Confirm that the custom abigen attributes are handled as expected when
outputting .abigen from banjo.

(Also fixes unnecessary space in array spec that became apparent from
looking at test output.)

ZX-3882 #comment [abigen][banjo] Tests for attribute markup

Test: new codegen test
Change-Id: Id51fbaf0a4c8286af3ef8079dba2518a900430e9

Ensure the ARG variable that holds the current argument value is set.

See also: https://fuchsia-review.git.corp.google.com/c/fuchsia/+/270197

Change-Id: I00a9569416f7a3dcfd73c937ae3b699dd63a0c74

TESTED=Compiled.
BUG=WLAN-895

Change-Id: Ia7b9eb70d6ae9a8de389e10b084f430468394412

TESTED=Compiled, run on device and code reached the beginning of
       iwl_op_mode_mvm_start().
BUG=WLAN-895

Change-Id: I2785b432127f1b7f786c305695834c4aa45e45fe

Test: CQ

Change-Id: I119265c417f68c37934367d200b4f6efbd8c09fb

Adds info capturing the last time each session applied an update to the
scene graph, and the timing of the most recently applied scenic frame.

Tested: ran fx debug-report on astro
Change-Id: Iea0ce669395384faa1b32553390ceadf67652b19

Change-Id: I6cad0ff500c4f89607e4ac3c32c51fb270df5d37

libc++fs is now a part of libc++, libc++abi.a is now merged into
libc++.a on macOS.

Change-Id: Ied6a95ccaffb98d8655e102c24e3e4ed735424b1

This gets the Intel magma ICD resources our of the system_image
package.

ZX-3897 #comment

TEST=build for NUC, verify graphics system starts

Change-Id: Ida2d88659be024658942f35173ec7d2490cd3f68

* some consumers of the sdk complain about this

Change-Id: Icbf676192ba7b666455b99445bafe549a56768a4

This copies the naming scheme and list pattern used by the
rewrite service, and slims down the number of configs used
in the listing tests.

Change-Id: Id4cd55c32401d57c281d7188bcd544395d61e323

Convenience method for creating a MerkleTree from e.g. a file.

PKG-672 #done

Change-Id: I9994d9e5f5d63ad63fcd9dbc87d7a67f2d61705d

Change-Id: I2bae931f3c8581c0091bbe679366c2df1258e7df

This allows PkgUri and RepoUri to be used as keys in HashMaps and
HashSets.

Change-Id: I30f7fb1a893ec84ce83bb2bff6c9ede4fd39bf1a

The new Provider interface replaces both openat and the get_handle
ioctl. The new approach to getting a controller client is to:
1. Open the device-controller device. This no longer creates a new
instance.
2. Call one of the new Provider methods on the service handle to the
device.
3. Maintain the connection by keeping alive two channels: one for the
device instance, and one for the FIDL interface.

Eventually those two channels should be merged, but that's blocked on
support for llcpp, see ZX-3889.

ZX-2826 #done
ZX-3277 #comment removes open_at from the display driver

Tests:
Boot device (tests virtcon)
display_test
gfxtest (tests framebuffer library)
vkcube_on_fb (tests vulkan swapchain library)
scenic app (tests DisplayWatcher)
recovery app (tests rust framebuffer library)
gfxlatency

Can't run:
display_capture_test (no HW)

Change-Id: Ie187e43692ed7c947920511188c98da38c9843f3

Don't call 1.1 apis if not supported.

Change-Id: I41ae1803f3b05ce110106646dcefba81e5c20926

The 'waterfall' demo is based on the soon-to-be-deleted
"PaperRenderer 1".  Delete it, and rename 'waterfall2' to
'waterfall'.

ES-193 #done

Change-Id: I19d8508754ed6f072340cc6a6bee2d5e928d3adc

This cl adds, ViewRef struct which will be used for Semantic
Registration process by the client.

Testing:
  * Build
  * Manually installed on device.

MI4-1736 #comment

Change-Id: I96defec65cc8f9f536490e56f844fd234fb3fbbd

When we handle the zbi_cpu_config_t type on ARM systems we will now
convert it to a new flat topology and feed it into the topology
subsystem. This allows us to have the Topology subsystem as the source
of truth for topology logic on all platforms. Eventually we will remove
the zbi_cpu_config_t alltogether.

ZX-3068
Test: Booted and "k ut all" on:
*Qemu arm
*Astro
*Vim2

Change-Id: Ia22eb29146b56e030d540dd92e469e48f0726e4b

This also remove the MessageLoopTarget as an interface and makes what
previous was MessageLoopAsync that. This means that now there is no
abstract interface for the target but rather a concrete implementation.

TC-447 # Removed from debug agent.

TEST=Manual: Running binaries and components.
     Unit & Integration tests (fx run-test debug_agent_tests)

Change-Id: Ia2b41ca660cebae70d8d35e1348c97cbddffab68

These are no longer valid build arguments. Please use
base_package_labels, cache_package_labels, and universe_package_labels
instead.

Change-Id: I3213b7d72b84b35c75d21106af91e648b415ca09

Change-Id: I543310f31119e8639e918a978ef940af3ea90f03

Change-Id: I94162ee8c71c057ed90b5f7a53e98c545cc309be

WLAN-1050 #done

Test: fx full-build
Change-Id: I083db44fadaad0d8e77179103b4ab47ae65d9e70

Change-Id: Id9c2406b7fb715b5bbb91ab999be687692569e29

This is intended as a transitional step on replacing abigen with banjo.
The goal of this backend is for it to be a relatively short-lived as a
migration path:

1. Create abigen backend for banjo
2. Replace hand-written abigen with output of equivalent .banjo file as
   a Zircon build step
3. Create additional backend(s) that generate the current outputs of
   abigen, but as there are currently 13, proceed one or a handful at a
   time
4. Once the build is consuming files generated directly from
   banjo, remove the abigen backend (and the abigen tool).

ZX-3882 #comment [abigen][banjo] Add beginnings of abigen backend for banjo

Test: new tests added in tools/banjo/banjo/test/abigen

Change-Id: I96a332e96c6d6332cfe74b62fbffd852f8a9e9b7

CLion has an incomplete index w/o these additions.

Test: Verified index is now complete.
Change-Id: Ifd3638e9197af515c3396a4373034c0eb9bea6af

Change-Id: I631566d0d9c5ecc96f7d45238fe5c094b4ef2842

The old gen/build_arguments.md files from the pre-fuchsia.git repos
have not been updated in a long time.  The docs-roller bot now
updates //docs/gen/build_arguments.md instead.  But since that file
didn't exist, bot runs have not been updating *any* file for the
Fuchsia GN build arguments documentation, only the new Zircon file.

Remove the defunct peridot/ file and move the garnet/ file to
//docs where the next bot run should update it properly.

Change-Id: I6430b4af4ca796fa12ae366ce899c1c3afc0856b

Bug: DX-1201
Bug: DX-1269
Change-Id: I19936ae5489f485c745fff86c464c879c897cfb7

Change-Id: I8260bb84c136fa4f088c99934cacad575b8c4df3

This fixes the huge volume of bad readings seen due to unimplemented
wlan_rx_info for data frames. Subsequent work should resolve these instances
of missing info.

WLAN-1018

Test=ran existing unit tests and verified histogram fixed in `wlan stats`

Change-Id: I3395b65c4060664f20a3bdff0d08b03fbc8458cd

- actions: convertTrace() accepts multiple files

  - Pass additional zedmon file when -zedmon specified

- Temporary: Local copy of minimal tracing event structs required for
  zedmon trace

- Hardcoded zedmon pseudoprocess PID: "zedm" ASCII

- Channel-based zedmon interface

  - On run(): Stream CSV into structs, then JSON bytes

    - Return channels for data and errors

  - On stop(): Stop streaming and cleanup

Change-Id: Ibf3e9693714d310f62c99aec7f10474a571fc829

Change-Id: I7d190d7293ebb77184fd36bfed934c2c1cb27eda

Change-Id: I512517aa1d0f76a07a6c54dc504db8f4da1366e2

As part of migrating towards a unified A/B/R boot image configuration
for all supported boards the following changes were made:
* A chromebook board file was added.
** VBOOT images are only built for the new target.
** VBOOT images are treated similarly to how signed images on other
   boards are treated.
* The EFI image was deduplicated into bootloader.
* The KERNC image was deduplicated into ZIRCON-A.
* Support for a VB_META_R partition was added.
* zedboot version was updated (due to breaking changes)

ZX-3861

Tested: Paved pixelbook, NUC, astro, and vim2
Change-Id: I2344048ef69dd7b5e53004c707a261f5e0e0bb93

This removes some repetition in the items
`fuchsia_uri::FuchsiaBootUri` and `fuchsia_uri::FuchsiaPkgUri`.

Change-Id: I400a15b98a8fc6e476fa15c3d920c1d5a13780ab

Change-Id: Icd85ede1a0f4f9492e91907039ba012a908db456

Previously, since a waker was reused between loops in recv_all, a
deadlock was possible:

    TASK 1  -> poll_recv_event
            <- pending, will wake task 1 on next chan.recv_from msg
    TASK 2  -> send MyRequest, poll_recv_msg_response
            <- pending, will wake task 2 on next chan.recv_from msg
    **event comes in, chan.recv_from's waker is called**
	    <- wake task 2
    TASK 2  -> poll_recv_msg_response
            <- pending, will wake task 2 on next chan.recv_from msg
	    <- but hey also event is ready so wake task 1
    TASK 1  -> poll_recv_event
            <- ready, will wake task 1 on next chan.recv_from msg
    **task 1 calls await!(Timer(1 year))**
    **MyRequest reply comes in, chan.recv_from's waker is called**
            <- wake task 1
	       ^ this will not poll or get recv_all called, since task 1
	         is waiting on a timer, not a FIDL event, so TASK 2 will
		 never be woken up with its now-ready response.

Eventually, after one year, task 1 will stop awaiting and either drop
the EventReceiver or await!() on it again, either of which will call
recv_all and wake task 2. This may be a contrived example, but I ran
into it when Task 1 requested a (futures-aware) lock that task 2 was
holding. Task 1 was blocked waiting on the lock, and task 2 never got
woken because of the logic above.

This CL makes several changes:

- recv_all now uses the same logic as wake_any to pick a Waker, instead
of using the initially-used waker, which may not be polling if the
Future is Ready after recv_all. In the example above, the last
poll_recv_event would cause the still-waiting task 2 to be registered as
the recv_from msg waker, instead of the no-longer-waiting task 1.
- poll_recv_msg_response and poll_recv_event now update their waker
*before* calling recv_all, so that recv_all can correctly select their
waker to be the next chan.recv_from waker if they'll still be Pending
after recv_all is called.
- When the EventListener is woken, it is now set to
EventListener::WillPoll to match the Received behavior in responses.
This means the comment in wake_any() is not necessary, since
already-called events that are now awaiting on something else will not
be selected in get_pending_worker(). EventListener::New was also renamed
to EventListener::WillPoll to clarify this new use of the state. This
also means the previous comment in wake_any is no longer relevant, since
an event receiver will only be the waker if got Pending on its last
poll.

Change-Id: Id3e23fd4575d220117207e43b6a5ce6b1e419f33

Change-Id: Ic129292339bda65c5a59e98a72cca5e257e40953