Solve problematic missing override declarations of mocks by temporarilly
adding -Wno-inconsistent-missing-override for clang. (It appears
googletest on clang-cl isn't handling something that our gcc, clang and
msvc-cl builds all ignore without the warning override.

Workaround THROW tests which googletest with clang-cl are framing as
as -Wdiscard errors.

Workaround nodiscard next() method error

Follow up about discarded results of singleton construction and other
discarded odd or apparently irrelevant function call results.

Signed-off-by: William A Rowe Jr <wrowe@vmware.com>
Co-authored-by: Sunjay Bhatia <sunjayb@vmware.com>
Co-authored-by: William A Rowe Jr <wrowe@vmware.com>
Co-authored-by: Sunjay Bhatia <sunjayb@vmware.com>
Co-authored-by: William A Rowe Jr <wrowe@vmware.com>

- Adds a second build clang-cl using the LLVM toolchain

- Introduce debugging output for msvc-cl and clang-cl opt binaries in
  order to be able to unwind stack traces from windows core files.
  (required increasing the available heap on the RBE GCP workers)

- Fix lld-link.exe errors caused by unexported destructors

- Mark additional tests flaky/fails on windows based on clang-cl build
  observed test failures

Potential additional changes for more efficient fastbuild linkage

- Stop using alwayslink/whole_archive logic to link all //source/...
  libraries used by //test/... binaries, because this increases the .exe
  and .pdb output file sizes and computational time. This requires we
  identify the libraries which must be resolved at run-time such as
  singletons, filter configs, etc, vs. those libraries which do not need
  to be preserved in full, and then override any default alwayslink.

- Simplify binary of the objects used. (-OPT:REF optimization) due to
  tremendous size of resulting .exe and .pdb files, one alwayslink is
  resolved.

- Enable dynamic envoy libs and clib/stdlib for test fastbuild objects
  (this presumes the objects would always be invoked on the machine with
  the same compiler and dynamic runtime already installed.) Requires that
  we fix the luajit/moonjit patches to toggle the /MD or /MT flag based on
  bazel config (much like we vary their msvcbuild.bat file for debug vs.
  non-debug.)

Potential changes to bazel rules_cc for better builds

- Deduplicate `-ignore:4221` flag and system library names from link params.

- Allow /DEBUG:FASTLINK to be replaced with /DEBUG:FULL or /DEBUG:NONE,
  and validate that FASTLINK results in faster linkage+invocation times
  for tests than /DEBUG:FULL. Also determine whether /DEBUG:NONE is in
  the spirit of "minimal debugging information" as documented by Bazel.

Signed-off-by: Sunjay Bhatia <sunjayb@vmware.com>
Signed-off-by: William A Rowe Jr <wrowe@vmware.com>
Co-authored-by: Sunjay Bhatia <sunjayb@vmware.com>
Co-authored-by: William A Rowe Jr <wrowe@vmware.com>
Co-authored-by: Sunjay Bhatia <sunjayb@vmware.com>
Co-authored-by: William A Rowe Jr <wrowe@vmware.com>

Add the broken_clang_cl_dtor tag to tests which cannot compile due
to some windows-specific clang scope or optimization issue.

Issue is observed when the class used is based on a : public virtual foo,
base class, that ~foo() = default; destructor declaration is ignored, and
the linker reports that foo's dtor symbol cannot be resolved at link time.
It appears that when the base is redefined as : public foo the defect goes
away. (The virtual base class may be nested in a non-virtual base class).

It isn't possible to solve this by dropping every virtual class inherited,
due to diamond inheritance patterns.

This may be a compiler bug, or an optimization toggled by /O2 in clang-cl
which they do not enable in -O2 on linux. Tableing this for the time being.

Co-authored-by: William A Rowe Jr <wrowe@vmware.com>
Signed-off-by: William A Rowe Jr <wrowe@vmware.com>
Co-authored-by: Sunjay Bhatia <sunjayb@vmware.com>
Signed-off-by: Sunjay Bhatia <sunjayb@vmware.com>
Co-authored-by: William A Rowe Jr <wrowe@vmware.com>
Co-authored-by: Sunjay Bhatia <sunjayb@vmware.com>
Co-authored-by: William A Rowe Jr <wrowe@vmware.com>
Co-authored-by: Sunjay Bhatia <sunjayb@vmware.com>

This restores commit 01c6aa7.

This commit needs a finer grain of precision to avoid building on
clang-cl but not avoid building on msvc-cl.

Signed-off-by: William A Rowe Jr <wrowe@vmware.com>
Co-authored-by: Sunjay Bhatia <sunjayb@vmware.com>
Co-authored-by: William A Rowe Jr <wrowe@vmware.com>
Co-authored-by: Sunjay Bhatia <sunjayb@vmware.com>
Co-authored-by: William A Rowe Jr <wrowe@vmware.com>
Co-authored-by: Sunjay Bhatia <sunjayb@vmware.com>
Co-authored-by: William A Rowe Jr <wrowe@vmware.com>

These fixes reflect that the timing on windows is often less
synchronous than linux observations, due to the design of the
underlying socket stack.

- listener_impl_test
    TcpListenerImplTest.SetListenerRejectFractionIntermediate

- Enable no-longer-failing tests (no failure observed in RBE or locally)
    //test/extensions/grpc_credentials/file_based_metadata:integration_test
    //test/extensions/transport_sockets/alts:tsi_handshaker_test
    //test/integration:http2_flood_integration_test
    //test/integration:http2_upstream_integration_test
    //test/integration:overload_integration_test
    //test/server:guarddog_impl_test

Co-authored-by: Sunjay Bhatia <sunjayb@vmware.com>
Co-authored-by: William A Rowe Jr <wrowe@vmware.com>
Signed-off-by: Sunjay Bhatia <sunjayb@vmware.com>
Signed-off-by: William A Rowe Jr <wrowe@vmware.com>
Co-authored-by: William A Rowe Jr <wrowe@vmware.com>
Co-authored-by: Sunjay Bhatia <sunjayb@vmware.com>

- These skipped tests are all missing exported destructors of pure
  virtual classes. Enable the compilation to fail the clang-cl build
  and demonstrate what needs to be fixed (this may be a clang-cl
  compiler bug, or a gmocks bug, or buried elsewhere, but they do
  not appear to be envoy source code bugs.)

Co-authored-by: William A Rowe Jr <wrowe@vmware.com>
Co-authored-by: Sunjay Bhatia <sunjayb@vmware.com>
Signed-off-by: William A Rowe Jr <wrowe@vmware.com>
Signed-off-by: Sunjay Bhatia <sunjayb@vmware.com>

Signed-off-by: William A Rowe Jr <wrowe@vmware.com>

Signed-off-by: William A Rowe Jr <wrowe@vmware.com>

Signed-off-by: William A Rowe Jr <wrowe@vmware.com>

- per htuch feedback keep these expectation simple
- macros can be enhanced, but discard vs non-discard flavors would be required
- work around the error messages for clang-cl only, and clean up c[xx]opt formatting

Signed-off-by: William A Rowe Jr <wrowe@vmware.com>

Signed-off-by: William A Rowe Jr <wrowe@vmware.com>

We require Bazel > 3.7.0 to support clang-cl 11.0.0 due to faulty error detection

Signed-off-by: William A Rowe Jr <wrowe@vmware.com>

Still-failing tests specific to clang-cl 11.0.0 are;
//test/common/network:dns_impl_test
//test/extensions/filters/network/sni_dynamic_forward_proxy:proxy_filter_integration_test

Signed-off-by: William A Rowe Jr <wrowe@vmware.com>

Signed-off-by: William A Rowe Jr <wrowe@vmware.com>

Signed-off-by: William A Rowe Jr <wrowe@vmware.com>

This does not work with simply updating rules_cc for the llvm 11.0.0 fix,
the new envoy-build-tools and bazel bundled with the fix are requirements

Signed-off-by: William A Rowe Jr <wrowe@vmware.com>

Signed-off-by: William A Rowe Jr <wrowe@vmware.com>

Co-authored-by: William A Rowe Jr <wrowe@vmware.com>

- Factors in the Level Triggger PR now on master
- speed up bazel query by using startup option output_base

Signed-off-by: Sunjay Bhatia <sunjayb@vmware.com>
Co-authored-by: William A Rowe Jr <wrowe@vmware.com>

clang-cl pipeline is waiting for llvm 11.0.0 bazel 3.7.1 toolchain

Signed-off-by: William A Rowe Jr <wrowe@vmware.com>
Co-authored-by: Sunjay Bhatia <sunjayb@vmware.com>

Prevents issues when text file test fixtures are cloned with LF or CRLF
endings

Signed-off-by: William A Rowe Jr <wrowe@vmware.com>
Co-authored-by: Sunjay Bhatia <sunjayb@vmware.com>

Signed-off-by: Sunjay Bhatia <sunjayb@vmware.com>
Co-authored-by: William A Rowe Jr <wrowe@vmware.com>

- This activates a few tests observed to fail locally, to confirm
behavior in the CI pipeline when the clang-cl pipeline is added next
week.

Co-authored-by: Sunjay Bhatia <sunjayb@vmware.com>
Signed-off-by: Sunjay Bhatia <sunjayb@vmware.com>
Co-authored-by: William A Rowe Jr <wrowe@vmware.com>
Signed-off-by: William A Rowe Jr <wrowe@vmware.com>

Signed-off-by: William A Rowe Jr <wrowe@vmware.com>

Signed-off-by: William A Rowe Jr <wrowe@vmware.com>

Signed-off-by: William A Rowe Jr <wrowe@vmware.com>

Signed-off-by: William A Rowe Jr <wrowe@vmware.com>

Signed-off-by: William A Rowe Jr <wrowe@vmware.com>

Signed-off-by: William A Rowe Jr <wrowe@vmware.com>

Signed-off-by: Sunjay Bhatia <sunjayb@vmware.com>