* Some variants have autogenerated fingerprints that are >= 92 chars
on official nightlies, and overriding them in init is failing
* To workaround this, force a build fingerprint that's shorter
Change-Id: If3909ad6f6f00b77559aff16feb6291b42ddbc47
* P made changes to pthread_mutex_init() that are incompatible
with our libsec-ril libraries. In order for rild to even see a
radio as being availble we need the legacy behavior.
* Even though our libsec-ril libraries are from 5.0, which is API
level 21, we do not appear to require any of the other behaviors
from API levels earlier than 27.
Change-Id: I3b3d0160236df09bdae95c5d9f10a61bf91ebc72
* This singularly handles both text relocation (M) and mutex (P)
linker restrictions
* mediaserver - Needs text relocations
* mm-qcamera-daemon - Needs both text relocations and mutex restrictions
relaxed
Change-Id: I97a79fb57aaadf5cf62894489522b2de6d8c2e36
* We have included the appropriate android_atomic methods in libril,
so the symbols are loaded into the process space and the libsec-ril*
dependencies for all builds are handled.
Change-Id: I93289af789df7e263488e1db29bdbde0d0964e97
* Recent binder changes caused our approach of shimming libcutils to
re-add the non-inlined android_atomic* methods to start getting
slapped, at least with NDK apps. This looks oddly similar to the
heads-up we received in 7.0, so no clue why it let us do it in 8.*
prior to now.
* See: https://developer.android.com/about/versions/nougat/android-7.0-changes#ndk
* We did this because all of the libsec-ril* blobs and libuiblur
needed it and doing each and every libsec-ril seemed fragile and
fraught with the potential for error, but "fragile and fraught with
the potential for error" is better than "phone breaking". And blur
doesn't seem to ever be coming back.
* We'll leave the "standard" libsec-ril.so here and let the
individual devices handle anything specific if they have variant-
specific blobs
Change-Id: Ib11048e4924f34ade20f44b707f0106e139f2f82
This reverts commit aee12a34df.
* No Samsung msm8974 device is going to develop an OSS camera HAL
at this point. Move this common stuff to msm8974-common.
Change-Id: Ib0bb18fe819e2ebbb39c9b278ccf687444f65488
* Mounting /system partition is handled in kernel now,
however removing the entry from fstab caused issues building
the OTA. The workaround was to have a separate fstab, but turns out
that simply setting the recoveryonly flag does the trick because
those are then ignored during a normal Android boot.
Change-Id: I2944384d0a1c41bc9f9f51e2e29daff2bed0a0f4
* In particular, the RIL on CDMA variants seems to only work reliably
on first boot after flash, when things are all slowed down while
dexopting. After that first boot, it's hit and miss whether a
particular boot will have function RIL
* Slow things down by limiting ourselves to a single CPU on boot,
bringing the rest online when boot has completed
Change-Id: Ie194740cf0487268dc0dbd3377bbb790cdd1b04d
This reverts commit 6daa361afe.
* We're out of cmdline space and we don't even make use of the symlink
that this used to automatically create.
* If we ever want to use the bootdevice symlink, just create it
explicitly (without prop) in init.qcom.rc
Change-Id: I1c45c3ed3605f4020d824be2775639c04ba8b60d
* Replace the LOCAL_PATH instances with a variable not prone to
getting stomped
* Replace all expanded values with the variable
Change-Id: I91ef536c038fb38d250633bfdad82c2a6537c1da
* All of the variants are close enough to each other that
carrying these in the device-specific trees doesn't
add much.
* Sizes picked by selecting the smallest value from each
of the device-specific trees for each variable.
Change-Id: Ibe024ec857db971ff1522d05cf55e8492816b7ec
* This removes /system entry from fstab, since this partition
is now early-mounted and the fstab entry is specified in device tree.
* At the same time, also include a fully populated fstab, in order to avoid
build breakage while generating recovery updater scripts.
Change-Id: I8e558458b984e97a45a920c3751a6b7c3f2ebc5c
* cmhw is dead, but we don't actually carry any local classes. Just
stop appending to the variable altogether
Change-Id: I075ad79ed99b0b85a21af594db0036136c984542
AOSP and mainline are going towards removing old 32-bit binder API support.
64-bit binder is 100% compatible with 32-bit kernel and userspace,
so there is no reason for us to keep using the old solution anymore.
We've switched to new API in our kernel, same thing should be applied to our userspace.
Change-Id: I3c95bd7fbd023c5cb08e856b3a3889c03228e843
