This is to nudge developers to use RBE especially when they run into
long builds and do NOT have RBE turned on.
Change-Id: I7da94763f5387310e522931bf650f9d7c01dc7f2
This is so that developers don't have to remember to source a separate
script to source RBE related variables.
I'll modify the docs once this change is submitted.
Test:
Ran forrest build with fix for previous build failure - https://android-build.googleplex.com/builds/abtd/run/L72300000950316038
Bug: b/193407413
Change-Id: I16f057086e2ee4f5918816ea70dbb33d0f5e99ab
This is so that developers don't have to remember to source a separate
script to source RBE related variables.
I'll modify the docs once this change is submitted.
Test:
USE_RBE=true source build/envsetup.sh sources RBE variables.
USE_RBE=false source build/envsetup.sh does NOT source RBE variables.
source build/envsetup.sh does NOT source RBE variables.
Bug: b/193407413
Change-Id: Ib8abf4c307b02fbb0c63b8a62d7aeda8be42d0fb
This CL extends the `m` function to short circuit to Bazel iff:
1) USE_BAZEL_ANALYSIS is set to 1 or true
2) There is only 1 Soong module requested to be built
3) The Soong module has a corresponding Bazel alias target in @soong_injection//targets/BUILD to the real target in out/soong/workspace.
Test: TH
Change-Id: I976ffa53106c0b52c75b71f1e0e1e8b3ef5cb3d1
This is less surprising than trying to use --config=bp2build as the only args, which is invalid.
Test: source build/envsetup.sh; b
Change-Id: I0b3724d05b8cc149268e660eb672192fe9bad19d
`b` is a function that combines:
1. Integrated generation of a synthetic bazel workspace, containing
symlinks to BUILD and bzl files alongside symlinks to the source tree.
This is the --package_path of the bazel build.
2. Running the Bazel build itself with b's entire argv.
A user accustomed to typing `bazel build <targets>` would
now type `b build <targets>`.
Test: source build/envsetup.sh; b build //bionic/...; b cquery --output=label_kind //bionic/...
Fixes: 188490434
Change-Id: I36e366108b024c09945d764a1115786658e03681
When metrics collection is enabled but
there's no gcert, print a warning and ask the user to run gcert so that
metrics collection can work.
Bug: b/184040091
Change-Id: Ib79c61e89a42dc7da9ef452b7101dac955859a7d
Test: m <tab><tab> before having done a full build.
Test: m <tab><tab> after having done a full build.
Change-Id: I35154151e6902d0337a26eab27d9d4fc797beed7
An old TARGET_PRODUCT value pointing to a removed product or an invalid
TARGET_BUILD_VARIANT value cause lunch to fail to read
COMMON_LUNCH_CHOICES, clear them before calling into the build with
get_build_var.
Test: TARGET_PRODUCT=foobar TARGET_BUILD_VARIANT=foobar lunch
Change-Id: I53d205784149ba7e32cc093f9d81d1a82c08bfc0
Necessary since the products used for APEX modules are different in
internal builds.
Test: banchan com.android.art
Test: banchan com.android.art arm
Test: banchan com.android.art art_module_arm
Test: banchan com.android.art module_arm
Test: banchan com.android.art x86_64
Test: banchan com.android.art art_module_x86_64
Bug: 179779520
Change-Id: Ib59a86b70e409537aaad7258465b6874b589b858
It currently uses TARGET_BUILD_APPS just like tapas, but the use case is
different and it may diverge more in the future.
Test: banchan com.android.art
Test: banchan help
Test: banchan
Test: hmm
Bug: 179779520
Change-Id: Iae718e65a2a7212c741c397e03c6f9a6d5ee8951
New ktgrep is added and sgrep/treegrep also scan *.kt files.
Test: kgrep/sgrep/treegrep for Kotlin files.
Signed-off-by: Taesu Lee <taesu82.lee@samsung.com>
Change-Id: Ie615aa86c4958310785392f18ee7bddaf64706a9
It will query module-info.json for modules inside a given directory.
Test: . build/envsetup.sh ; dirmods frameworks/base
Change-Id: Iad3dcb038315f49c156764e34fb5f52ec3ad29be
rgrep is a common alias for 'grep -r' and this new macro was confusing.
Test: cd external/rust/crates
rsgrep test
Change-Id: If71cdcc926455983a04a80982356e9feb61968fe
These are analogs to pathmod/gomod that find/install
the module's apk instead of cd'ing to its source
directory.
Bug: None
Change-Id: Iad750e154397f7cfcdd620f1ed6478b05c1225bb
Test: Manually, with bash and zsh
Some host tools (launch_cvd, etc.) have been using HOST_OUT.
As their dependencies are now installed by soong, the tools should refer
to SOONG_HOST_OUT. Export it as ANDROID_SOONG_HOST_OUT.
Bug: 172065015
Test: build aosp_cf_x86_phone, execute launch_cvd
Change-Id: Ie175049f15b89485feec0d33d82e756b12a7ad8a
Get a relative path to out by using $(get_build_var PRODUCT_OUT)
instead of hardcoding it to out/target/product/$TARGET_DEVICE.
That will correctly return the path to the out directory
when setting an external OUT_DIR_COMMON_BASE.
Change-Id: I8a990b710222bc72755c6b6b88fd0c9e80711e14
When sourcing the envsetup.sh script from a deep directory
inside of the repo, the vendorsetup.sh scripts are not being
sourced. This was causing the ANDROID_ENABLE_METRICS_UPLOAD
not to be set and developer's metrics were not being uploaded.
Provided the top directory to each directory path. Also cleaned
up the function to handle whitespacing correctly.
Bug: b/169699936
Test: *Ran source ../../build/envsetup.sh from external/libchrome dir
*Renamed repo dir to "aosp master" and source build/envsetup.sh
*Ran the script in zsh command line interpretor
*Ran the script in MacOS
Change-Id: I96eb73d50a1923cdda782792778b0d185a341cf9
oh-my-zsh common-aliases plugin defines `alias -g T='| tail'`, which causes an
error with the line `local T dir f`.
Making these lines more explicit resolves this issue, while also being clearer
that the intent is to clear out the local variables.
Bug: 169191000
Fix: 169191000
Test: source build/envsetup.sh; lunch <target>; m
Change-Id: Ief27aa3d9f541d0c39c6c11d7d771818d6493d31
This function:
- disables verity on the device
- reboots the device only if it is needed
- remounts partitions on the device
Bug: 167697453
Test: syswrite
Change-Id: I23c3e88daaa2b334233f4faf7fe8f8244cd56225
This is similar to the old `m showcommands`, but operates on specific
output paths instead and does not require a build aside from the
initial invocation of soong to generate the ninja files. The
`--regenerate` option is accepted that causes the ninja file to be
regenerated first, but this is not the default because it is slow.
Just note that without this flag results may be stale.
It unfortunately doesn't work for everything in out. For example,
`$OUT/system/lib/libc.so` won't work because it is a symlink that
isn't in the build graph, but `$OUT/system/lib/bootstrap/libc.so`
works.
Test: showcommands out/target/product/walleye/system/lib/libz.so
Bug: None
Change-Id: I4cfded3db994044870fc71ae4bcf9a69224b1e15
This reverts commit d92718356d.
Reason for revert: Adding the opening brace to the end of the function breaks lunch.
Change-Id: Ia3a9ff01964de0e616a534058d568150de32558a
This also affects pathmod here.
Since it's updating module-info.json, we could say that as well (this
should affect any tools which parse this file), but I haven't seen any
indication it's necessary in this context.
Bug: N/A
Test: N/A
Change-Id: Iecb6f454112b5c0095ee7063bc5442fdb1e3bd09
In some environments (e.g. with the master-art branch), the default
product (aosp_arm-eng) cannot be built, and thus prevents `lunch` from
printing the lunch menu (as it cannot initialize the build system in
order to fetch the common lunch choices). When this happens, show an
explicit error message from `lunch` instead of a Soong UI error.
Test: Check that this command line displays the expected error message:
unset TARGET_PRODUCT && . build/envsetup.sh && lunch
Bug: 152762648
Change-Id: I40c66bca5b075d88dbc8364b36b2db713e6ad93e
