* commit '054b0274fc27babb42a60002ad25c9e608eac257':
add support for more LOCAL_*_arch variables
don't rename 32-bit executables to *_32
remove 2nd arch from ARCH_ARM_* defines
Depending on the file extension of the generated C++ file,
bison will generate a #include of a .h or .hpp. So both files
must be kept in the generated directory.
Change-Id: Id0aac7f407bdc69c7f5012c0d021761b0fceb427
Analyzer needed by WITH_STATIC_ANALYZER and WITH_SYNTAX_CHECK is
moved from prebuilts/clang/{linux-x86,darwin-x86}/host/3.3 to
prebuilts/misc/{linux-x86,darwin-x86}/analyzer
See https://android-review.googlesource.com/#/c/83852/
BUG=13243591
Usage:
"WITH_SYNTAX_CHECK=1 make ..." instructs build system to invoke "clang -fsyntax-only"
to utilize clang's better diagnostics before calling LOCAL_CC/LOCAL_CXX for code generation.
The compilation time is slightly longer, and the generated object file should be the same as
w/o WITH_SYNTAX_CHECK
"WITH_STATIC_ANALYZER=1 m/mm/mmm/mma/mmma ..." instructs build system to run static
analyzer via "clang --analyze" on a successful build. If analyzer finds any issue, instruction
to open report is displayed. See http://clang-analyzer.llvm.org/scan-build.html for details.
WITH_STATIC_ANALYZER trumps WITH_SYNTAX_CHECK if both exist. Project use lots of GCC extensions
(eg. nested function) not supported by clang may opt out by adding LOCAL_NO_STATIC_ANALYZER:=true
Change-Id: Ib3dda3ffb0fd3aaf2eadec867a966d1dd2868fb1
- external/skia now builds on arm64, frameworks/base/libs/hwui
depends on it.
- external/bluetooth/bluedroid builds on 64 bit, though there
isn't any obvious way to test it without real hardware.
- frameworks/ml now builds, as does external/srec
- frameworks/opt, frameworks/ex and frameworks/wilhelm don't
build because of their dependency on frameworks/av.
frameworks/av should probably be dropped out of the blacklist
and replaced by individual markers on targets that we want
for 64 bit (we don't want all of them for arm64).
Change-Id: I9735108840fcba21ac8918bedf0d6de8989de086
This is until aosp gets the latest libpcap+tcpdump currently only on internal.
This reverts commit 5c68e265ac.
Change-Id: If6a990c72cb3a8d699cf0881f7d5fb8b725fdf2e
Include apps that appear under /system/priv-app
to also be included with the zip of all the apps.
Change-Id: If4687cf4c471877f11d78b68bad96f1842e49d87
Signed-off-by: rpcraig <rpcraig@tycho.ncsc.mil>
Also we don't need to include module_arch_supported.mk again, if we are
currently substituting the source build with LOCAL_PREBUILT_MODULE_FILE.
Change-Id: I444b0397d74c3153b398a050b762e49418062a86
We'll probably remove external/oprofile soon, but this gets us closer
to being able to turn on an x86_64 continuous build in the meantime.
Change-Id: Ic1d5331d41dafee9ffed222dc332afed2d4ae356
So a library can export the proto's include path that can be used with
both archs in multilib build.
Change-Id: Ia0f92f0b40e39dc3fa426c69c52139a0a8f04077
For host executables and shared libraries, the global LDFLAGS were being
inserted into the linker command line after the module-specific ones,
making it impossible to override the default settings. Change the order
to match target linker invocations.
Change-Id: Icd5f6f83df9f27a5be97ddb197ee245c1ab8c2be
This makes sure copy_headers.mk only be included onces, no matter
it's for the 1st arch or the 2nd arch.
Change-Id: I80a558fbdb52861f176bd27a21c302069a5cc3ce
It seems the code was meant to add the toolchain to ANDROID_BUILD_PATHS
and then export to PATH by lunch.
But now we export the toolchain to PATH explicitly in envsetup.sh and
the code is unnecessary.
This reduces the places to change when we upgrade the gcc version.
It also fixed the bug that the mips toolchain was always exported
regardless of the target arch.
Change-Id: Iee3b895b4c4e0df8971c27c01b9ecbd591848b12
Fix several wrongly configured tests that were either
looking for tests in the wrong jar (apache-harmony-tests
instead of core-tests) or using the wrong prefix.
Also, this change creates subsets of the harmony tests based
on subpackage names (java.net, java.io, java.nio etc.) instead
of the earlier harmony groupings.
Change-Id: Iea0e20d23512611d1aac92b2f8219031b6396c77
Prebuilts often support only a single architecture, allow them to
use LOCAL_MODULE_TARGET_ARCH to specify it.
Change-Id: I2514f66f682ef267bbf1a1ab78510faff0a18b64
The LOCAL_*_$(TARGET_ARCH) variables don't make sense for host
modules, only append use them for target modules.
Also complete the list of LOCAL_*_arch and LOCAL_*_32/64 to be
consistent.
Change-Id: I00c83e5c4e08ed9a844f9f99a79ce4bcc3f0bf11
combo/TARGET_x86*.mk mistakenly added TARGET_GLOBAL_CFLAGS to their
linker command lines. This results in clang builds not working properly,
since they strip some unknown flags from TARGET_GLOBAL_CFLAGS.
Change-Id: I60a1ff5df70305323134435e4ae107ea7acfe8ea
Add four new variables for module makefiles:
LOCAL_MODULE_TARGET_ARCH specifies that a module is only supported for
one or more architectures. Any architecture not in the list will be
not attempt to build the module. The expected use case is prebuilts
that are only suitable for a single architecture, or modules like llvm
that need per-architecture support.
LOCAL_MODULE_UNSUPPORTED_TARGET_ARCH specifies that a module cannot be
built for one or more architectures.
LOCAL_MODULE_TARGET_ARCH_WARN and LOCAL_MODULE_UNSUPPORTED_TARGET_ARCH_WARN
are the same, but warn that the arch is not supported, which is useful
for modules that are critical but not yet working.
The logic for whether or not to build an architecture is fairly
complicated, so this patch consolidates it into module_arch_supported.mk
Change-Id: I120caf4a375f484e1fd6017b60c2f53882ae01e6
-- Added TARGET_PREFER_32_BIT, which sets LOCAL_32_BIT_ONLY for an
executable, if LOCAL_NO_2ND_ARCH is not true.
Name resolving in 64-bit multilib build:
-- Name resolving in PRODUCT_PACKAGES:
foo:32 resolves to foo_32;
foo:64 resolves to foo;
foo resolves to both foo and foo_32 (if foo_32 is defined).
-- Name resolving for LOCAL_REQUIRED_MODULES:
If a module is built for 2nd arch, its required module resolves to
32-bit variant, if it exits;
Otherwise for executable and shared library, a required module
resolves to the default 64-bit variant; for other module classes,
required module foo resolves to both foo and foo_32 (if foo_32 is
defined)
Bug: 12898862
Change-Id: I5fda1a77f58814097b10b5ad2743ee25adfaecc4
1. Following the setup of gcc in build/core/combo/,
we added the [HOST|TARGET]_<arch>.mk clang config files,
and load only the configs needed by the current product.
2. Added support for the 2nd arch.
Change-Id: I2a383418a9688a050b39492f8e489d40eeeb5f2d
I don't think we can realistically turn this on for 32-bit builds any
time soon.
Also, fix the arm64 stack-protector hack.
Change-Id: Ie1e7c875bbc06fb21bb372b8ca99879a23ef53d4
2ND_TARGET_DEPENDENCIES_ON_SHARED_LIBRARIES was not set,
which was causing the later += to act like = instead of
:=, and the dependencies would disappear as soon as
LOCAL_MODULE was cleared.
Change-Id: Idea291524fc06377deafec62f37d20eaa7f93bca
Users of ARCH_ARM_* defines don't care about first vs. second arch,
set ARCH_ARM_* regardless of which arch is arm.
Change-Id: I2ae83ec5c3f839ff91a0e352c95d76ec2cbd5dc5
