While this doesn't solve the long-term issue of out-of-date docs,
I hope it can smoothen the initial set up.

TEST: Manually went through the steps on QEMU to make sure it works.
Change-Id: Ia1ab7168c05af48e9316e3acb971537476103434

Use a FIDL service to transport the root resource between bootsvc and
the next process in the boot sequence, which is currently devmgr.

The FIDL service is currently provided under /bootsvc in order to not
collide with /svc. This will change in the future once componentmgr is
introduced into the boot sequence.

ZX-3385

Test: Added to bootsvc integration test, and ran DDK tests.
Change-Id: Id6f8219c971a4dc9958040a1a56a05d718e83eac

Modifying fvm.blk in-place was a mistake, and we must extend it to make space
for ephemerally cached software. As such we no longer really have a choice
but to make a copy before we extend. Wrapping in qcow2 is no longer an
advantage, so that's removed.

Change-Id: I8aae0e15c7b78c6715fa98da3ce970d1ba4fb808

The upcoming automatic variable initialization (pattern mode) CL catches
an uninitialized bool in CodecClient test code.

Tested: use_h264_decoder_test with I81a22db76c925b14731def42abd49168aa0265e4 patched in
Change-Id: Ia213de664aebaacece43ee401b0ecb665f58e6e2

This change adds DEF_UNSHIFTED_FIELD, which is a field you read/write to
as if the other fields in the register were masked off, allowing users
to read/write values without shifting.

Change-Id: I56c172d11d43c23c9d8816628b40d33509749fc4

Apparently if we do not do this explicitly, `Send` is not automatically
inferred and that makes it impossible to construct a pseudo directory
tree and then send it to another thread for execution.

Seems like this has a drawback in now requiring all the callbacks to be
implemented using only `Send` types, rulling out, for example,
`RefCell`s.  I am not sure if this would be a considerable inconvenience
for usage in actual applications. It would be great if we could make
this generic over `Send`, but I do not understand yet how to do it.

Change-Id: I4543c33f4ba12011f9a91847736fdb21205f134a

This parses function calls in expressions like "print Foo()" or
"print reinterpret_cast<Foo*>(bar)" (where the function is
reinterpret_cast in this case).

Currently these are not hooked up to anything and will issue an error if
one is ever executed. This will be used to implement reinterpret_cast
and static_cast in the future.

Fix a problem in the template parameter list parser that allows trailing
commas. Test added for this case.

Change-Id: I24299fbd3e8242b778c6dfffe8cbcc1d9e66b9f9

This CL uses a forward declaration for fxl::Thread to minimize includes in
Dockyard.h.

Test: none
Change-Id: Ib81d768ed4695ce7c224f6a361869ae82fdde803

This CL is the first in a series that converts the Bluetooth APIs
to use integers to identify remote devices.

* Introduced a new opaque identifier template to lib/common which
can be specialized with any trivial data type.

* Introduced common::DeviceId as a distinct subclass of the new
Identifier type. ID types that don't represent Bluetooth devices are
expected to be declared as distinct types, preventing ID type conversion
through type safety.

* Migrated drivers/bluetooth/lib to the new types instead of std::string
  for RemoteDevice and advertising identifiers.

* The FIDL APIs still express identifiers as strings. The FIDL servers
  currently convert between strings and identifier types when passing IDs
  between FIDL clients and btlib objects. This is temporary until all FIDL
  interfaces have been updated to use integer IDs.

Bug: BT-305
Test: 1. Run all tests in the bluetooth-tests package.
      2. Tested device discovery and connections using bt-cli.
      3. Tested LE peripheral role using bt-peripheral.
      4. Tested LE central role, pairing, and GATT client using
         bt-le-central.
      5. Tested GATT server using bt-le-heart-rate-peripheral.
Change-Id: Iee42a9cc08d555c41ea024ae20bae21fdc3a50cf

Add operator<< for logging requests and responses. Helpful for debugging.

Test: none
Change-Id: I3405fa4bdfe45aeb06139eac75b68d033e4d7a13

Test: CQ
SCN-1291 #done

Change-Id: Id8d8c2921855fe60973cfabbfa46bf7a30339aea

Added details on how "-v" is interpreted as it's interaction with the
Rust test runner is different from what was documented.

Change-Id: Ib82802d7edca992cd0b37c284de43a0403739ee2

Zircon saves and restores extended register state on context switching
but not when moving between userspace and the kernel. Therefore, the
hypervisor needs to save and restore on exit and resume to avoid
collisions with the VMM userspace.

Additionally, Zircon does not save all possible extended register state
so for the sake of completeness and safety the hypervisor should save
what the guest is configured to use by setting XCR0 before save/restore.
The value we used is based on the guest's XCR0 and CR4 in the following
way:

x87 - From guest XCR0 (must always be set to 1.)
SSE - Can be enable independently of XCR0, so check CR4.OSFXSR.
AVX - From guest XCR0 (XSAVE-enabled.)
MPX - From guest XCR0 (XSAVE-enabled.)
AVX-512 - From guest XCR0 (XSAVE-enabled.)
PKRU - Can be enable independently of XCR0, so check CR4.PKE.

There are also two pieces of state (PKRU and HDC) that saved/restored
based on IA32_XSS_MSR but the hypervisor does not support that MSR
nor the MSRs that are saved and restored so we never handle those
states.

TEST=Added test to hypervisor-test

Change-Id: I99d4591f4f29408e07d189671d76e9cb2d55b7cf

It's impolite for a library of this sort to spill to stderr, and we
don't have a good alternative on the host side, so instead we'll simply
collect our warnings on the ElfLib object itself and allow interested
users to grab them when they want them.

Change-Id: I35db8177d39ca309bf67fcaff355dbb49697c57b

This is the first version of MediaTek DSI driver. It uses the new
DSIIMPL protocol.

This CL makes some modifications to the DSIIMPL protocol. It also add
support for the LCD panel used on the mediatek reference board.

Test: Tested on Astro, Sherlock and mt8167s_ref since they all use the
DSIIMPL protocol

Change-Id: I5b2ddef8509340d00e06303882013c1dc046989d

Current compiler default is for new alignment to be 16 bytes, this mismatches our
heap, which has 8 byte alignment.

This explicitly sets `new` allocations to 8 byte default alignment, and adds a
static_assert to prevent divergence in the future.

ZX-3568 #done

Test: Build and run kernel tests on gcc+clang

Change-Id: I7f265dca8b9fef90dc1a0e128480569aba0ff8a3

Vec is Deref<[T]>, so that's not needed here.

Change-Id: If432e2bafc45b70f1e5d4b90af6218f249c37240

Use --board vim2 instead

Change-Id: I23fa92495851fb2d41ca17009c639cb87ed0a25f

Adding another function to init_sequences.cc which is run
when the ISP starts up.  These are custom configurations for
sherlock.

Test = CQ

Change-Id: I63b382720b9572c242a199656ab04db7f6d67819

Minor improvements to docs (formatting and clarifications).

Test: none
Change-Id: Iff2461bd37de1636d82fd8e4e19cb014f4eecfee

Improved the verbiage onhow handles are treated during a channel
read.

Test: Inspection.
Change-Id: I61e157909a63a91d4069326a0f2d7d72be3b7663

This change adds the banjo and necessary plumbing for ZX_PROTOCOL_POWER_IMPL.
The power impl protocol is added to the platform bus and will behave
same as GPIO_IMPL protocol for now and will be
moved out later when the composite devices are implemented.

Test: fx full-build with a bare minimum powerimpl driver using this protocol.
    https://fuchsia-review.googlesource.com/c/fuchsia/+/260149/

Change-Id: Ibf2f1fbfee7aecae185c778e8612ed76a49c3bea

TEST=guest_integration_tests

Change-Id: I7122e07dc197b08130a627042396e2dbf703ae11

Change-Id: I2d9ff414f8e6d4719a9aaba4b885120a1c81a39a

This CL walks through the job tree (though it doesn't do useful things with the
data yet, that will come later).

Test: None
Change-Id: I50c52231516f0a208d80abab9eeb871d97b35a4f

Change-Id: I7f756d9209a7bcd9dc52fb3d70ccbd685444d9af

mediaplayer_test_util had the deprecated persistent storage feature
(access to /data) turned on in its .cmx file. Given that this
feature is deprecated and that /tmp serves just as well for most
use cases, this CL removes that feature.

TEST: ran mediaplayer_test_util to check for regressions

Change-Id: I3c84fcf59c2904c3b9342d2439048b99e6697da3

Change-Id: Ib3ce0b3999aee80e862376cd7bc789b7f320e7e7

Test: fx run-test wlan-mlme-tests
Change-Id: I3d63ea90c8218ecbbd51e43b8c9087356eb16854

Update instructions for the kernel command line arguments file so that
the file lives in $FUCHSIA_DIR/local, where it is ignored by git.

Test: Ran and built doc instructions on Mac
Change-Id: Iba2678537853c6b00d184cfe3c30e3d905edec1e

Support nouns for the "get" command, so one can type "process 2 get" or
just "process get" to get just the settings for the process.

Allows searching for a specific setting on a specific noun via "job get
filters".

Outputs the noun index for the setting summary, so "get" will print
"Process 1" and "Thread 5" (for example) so it's clear what the returned
setting context is.

Generalizes the retrieval of settings to avoid code that manually
checks different object types. Most of this is stored in an arrays with
one helper to get the right setting schema and object for a given type.
This should make the settings more extensible in the future.

Adds a "global" noun to the command syntax. This allows specification of
the global settings store via "global get". This noun will also be
needed in other contexts.

Removes printing of the settings schema when no object exists.
Previously "get" would fall back to printing the schema. when e.g. no
thread exists. But this gets either weird or complicated when you have
to support "thread get foo" on the schema. Since we don't have any
settings in this category, it's not possible to test and it's difficult
to tell how important this will be. We can add it back later if we run
into a need for it.

Change-Id: I3a40975946b8fc483edaca7335fbe80d13877c75

Ordinals are no longer. Long live hashing.

Change-Id: I1ef296e4fda4d18614b0b0fb6f6d17fc9d6aed1b

We refactored some of the functionality related to kernel debug
functionality from dmctl into a standalone service. However this service
only ended up being visible to the serial console and not the virtcons
(ssh/local input).

Test: tested local input on vim2 and ssh on qemu.
Change-Id: I132f681eb9f0c1438809e2e7a6cdd46582e009e0

Change-Id: I6f9346fdd9c9dd444a9bd31dc9796edd1b3a4784

Change-Id: I6143768fa8885a5ed844d84c83171a20ce16a173

This CL improves the comments for the fuchsia.media FIDL files
stream.fidl and stream_type.fidl.

TEST: no behavior change
Change-Id: I7b50bc018e4ac2270a181e05ff5e8c3e0f1ba1a4

When bt-gap errors and returns Error from the top level function, log
that error to syslog before returning it from the process. This aids in
diagnostics if another component requires a service from bt-gap, and
bt-gap fails during execution, as the error will now be easily visible
by reading syslog.

test: Run bt-gap in a situation where it will fail (For a motivating
example, in a hermetic testing environment where the Stash service is
not avaialble), and then run `$ fx syslog` and check for the error
output.

Change-Id: Ic1f0320289f15f36b96f4b2f0db9f34e90c7820b

When linking you can't link across a device or partition. This change
should defensively place the temp file used for atomic linking in the
same directory as where we'll be renaming it to.

Test: CQ + built Fuchsia and checked the out and .build-id directories
Change-Id: If804230d803c71379e119f6a4bda7fc1c41d5f29

This isn't used at all, and is being replaced by
fuchsia.sysmem.Allocator2 (which will be renamed).

Test: CQ

Change-Id: I32bccd0c0fc4911a2cc8bca2e8451dae34924ce8

Change-Id: Iaa147d82dccd1d43e204a925a863caa86c636892