The GetNumVirtualDevices API that was recently added to
virtualaudio can be used to eliminate the synchronization
issues that AudioDeviceTests faces, so this set of tests
can now be re-enabled.

Test: re-adds 67 test cases; no production code change

MTWN-293 #done

Change-Id: I6d4618a52835f2ef388bed1dddef9491fc2e2b8d

Previously this was copyable but not copied in practice.

Changes some of the code that took pairs because std::pairs to make the
call-sites better.

This should have no affect on functionality and I measured indexing
performance to be the same within typical measurement error.

Change-Id: I5adebf46f0be89090f8048e7b00db8b3c71c7d42

Test code needs to use the mock mmio library header while driver
code needs to use the real mmio library header.  Hence, a library
like ddktl or hwreg that is used in both contexts cannot rely on
this header directly.  Instead, make the APIs agnostic to mmio
implementation by using forward declarations and requiring users of
the mmio-related APIs to #include what they use.

The renaming of the mock library to aaa-mock-mmio-reg was a previous
attempt to work around this issue, which is actually wholly
inadequate in the current build system.  So the library is renamed
back to mock-mmio-reg.

In the hwreg library, this was achieved by using a template common
to the kernel and non-kernel cases, rather than an explicit forward
declaration.  To make that work, the argument order for
RegisterIo::Write was swapped to match MmioBuffer::Write so that
they are interchangeable in template instantiation.

Both implementations of the <lib/mmio/mmio.h> were using templates
where they should have been using simple forward declarations.  Fix
that so that separate forward declarations become tenable.

Some cases of non-driver code still using the real mmio library
remain and need to be untangled.

Bug: ZX-4062 #comment hwreg, ddktl libraries made mmio-header-agnostic
Change-Id: I824d395501b35f59ffb23ac37fffb628f163d46e

Tested: No behavior changed; base_view_unittest
SCN-1291 #comment

Change-Id: I1760e5b179b40210c37ba5eae743eff41431733a

Change-Id: I9c946b6fc2d8ded7141b07000f96f764760fa561

Performance improves by ~7% by optimizing allocation of vectors and strings
iin module_symbol_index and module_symbol_index_node before doing the indexing.

Performance improves by ~15% by optimizing the allocation of the "seen_attrs"
vector in the DwarfDieDecoder.

This was measured by running the normall-commented-out test
ModuleSymbolIndex.BenchmarkIndexing and giving it an ELF binary of a fully
symbolized release build of Chrome.

There should be no functional change.

Change-Id: I107a3d651ac05e9b2959da88a06e463173eae199

Previously, the first event of each window was not being counted, so
reported FPS was lower than it should have been.

Change-Id: I4c788c8ff2760d18c21e6abe98cfe9eab69ea3c0

This will allow creating buffer collections of fixed size that don't
contain images.

This is ignored for now, until the drivers are updated to set the value.

Test: in a follow-up patch once this does something.

Change-Id: I5e6e5e216e61f006645480f0f89d09fdd28638d8

Tested: Built and ran Escher on Linux
SCN-1391 #comment

Change-Id: I35bdcf15aae15b4724f4652c5fa9bb8e790b564b

(1) Refactors the AML Clock Driver for Testing (Inject MMIO regions)
(2) Add tests to the AML Clock Driver.
(3) Don't needlessly copy gate definitions into the driver.

Change-Id: Ib5c685ff0ac6cf23b3df8c4afd90daaffa299472

The fidl::EventSender is similar to fidl::Binding except the
fidl::EventSender just sends events and does not receive or decode
incoming messages.

Having a separate class for this functionality is useful because
EventSender does not need to couple to an async_dispatcher_t, which
means EventSender can be used in scenarios where an async_dispatcher_t
is not available or when the client wishes to send events from multiple
threads.

This CL also uses EventSender in lib/vfs/cpp in a location where an
async_dispatcher_t is not available. This change lets us remove a
dependency on the C bindings in lib/vfs/cpp, which is the last blocker
for adding lib/vfs/cpp to the SDK. The change also lets us remove one
more instance of manually constructing FIDL messages.

Test: Added event_sender_unittest.cc
Change-Id: If9d6c9cca4d16d248d3635be677d35b61d321e28

This reverts commit cf254ae8.

Reason for revert: Causes vim2 to fail to pave. See: https://fuchsia.atlassian.net/browse/DNO-519

Original change's description:
> [devhost] Wait for BindDriver in reply to Bind
> 
> Calls to fuchsia.device.Controller/Bind should only return once the
> operation completes, which is when
> fuchsia.device.manager.DeviceController/BindDriver is called.
> 
> ZX-3991 #done
> 
> Change-Id: I10607640072c8f779ac1d5be79d5a2da6de6288b

TBR=kulakowski@google.com,teisenbe@google.com,abdulla@google.com,brettw@google.com

Change-Id: I370d03abd9fa93a70a48f86f4df171f2cf87ee12
No-Presubmit: true
No-Tree-Checks: true
No-Try: true

Tracked by: SPN-8

Test: fx shell run fuchsia-pkg://fuchsia.com/hotsort_vk_bench#meta/hotsort_vk_bench.cmx"

- CL following 277184.

This CL replaces shell scripts with GN.

The GN build invokes the 'hotsort_gen' code generator and builds
architecture-specific shader "targets" (bundles of shaders and
algorithm configuration info) using a custom GN template.

The target outputs are placed in their respective $target_gen_dirs for
later inclusion of either a source set or its equivalent binary image.

The hotsort_target.gni template performs the following steps:

  1. Produce list of compute shaders that *will* be generated.
  2. Invoke hotsort_gen to generate compute shaders.
  3. Compile the shaders to SPIR-V modules with glslangValidator
  4. Invoke spirv-opt to optimize and shrink the SPIR-V modules.
  5. Invoke spirv-remap to improve compressibility of modules.
  6. Convert SPIR-V modules to array of C literals.
  7. Dump a binary image of the HotSort target bundle of modules.
  8. The GN template returns a source set of the target.

Steps to verify target outputs are being created:

  1. fx build graphics
  2. ls -R out/default/gen/src/graphics/lib/compute/hotsort/platforms/vk/tests/hotsort_vk_bench/

Change-Id: Ic98c374c70e4fda2c1d8d3284a8cbc46443eaffc

Tested: No behavior changed
SCN-1391 #comment

Change-Id: I98fdc6188963a381141448d36400d4a370ead295

Fixes hermetic dart test that require UI.

Test:
fx run-test slider_mod_tests

Change-Id: I8464abc271378031f2da133290a730c4f2b9d504

This:

MyTable t;
t.set_x(1);
t.set_y(2);
MyFunction(std::move(t));

Can now be reduce to:

MyFunction(MyTable().set_x(1).set_y(2))

Test: Table.ChainSetters; all existing tests using tables also aren't
failing.

FIDL-615

Change-Id: Ieeffa04417e92df7c0bb01d3b461dd968ba33852

This adds the Installer trait, and replaces the InstallPlan trait with
Installer::Plan, an trait to use with the associated InstallPlan type of
the Installer.

A stub implementation is provided for ease of writing tests and for
prototyping work.

Bug: PKG-465
Test: fx run-test omaha_client_lib_tests
Change-Id: Id644c3a41c78a1fce09fa8575027eb7c71c9943d

Tested: No behavior changed
SCN-1018 #comment
SCN-1291 #comment

Change-Id: Ic2dba8b13d6aaea74c19013cdb2561d601114a70

on a QEMU bot, the stub logger took longer than the timeout (100ms) to
send the first message (before it goes to sleep for 200ms) so we
increase the timeout to 500ms (and the sleep to 1s).

FLK-225 #done #comment

TESTED=`fx run-test feedback_agent_tests -t log_listener_unittest -- --gtest_filter=CollectSystemLogTest.Succeed_LogCollectionTimesOut`

Change-Id: Ia8e135e4ac2305076e73658916a5f6c6b0bddda6

Add a simulator configuration to lay the groundwork for driver
unit tests. This first step creates a build target that doesn't
use the zircon driver/device framework. To do so, we create two
sets of sources, one (for the actual driver) that instantiates
a devmgr device and one (for testing) that instantiates a
simulated device.

WLAN-1126

TEST: Built, ran brcmfmac unit tests, ran smoke test with
      fx-linksys. This patch is the first step in a unit test
      framework for the brcmfmac driver.
Change-Id: Icecadcfd75f727084447953c6a521775ca499737

Change-Id: Ia1db1813d4c6f491e0ca02945ca86ffa6f6f28c6

This is first cut to keep review rolling. The patch allows me to
incorporate feedback from review as I keep working on the tool.
There are plenty of TODOs that might serve as a hint to reviewers
what I intend to do in the future, There are codes that look noop
which I have retained because it makes testing easier.
The current version parses only one argument (a file name) and
runs load for few seconds or a minute at max.

When run - output looks like this.
$ odu /tmp/f
===== Aggregate Stats =====
Runtime:50.15060829s IOs:303 write_size(MB):62.40
IOPS:4.79 LAT(us):208579.46 BW(MB/s):0.98

If I print thread specific stats and stage stats the output would look
like

$ odu /tmp/e
===== For generator-0 =====
Generate TimeInStage:344791630 IOPS:292.93 LAT(us):3413.78
BW(MB/s):58.01
Issue TimeInStage:24016252450 IOPS:4.21 LAT(us):237784.68 BW(MB/s):0.83
Verify TimeInStage:6253520 IOPS:16150.90 LAT(us):61.92 BW(MB/s):3198.20
Runtime:49.78938866s IOs:101 write_size(MB):20.80
IOPS:4.21 LAT(us):237784.68 BW(MB/s):0.83
===== For generator-1 =====
Generate TimeInStage:494067320 IOPS:204.43 LAT(us):4891.76
BW(MB/s):40.48
Issue TimeInStage:22197400730 IOPS:4.55 LAT(us):219776.24 BW(MB/s):0.90
Verify TimeInStage:1960320 IOPS:51522.20 LAT(us):19.41 BW(MB/s):10202.42
Runtime:52.46426601s IOs:101 write_size(MB):20.80
IOPS:4.55 LAT(us):219776.24 BW(MB/s):0.90
===== For generator-2 =====
Generate TimeInStage:634360370 IOPS:159.22 LAT(us):6280.80
BW(MB/s):31.53
Issue TimeInStage:24582622520 IOPS:4.11 LAT(us):243392.30 BW(MB/s):0.81
Verify TimeInStage:21145570 IOPS:4776.41 LAT(us):209.36 BW(MB/s):945.82
Runtime:48.96960358s IOs:101 write_size(MB):20.80
IOPS:4.11 LAT(us):243392.30 BW(MB/s):0.81
===== Aggregate Stats =====
Generate TimeInStage:1473219320 IOPS:205.67 LAT(us):4862.11
BW(MB/s):42.08
Issue TimeInStage:70796275700 IOPS:4.28 LAT(us):233651.07 BW(MB/s):0.88
Verify TimeInStage:29359410 IOPS:10320.37 LAT(us):96.90 BW(MB/s):2111.76
Runtime:52.47356284s IOs:303 write_size(MB):62.40
IOPS:4.28 LAT(us):233651.07 BW(MB/s):0.88

Utility saves the arguments to generator in a file called
"./output.confing" whose content may look like
[{"magic_number":5720826294568186926,"process_id":3,"name":"generator-0",
"generator_unique_id":0,"block_size":4096,"max_io_size":1048576,"align":
true,"seed":0,"target_name":"e","target_range":{"start":0,"end":6990506},
"target_type":"target_file","issuer_queue_depth":40,"max_io_count":100,
"sequential":true}...]

Test: fx run-test odu_tests -- --nocapture

Change-Id: I219379093dfef14a0979678ce2e34a76fc5b9bbf

LE-618 #done

Change-Id: Ibdc6d9a9bf187b7127e82c614b5fb7540d0ede09

Test: CQ

Change-Id: I8382094ead019dbfc3a5447ec5a27dccbd1a5f3f

LE-618

Change-Id: Ib68941ffd47954c79dd8784c713c1e1d37607f3c

After moving fbl::RefPtr from std::shared_ptr, removing raw pointers
where appropriate and removing unnecessary uses of .get(), which can be
dangerous.

TEST: no functional change
Change-Id: I143faeb33c98013119fcb3e14d50f3d6648316ee

Moving from std::shared_ptr and std::set, etc. to fbl's intrusive containers
had been planned here for quite awhile as evidenced by johngro@'s TODOs;
this also brings us in line with the style guidelines as fbl::RefPtr is
half the size of std::shared_ptr.

TEST: audio_fidl_tests
Change-Id: Id9734cb8a5099f061a43b017b42a9224bc03fffa

Just like change Ib5d4d93b45152f8117ef2ad3fcdba6c70638bef9 but for the
(Singly|Doubly)LinkedList classes. HashTable is also supported since
it's backed by those.

TEST: new stuff in intrusive_(singly|doubly)_linked_list_tests.cpp
Change-Id: Id00347b4191744b4ebdb2c23f12a2fae65451f2f

Currently if you want to have your class able to live in multiple
WAVLTrees at once, you have to give it multiple WAVLTreeNodeState
members and define the node_state function in multiple NodeTraits
classes.

This change greatly simplifies this use case by allowing a
class to inherit multiple times from fbl::WAVLTreeContainable and pass
it a different tag type as the second template parameter each time. You
then pass the same tag type to the WAVLTree template and you're good to
go.

I also defined a helper template fbl::ContainableBaseClasses which allows
one to inherit from multiple Containable classes (even ones that aren't
necessarily WAVLTreeContainable) without also explicitly inheriting from
the fbl::HasTagTypes class, which is mostly an implementation detail.
This helper template also contains static_asserts to give understandable
error messages for common errors.

TEST: CQ, new stuff in intrusive_wavl_tree_test.cpp

Change-Id: Ib5d4d93b45152f8117ef2ad3fcdba6c70638bef9

Eventually zx_clock_get_new will replace zx_clock_get.

Tests: No behavior change.
Change-Id: I7884752bb866e72c338cf63bf64739894103ff01

appmgr is missing an OWNERS file, which can lead to changes being made
without CF being aware.

Change-Id: Iffa31b40ac2c78d592167c334e9614aa2f90c74f

Prior to this, if you had multiple references to a nullable xunion in
the flat AST, it would include multiple copies of that xunion in the
list of coded types.  Since the list of coded types is directly used to
generate xunion declarations (see TablesGenerator::Produce), it could
emit multiple versions of the same nullable xunion into the C generated
code, leading to compilation failure.

Also applies the same logic to nullable structs and unions, although
they don't have the same failure mode.

Change-Id: I840beaa607f390a9ec7e685d6c0e37477445d657

Calls to fuchsia.device.Controller/Bind should only return once the
operation completes, which is when
fuchsia.device.manager.DeviceController/BindDriver is called.

ZX-3991 #done

Change-Id: I10607640072c8f779ac1d5be79d5a2da6de6288b

When an environment() sets variant_selectors so it doesn't match the
default list used in common environments, things get even hairier for
environment_redirect().  Previously environment() documented that
default_variants must be in the list for environment_redirect() to work
right, but even that was not actually enough.  Instead, remove that
requirement and make environment() always define a dummy environment with
the "default redirect" variant name that environment_redirect() will
choose.  Make that logic choose only from default_variants and line up with
the fallback dummy variant name logic in environment().

Bug: BLD-430
Change-Id: I399a38ba494f4fafc28a3d9e93dece101eab3732

This allows the OTA tests to be used between two local builds.

Change-Id: Ib9e4088fe0d6d17dda57d6b628e5ca636e48b806

- Convenience methods GetProperties,GetConfig no longer that convenient
- move to "internal" namespace
- use weak pointer to access Controller functionality

Tested: fx run-test cpuperf_tests
Change-Id: I36002719309a4798447b6e3c79692298990fb1e2

This reverts commit 1651ebe6.

Reason for revert: The change rolled too much into one commit and did not get anything green.  It needs to be split up and reviewed properly.

Original change's description:
> [utest] Don't hard-code /boot location for zircon test files
> 
> This change fixes the zircon tests that were failing in the `bringup`
> build.
> 
> The tests were failing because they had hardcoded paths at "/boot/..."
> which failed when the tests were run from "/system". The fix is for
> runtests to pass a TEST_ROOT_DIR environment variable, which is
> "/system" or "/boot" depending on the location of the test. The tests
> use this variable instead of hardcoding the path.
> 
> Bug: BLD-463 #comment fix broken tests
> Change-Id: Id6478f68600a55f13dc8c0bde19253c909a03694

TBR=mcgrathr@google.com,geb@google.com,shayba@google.com
Bug: BLD-463

Change-Id: I9d5a296eee311d48ac45a7dd8339d889aca18d4d

Change-Id: Ida400311e4757c34aa23e42fee0c658966ce13a4

Happy to tweak wording as necessary.

Change-Id: I4db141b3ad0436a861d90905b9bf7c9040b649c5

Change-Id: I9378d445e53b88cade30ecdab28021bb4dc4137c