The recovery image already contains a copy of first stage init, so we
can boot unconditionally to the recovery image and instruct first
stage init whether or not to boot to Android or to recovery. In this
case, we need neither the kernel to mount /system as / nor a separate
partition for recovery, so this change modifies the build scripts to
allow this combination.
Bug: 114062208
Test: Boot pixel from recovery to Android with BOARD_USES_RECOVERY_AS_BOOT
But without BOARD_BUILD_SYSTEM_ROOT_IMAGE
Change-Id: Icd047afb7f22d2724b3bcaca1aa0c837426dcce7
This has been long dead since commit
17c83cf22c (2009).
Test: m checkbuild (via TreeHugger)
Test: No hit in code search.
Change-Id: I054241ae63436ade2f4e50ec2fc3757ae35635dd
d8 requires the full classpath if --min-api requires default method
or try-with-resources desugaring. Always pass the full classpath,
since it will already have been built to generate the input jar.
Bug: 116243313
Test: m checkbuild
Change-Id: I363b4d3145e23d4eed77455c1897430206f3ac22
For ASAN build the intermediate dir might differ from the default
directory.
Test: m (on an affected ASAN build)
Bug: 116020783
Change-Id: Ie4f282ea651d710351799f3fc24dba42e79ee63d
Adding -pie to LOCAL_LDFLAGS for host linking leaves it around
when doing the host cross Windows linking, resulting in a warning:
clang-7: warning: argument unused during compilation: '-pie' [-Wunused-command-line-argument]
Add it later in binary.mk after the flags have been copied to the
single-use my_ldflags.
Bug: 115776263
Test: m checkbuild
Change-Id: I70d71fa1ec6445631dc7e89ef61927a3304375e2
For example, if BOARD_SUPER_PARTITION_SIZE is
10GiB, then the sum of sizes of all dynamic partitions must
be within 5GiB.
Fixes: 115920069
Test: make
Change-Id: I800faff28e8483c92a1c0cf5e0b3bae4d638f86d
When the system partition is a dynamic partition, the device will be
using an initial ramdisk and should not have a root device specified on
the kernel command-line.
Bug: 115939310
Test: device boots when BOARD_SUPER_PARTITION_PARTITION_LIST contains
super and AVB is enabled.
Change-Id: I5c4b2304bf1ab6a59443c43c354845bf8a3949f0
Missing that doesn't seem to affect correctness, but it upsets the
syntax parsing in my editor.
Test: `m -j out/target/product/marlin/recovery/root/prop.default` with
aosp_marlin-userdebug.
Change-Id: I272f6a6e4fe67b3ef316aa38896a0df0b42228d0
dex_preopt_odex_install.mk should always be included, it handles
setting a default value for LOCAL_DEX_PREOPT.
Use dexpreopt-copy-jar when preopting to strip the dex file if
necessary.
Bug: 115967961
Test: m SystemUI
Change-Id: I74f77ed0e88a67043836407261f3845a16cbaf37
$(HOST_OUT)/obj/PACKAGING is cleaned via `installclean`, so regular
built modules should not be depending on it. Explicitly call out the
dependencies instead.
Test: m; m installclean; m
Change-Id: I0f318257e4742d6b7bd099dc06958ca8aba0cfb8
Also remove the msg.mk files used for old apicheck.
And export jdiff-doc.zip generated by Soong.
Test: api-stubs-docs-diff
Bug: b/78245848
Change-Id: I329aa375623a7b7a121d58314dab694a326c9f0f
soong_zip produces more reproducible zip files that are not
affected by umask values or timestamps.
Bug: 69500920
Test: m checkbuild
Test: contents of zips is unchanged
Change-Id: I2eca2f7618e06c0b5893ff54c79261921fe051d6
zip maintains permissions when adding files to zips, which causes
the final jars and apks to change based on the umask of the build.
Move some of the rules that were using zip to soong_zip instead.
Since soong_zip doesn't support adding files to existing zips,
create a new zip file for each part of the package creation
process and then merge them together at the end with merge_zips.
Bug: 69500920
Test: m checkbuild
Change-Id: Id253df776ce19ec4cac3a36fa470a50461ea3cad
In preparation for rearranging the way jars and apks are built.
Relands Ib9f0da132196942062c58fd5b12e84588106724e with a fix to
correctly include classes.dex in hostdex jars.
Test: m checkbuild
Change-Id: I397544456b8ce5827716fcd1828de707b719c2bb
The change https://r8-review.googlesource.com/c/r8/+/26482 made R8
process data resources by default. This is not expected for the
Android Platform build
Test: m -j PRODUCT-aosp_x86-userdebug core-oj
Change-Id: I8936a39ea3cc0f46e455087076635a8fa2c9e0d3
If something needs to distinguish between _OVERRIDE and non-_OVERRIDE,
it is a hack (these should be the same property).
Bug: 114488870
Test: m nothing (exercises this)
Test: searching all of these reqs in the code.
Change-Id: Idc22150a3179d025d7d95beafad39e8b07946ef8
This makes the product-installed-files macro more accurately reflect
the files installed for a given product, as well as fixing bugs in
the previous implementation.
Specifically, the complete list of overrides found so far is stripped
in each round of expanding required modules. Previously, overrides were
stripped out *after* expanding required modules. This meant that for a
scenario where B depends on C, and A overrides B, C could get installed.
It's unclear if this was a problem in practice.
The other effect is that the offending artifacts txt is more accurate,
since overridden modules are now correctly removed.
Bug: 80410283
Test: build_test downstream
Change-Id: I8bfc7c40bedd5cb2afba567bae4b998f51770793
This makes the product-installed-files macro more accurate, as it will
now include the eng/debug/test/asan packages for each product involved
in the analysis.
It also fixes a bug, where packages in e.g. PRODUCT_PACKAGES_DEBUG
would not get their REQUIRED modules installed prior to this change
(I think).
Bug: 80410283
Test: diff modules_to_install on an internal product
Change-Id: I8e148454df2f96523bbed25e2695d050559968f1
Treehugger Robot