userdata(SSL_CTX *ctx, void *u);
^
PR: 214793
Submitted by: matthew@reztek.cz
Obtained from: upstream (via Gentoo)
Approved by: portmgr blanket
- Clean up the Makefile.
- Follow some upstream recommendations (--with-data-packaging=archive,
--disable-renaming, -DICU_NO_USER_DATA_OVERRIDE).
- Patch makefiles to install static libraries with INSTALL_DATA so they
aren't stripped.
- Patch config/mh-bsd-gcc to sync with config/mh-linux-gcc.
- Fix endianness detection in ICU. The code wanted to use BYTE_ORDER
defined in machine/endian.h, but this isn't visible because ICU is
compiled with _XOPEN_SOURCE. Patch the code to use _BYTE_ORDER instead.
- Compile ICU with C++11 compiler to enable move constructors.
- Patch ICU to fix a problem with atomics in the case of a C++11 compiler
without C++11 header <atomic> (like Clang on FreeBSD 9).
- Bump all ports that depend on it due to library version change.
- Add USES=compiler:c++0x to some ports that pick up -std=c++0x from ICU
pkgconfig files.
- Add USES=compiler:c++11-lib to graphics/libcdr01 because it also needs
a C++11 runtime library now. Add this to all ports that depend on it
so their executables load the right libstdc++.so on FreeBSD 9.
PR: 205120
Exp-run by: antoine
Approved by: portmgr (antoine)
it enabled by default to avoid a regression in the science/orthanc unit
tests. Set USES=iconv whenever ICONV is enabled, rather than only when
ICU is disabled. With the default option settings, this will bring in
converters/iconv on FreeBSD 9, which will unbreak the failing unit tests
in science/orthanc. At least one of the ICONV or ICU options must be set.
Fix PR 198658 in the default case by setting
BJAM_ARGS+=-sICONV_PATH=${ICONV_PREFIX} whenever iconv support is enabled,
not only when ICU is disabled as was done in r398027 (before that was broken
by r398060 and fixed again in the ports framework by r398258).
Pet portlint.
PR: 198658
USES must be set before including bsd.port.pre.mk, which is included
by boost-all/compiled.mk. The easiest fix for the conditional
USES+=iconv being too late seems to be to use an option helper.
For consistency use an option helper for the LIB_DEPENDS.
PR: 203285
building with recent clang. Four of these were included in the 1.56
release, the fourth was included in 1.59. One of these needed the
addition of "|| defined(__clang__)" because the __GNUC_MINOR__ check
excluded clang.
when compiling code that uses the BOOST_STATIC_ASSERT() macro with
recent versions of clang. Upstream has a slightly different fix in
their develop branch (removel of the __GNUC_MINOR__ check, which works
because clang claims to be __GNUC__ == 4), but it has not been merged
into the master branch.
This change reduces the size of the OpenOffice build log by 66%.
when the files contain pre-processor directives.
The attached patch passes these files through the compiler as is done in the
base system.
PR: 193836
Submitted by: andrew
Approved by: bapt (mentor)
ccache's documentation says to set CCACHE_SLOPPINESS=time_macros and
CFLAGS+= -fpch-preprocess (it's a magic check in ccache for the string),
but doing these does not actually fix the build when files are cached.
Disabling PCH for ccache still results in an 18 second build, down from 143,
on a 12 core machine.
- While here, strip libraries with STRIP_CMD (not bumping rev as this is
not significant enough to warrant a rebuild).
compilation with clang 3.4:
Commit: 6bb71fdd8f7cc346d90fb14beb38b7297fc1ffd9
Fixed incorrect initialization of 128-bit values, when no native support for
128-bit integers is available.
Commit: e4bde20f2eec0a51be14533871d2123bd2ab9cf3
More compilation fixes for the case when 128-bit integers are not supported.
PR: 187958
Approved by: bapt (mentor)
Sponsored by: Bally Wulff Games & Entertainment GmbH
switch back to normal library naming
Add a compat symlink for .so.5
Use compiler.mk to detect compiler
Use a modern compiler (doesn't work with base gcc)
- pass maintainership to office@
- detect if compiler is gcc or clang and correctly set the toolset for bjam
- enable long-double support
- some cleanup
Approved by: maintainer timeout (3 weeks with no reply)
