Newer AOSP deprecates LOCAL_MODULE_TAGS := eng debug.
We now remove that and instead add the package to
PRODUCT_PACKAGES_ENG or PRODUCT_PACKAGES_DEBUG as applicable.
To ensure that we don't miss out on any package during this switch,
we add these explicitly, even though it might have been already
added under PRODUCT_PACKAGES at some other location.
Change-Id: Id1f94f34e100d616d91292efe514c201f7d24642
* Switch to binderized HAL services as possible and update
HIDL manifest accordingly.
* Following considerations regarding those without any change:
- Sensors left out of the game because wouldn't work
otherwise.
Change-Id: Id50291488d655187aa013c51bdd6890dca010564
Recent changes to adbd have enabled nonblocking_ffs by default.
These have seeminly uncovered some bad memory corruption in the
kernel ffs gadget driver on every kernel I've tried so far.
Unfortunately I've not been able to diagnose what is going wrong,
so for now, lets disable the nonblocking_ffs.
We still need to root cause the issue as its a problem even with
the mainline kernel on HiKey, so its likely a generic kernel issue.
Change-Id: I68a94f8f26b65ab383c37274e4f631227a3c7d5f
Signed-off-by: John Stultz <john.stultz@linaro.org>
This makes bionic and various art tools that compile/runs cpu-specific
code choose the kryo variant at runtime. It does not affect
build-time compilation.
Bug: 120773446
Test: bionic benchmarks with and without this change
Change-Id: I766f7bbd9f1fab35991c313da7bfad8563108ef8
Follow blueline which has not demonstrated any regressions
after switching to a generic arch variant.
Bug: 120773446
Test: make
Change-Id: If64385d968cb923177f74b03a94633f0dbf42fbb
Add zero clearing of allocated nmea buffer to ensure
the nmea string is null terminated.
Change-Id: I78e8b03bc7d0f9b69e1a51356da379a57a647cad
CRs-Fixed: 2041933
some msm8226 devices (like the grand 2 and the S5 Mini duos) supports full mdnie display capability, while other (like the s3 neo) not.
drop the livedisplay manifest entry to device specific tree so every device can enable the service that it suports
* Rationale: speed has a lot of dex code expansion, it uses more ram and space compared to quicken.
Using quicken for shared APKs on Go devices may save RAM.
Note that this is a trade-off: here we trade clean pages for dirty pages, extra cpu and battery.
That's because the quicken files will be jit-ed in all the processes that load of shared apk and the
code cache is not shared. Some notable apps that will be affected by this are gms and chrome.
As seen here : https://android.googlesource.com/platform/build/+/refs/tags/android-9.0.0_r48/target/product/go_defaults_common.mk#45
* The permissions were not set to 222 instead of 220,
thus introducing regular logs from libcutils:
"cutils-trace: Error opening trace file: Permission denied (13)"
* Refers to frameworks/native/cmds/atrace/atrace.rc
Change-Id: I236ea25d2f2068f9f86a43492807f5f05e92da96
This file is used by hciattach when turning on bluetooth,
and without correct permissions, MAC address will be wrong.
Change-Id: Ie700a4be1a6b83932a9cffac45cd63f179f3edda
