Update bootsvc.next to support optionally specifying arguments to the
process launched by bootsvc, delimited by commas.

Note that the bootsvc tests still only run manually, though we are
investigating running them automated on CQ similar to the core-tests.
Tracked in ZX-4000.

ZX-3662 #done
ZX-4000 #comment

Test: fx --dir out/x64 build && ./zircon/scripts/run-zircon-x64 -o
out/x64.zircon -c "bootsvc.next=bin/bootsvc-tests,testargument"

Change-Id: Iecfdf22be93c8d0893ffc1b58d541a2b3ad7e97c

This is the end of the truncation madness. We're no longer stuffing GNU
build IDs into a Microsoft UUID field. We get the whole, correct build
ID every time. (Unless your crash came from a release before today, in
which case we're now pretending you don't exist).

Change-Id: I89567f399ff32db933350be8705ec0d79fa71619

Change-Id: I5d81f8021f51a09f98f1e335280bde0ea1d4f367

Include the default keys from $FUCHSIA_DIR/.ssh

DX-1308 #comment

TEST=build, 'fx netboot', verify 'fx shell' works

Change-Id: I299c4648302f205a1d5cb2d05468e7f0ac21c531

Change-Id: I7ee0f6c26007edcd5320c9998f966f2a289b5af1

These files were generated using:
https://fuchsia.googlesource.com/tools/+/refs/heads/sandbox/kjharland/fuchsia-proto-migration/cmd/generate_spec_files/

(some files failed because there's not been a green build or there was some other error.
I'll need to manually add those).

This change MUST be submitted after http://fxr/277564.

Bug: IN-1102 #comment
Change-Id: I0a2f6dab69222c81e2b8a9a74b3c08cb28ac175a

This partially reverts 3dcf5826, as it turns
out there are load bearing dependencies on these flags in infrastructure.

New tests are not added, this code is a revert to fix pipelines.

Change-Id: I4675eced5c517e99e174a991dd11481924b199e2

Tested: No behavior changed
SCN-1291 #comment

Change-Id: I60dfb9c246c5a5ee3b115b85a6e61c6e50ff0f7e

Test: CQ

Change-Id: I58d8c2b2d058a13a6c0944369b3a33f7abb8d28b

Compiling zircon with clang causes the guest integration tests to fail
because it emits instructions (that trap to the hypervisor) that both
read and write memory such as "and" and "or". This temporarily prevents
the compiler from using those instructions until the hypervisor's
instruction decoder supports them.

TEST=fx run-test guest_integration_tests

MAC-250 #done
MAC-251 #comment

Change-Id: I0f8779a6ffdbf3d75690112e3a2304aa95a800a6

Add the beginnings of monitor timeouts, implementing a subset of the
required functionality. Functionality which is present in the cited
specifications, but not listed below, will be added as needed. Note
that some of the functionality in the cited specs is conditional on
which features the stack is qualified for.

* Per Section 8.6.5.7, when the transmitter is in the XMIT state,
  and RetransTimer-Expires, we should start the monitor timer, and
  transition to the WAIT_F state.
* Per Section 8.6.5.8, when the transmitter is in the WAIT_F state,
  and MonitorTimer-Expires, we should send a receiver ready frame, and
  restart the monitor timer.
* Per Section 8.6.5.8, when the transmitter is in the WAIT_F state,
  and we "Recv ReqSeqAndFbit" with F=1, we should stop the monitor timer.
* Per Section 8.6.5.8, when the transmitter is in the WAIT_F state,
  and we "Recv ReqSeqAndFbit" with F=0, we should _not_ stop the
  monitor timer.

(References above are to Core Spec v5.0, Volume 3, Part A.)

Note that we do not explicitly represent the XMIT vs WAIT_F
states. Rather, we observe that the WAIT_F state corresponds precisely
to whether or not the monitor timer is running.

Bug: BT-442
Test: bt-host-unittests
Change-Id: Ic4064030e6f970821766fc39b7bee195e2d22018

Up until now, we emitted the raw value as it appears in .fidl source
files into the JSON IR. With the support of binary literals, backends
for target languages which did not support that syntax (e.g. Dart)
would have to resolve an interpret constants. Since we're already doing
this work in fidlc, we're simply emitting the resolved value, along with
the original value.

FIDL-486 #comment
FIDL-605 #done

Change-Id: I7e6bb28a4392f0a43be59f236a3222bb662ebb96

Turns out failure to find rssi may sometimes report 0, which leads to a debug
assert failure right now. Let's just pass these occurrences to cobalt instead.

Test: Wlan unit tests pass. Added new test for 0 rssi cases.

WLAN-1103 #done

Change-Id: Idae6a21df732f98d770a7d81e7d0fc4530886028

Revert "[mediacodec] Disable use_h264_decoder_test"

This reverts commit 70c5c544.

Reason for revert: flake fixed by I78efcf2a01dd6b918e0663a783054e47a29c423b

Tested: use_h264_decoder_test on CQ

Original change's description:
> [mediacodec] Disable use_h264_decoder_test
>
> Flaked in GI
>
> Change-Id: Ib7b4e7bc7f269b1fde89ebd1fc58e23622f236a8

TBR=vardhan@google.com,dustingreen@google.com,turnage@google.com

# Not skipping CQ checks because original CL landed > 1 day ago.

Change-Id: Ib41265b70601daf1e542eea01c2d9c0cb1589a61

In order to have additional Scenic systems depend on e.g.
FrameScheduler without depending on Gfx, this CL starts
to break //garnet/lib/ui/gfx:gfx into smaller source_sets.

SCN-1398 #comment

Testability: no behavior change.

Change-Id: I87bc89b51d27da64966d93bbe1e339fa2d33f909

Eventually the infra/config/generated directory will contain
many configs generated from starlark code.

* Remove /infra/ from .gitignore
* Document the new directory in layout.md
* Add a few infra folks to OWNERS
  * garymm: working on starlark configs for all infra
  * nmulcahey: has been reviewing recipe proto API changes.

Bug: IN-1102 #comment
Test: CQ
Change-Id: Ie231f5cab1856f9280e283006e3acc12751385d2

Instead of manually closing handles, we now use libzx smart handles in
some places.

Test: No behavior change
Change-Id: Ia5d0078ec8aae829f588aa026ce6b7b5d71f7116

Add ability for the elf runner to pass items from program.args in the
manifest to the loader service. The added integration test also covers
the elf runner better.

CF-730: #done

Change-Id: I94671d6250c9fc9d9c7f8931be50b461707fcd64

By not having buffer_index == packet_index, we should avoid building up
client dependence on them being equal.  In general they can't be equal
for VP9 and we don't want to promise they'll be equal for h264 either.

Tested: use_h264_decoder_test (with and without amlogic-video)
Change-Id: I353a7e60d906d066504955b6550cbf976484efd0

The BrwLock previously did not support PI. This splits the BrwLock semantics, through a
template parameter, into one that either has always correct PI handling, or makes
absolutely no attmept at PI.

For correct PI handling the enabling primitive is a 16byte compare-and-swap that removes
the race between setting kBrwLockWriter bit and the writer_ value.

Test: Kernel unit tests on NUC + vim2

Change-Id: I766addc50ddac95e598c5f6d1c97e1fa915b805d

These unused imports generated warnings in the build:

  warning: Library fuchsia.router.config imports zx but does
  not use it. Either use zx, or remove import.

  warning: Library fuchsia.router.config imports fuchsia.net
  but does not use it. Either use fuchsia.net, or remove import.
  using fuchsia.net;

Test: Removing unused imports makes the warnings disappear.

Change-Id: I4aac3eaca52474eddf52d6dd1162e67f254b71fe

Making it clear that classes in this library are not thread safe and
what contraints clinets will face when serving nodes over a second
loop.

Also update documentations of dependent classes.

Change-Id: I2a795450e29675951944272f79d9cda1f81402c6

This is a reland of ee539081

Original change's description:
> [run] Integrate with component_index
>
> CF-162 #comment
>
> Change-Id: Icb811b1073772923d1c471e7c1d68eac0b57281b

Change-Id: I20bba5a05ce8575a3f2e8ba1f076eee86fd44eb5

Test: No behavior change
Change-Id: I740edf02d6efc786d3696e42b5dbcf5a0d42e234

Test: CQ

Change-Id: I22cc45b8de4ff9c25cee96ee0445293498b4f644

This is just step 1. Step 2 is to convert the crosstool file into
starlark.

Bug: DX-1061

Testing:
 * Built the SDK archive, generated an SDK workspace and ran the tests.
 * This change fixes test issues found in my other bug, DX-538

Change-Id: I8bc9c423e931fb676f7acbe12249fb727610889c

This reverts commit ee539081.

Reason for revert: Fails to build in CI.

Original change's description:
> [run] Integrate with component_index
> 
> CF-162 #comment
> 
> Change-Id: Icb811b1073772923d1c471e7c1d68eac0b57281b

TBR=jamesr@google.com,anmittal@google.com,shayba@google.com,leolung@google.com

Change-Id: Ic7949fa65f76f3db8b8b713f274118d02436320f
No-Presubmit: true
No-Tree-Checks: true
No-Try: true

Bug: AUTH-161
Change-Id: I81c5e8cef2eccad45706a6ec409602944b43533e

tables to basemgr.

* Peel PseudoDirServer (and another utility) off so it can be reused by
TestHarnessImpl and ModularConfig's test.

* TestHarness was manually translating BasemgrConfig and
SessionmgrConfig to flags for basemgr. This change will make TestHarness
give basemgr /config_override and pass through all config in
TestHarnessSpec. This is tested by all the tests that exist that use
TestHarness.

* Now that config is being passed through, making sense of ShellSpec is
difficult, considering how much extra functionality BasemgrConfig gives
you (e.g., multiple session shells). This CL removes the ShellSpec way
of supplying things, and has everyone use BasemgrConfig, which is
consumed directly by modular. All test harness tests are updated and
exercise that this works, along with the
TestHarnessImplTest.MakeBasemgrConfigDir test.

* TestHarnessFixture provides helper methods for constructing
intercepted shells. `test_harness_fixture_test` test package contains
tests for each helper method introduced.

Change-Id: I4b0eec5510e84365e6d420ea35936211ad868280

When adding this, it was thought that it would be quite rare, but
this ends up firing even when tests don't fail.  Add useful data
to the log so we can track trends in them.

Test: Ran e2e locally to verify it printed.
Change-Id: I09c3458cdd884a112d401aec29bdecc684e9183c

Change-Id: Ie5597d5cce0bf4e6179946762e8158fc930a6745

Change-Id: Id13cd53cd3e75ffed5a8f05611850f581b9e9ca6

Change-Id: Ib99f7cd57f68f6624797816f4098f0812c27b450

CF-162 #comment

Change-Id: Icb811b1073772923d1c471e7c1d68eac0b57281b

Implements synchronous event handlers that are safer than the
ExpectFoo() versions found in Go.

Given the FIDL protocol

    protocol Foo {
        -> OnOpen();
        -> OnSomthing(int arg);
        A(int a) -> (int b);
    }

We can build an "event dispatcher vtable" like this:

    struct FooEventHandlers {
        fit::function<zx_status()> on_open;
        fit::function<zx_status(int arg)> on_something;
        fit::function<zx_status()> unknown;
    };

and have the client call:

    Foo::SyncClient client(std::move(channel));
    client.HandleEvents(FooEventHandlers {
        .on_open = [&] { ... },
        .on_something = [&] { ... }
    });

Doing so will not interfere with regular method call replies, since all
method calls in llcpp are sync for now. In addition, we force the caller
to handle all possibilities, so there is less chance of abuse/dropping
messages.

TEST: All llcpp tests involving events are rewritten to use this new
API.

FIDL-604 #done

Change-Id: Ib59b77f113ccb3d975dbff24f57b420b6690c12a

This is not yet implemented, and it's better to warn than silently
misrepresent what features exist.

FIDL-609 #comment

Tes: added
Change-Id: I1393a2dfffa269e757f70a75ee15fd504c734fac

AUTH-196 #comment

Tested: Existing coverage.
Change-Id: I917dcb79e38381896c50eb486ee0c60203b34eb1

TEST: No behavior change. fx run-test mediaplayer_tests
Change-Id: I48c2bee4b3c7c637d31933c5487e8a38645bfe48

This allows targets to configure themselves based on whether they are using
using the Fuchsia SDK or are directly wired into the Fuchsia buildroot.

Change-Id: Ifccf2cbc0422682336fc5b804476bfc0e23adab2

Makes it easier to validate flags and also define if node is a
directory, file, etc.

DX-387 #progress

Change-Id: Ic2ccf7ffc13bd713efa865e2ced93e7ac1de53b7