Thanks for tackling this, @tymcauley !
* big endian docker test -- wip
* tweaks
* tweaks
* tweaks
* docker tweaks
* fix conditional compilation issues
* reactivate other docker tests
* try some more cross-platform config (from tymcauley)
* Update tests/docker/languages/Dockerfile.testing.rust.big_endian.1_30_1
Co-Authored-By: rw <rw@users.noreply.github.com>
* Update tests/docker/languages/Dockerfile.testing.rust.big_endian.1_30_1
Co-Authored-By: rw <rw@users.noreply.github.com>
* Update tests/docker/languages/Dockerfile.testing.rust.big_endian.1_30_1
Co-Authored-By: rw <rw@users.noreply.github.com>
* Resolved Rust warnings during big-endian builds.
* Unify Rust test suites for x86 and MIPS builds.
Note that I had to add four extra packages to the MIPS `Dockerfile`:
`libexpat1`, `libmagic1`, `libmpdec2`, and `libreadline7`. For a reason
I couldn't identify, even the simplest Rust MIPS binaries run with
`qemu-mips` would fail with a segfault when run through this
`Dockerfile`. After installing the `gdb-multiarch` package to attempt to
debug the issue, the binaries ran successfully. I pared down the
packages installed by `gdb-multiarch`, and these four packages are the
minimum subset necessary to get Rust MIPS binaries running under
`qemu-mips`.
* Changed Rust tests to use `Vector`s instead of direct-slice-access.
The direct-slice-access method is not available on big-endian targets,
but `flatbuffers::Vector`s provide an array interface that is available
on all platforms.
* Resolved FooStruct endianness issues using explicit struct constructor.
This more closely resembles how FlatBuffers structs are constructed in
generated Rust code.
* Added explanation of how `FooStruct` parallels generated struct code.
Also collected duplicate implementations of `FooStruct` into a common
location.
* Fix lifetime in union _as_ accessors
In the accessors for union field, the return value is implicitly taking the lifetime of &self.
This is irrelevant and prevents usages of the returned value, because it is needlessly bound to the parent field lifetime.
This patch makes the return value inherit the lifetime of the data, like other methods do.
vtable and vtable size depends only on `Table#bb_pos` but calculated in
`Table#_offset` method on each field lookup.
Doing this with every call of `Table#__offset` is redundant.
These values can be read once with change of `Table#bb_pos` and reused
for any field lookup.
* Remove byte* property in ByteBufferAllocator.
This allows consumers to read/write into native memory, but without
having to always pin the managed `byte[]` when working with managed
memory. This allows for users to not need to Dispose() ByteBuffers
when they are using the default ByteArrayAllocator class.
Instead, we use `Span<byte> GetSpan()` methods to get access to the
underlying memory buffer.
Fix#5181
* Add a set of benchmark tests.
* Add ReadOnly spans.
This allows consumers to use ReadOnlyMemory<byte> as the backing storage
for ByteBuffers, which is useful in read-only scenarios.
* Run tests using ENABLE_SPAN_T in appveyor.
* Fix FlatBuffers.Test.csproj to work on older MSBuild versions.
* Change the test script to test UNSAFE_BYTEBUFFER
* Address PR feedback.
Remove IDisposable from ByteBuffer.
* Respond to PR feedback.
* Add utility for checking the encoding of source files
- accept source files with ASCII or UTF-8 without BOM
- accept only CRLF line ending
* Fix non-ascii symbol in idl_parcer.cpp
* Remove BOM from test.cpp
There is a test code error that causes the CanReadCppGeneratedWireFile test to fail when ENABLE_SPAN_T is defined. When TestarrayofboolsLength is not 0, then the GetTestarrayofboolsBytes() should have a length.
Introduce a HeapByteBufferFactory singleton instance in order to reduce allocations.
Clarify the usage of LITTLE_ENDIAN ByteBuffers in ByteBufferFactory.
This is to protect against cases where part of a project is
compiled with or without this flag, making for very fragile
and hard to find bugs, such as sizeof(Verifier) changing.
Change-Id: I01c895cdc5b44f860e4b0b9c9613bff1983e2b9d
See: https://bugs.chromium.org/p/chromium/issues/detail?id=929847
* [Rust] Added global namespace imports
* Documented the need for global imports
* Added white_space params to GenNamespaceImports
* Removed a \n from GenNamespaceImports
* Add `const` keyword to the `operator-(const uoffset_t &)` function in
`VectorIterator`
* Support reverse iterator in Vector
Introduced a new VectorReverseIterator type. We cannot directly use
`std::reverse_iterator<VectorIterator>` because the signature of
`operator*` and `operator->` in the VectorIterator class are not
standard signatures.
Also added `rbegin()`, `rend()`, `cbegin()`, `cend()`, `crbegin()`
and `crend()` in the Vector class.
* Fix high certainty warnings from PVS-studio
- Introduced FLATBUFFERS_ATTRIBUTE macro to use [[attribute]] if modern C++ compiler used
* Update the note about __cplusplus usage in the MSVC
* Add `NaN` and `Inf` defaults to the C++ generated code.
* Refactoring: add FloatConstantGenerator
* Refactoring-2:
- remove isnan checking for all float/double values
- add most probable implementation of virtual methods of FloatConstantGenerator
* Add conditional (FLATBUFFERS_NAN_DEFAULTS) isnan checking
Track and emit required FlatBuffers namespace imports in generated Go code.
Update Go code generator by moving most functionality into the generator class, to facilitate namespace tracking. (Note that the git diff in this combined commit may appear large due to this refactoring, but very little code was actually changed.)
Update Go code generator by tracking namespace imports when generating FlatBuffers code.
Update Go code generator by emitting package imports to correctly reference code in other FlatBuffers namespaces.
Create Go test that checks the usage of InParentNamespace objects (as defined in the example schema).
Create Docker test that checks the Go language port.
Fixes#4883Fixes#3927
Individual commits:
* remove "static" from soon-to-be method functions
* move almost all functions into class as methods
* set current namespace and emit package names if needed
* track imported namespaces
* parent namespaces work
* docker test for go ^1.11
* update base image name for go docker test
* remove cerr debugging
* formatting fixes
* re-run generate_code.sh
* explicitly test namespace imports and usage
* Enable flatbuffer to initialize Parser from bfbs (#4283)
Now its possible to generate json data from bfbs data type and flatbuffers data
and visa versa.
* add deserialize functionality in parser from bfbs
* add small usage sample
* Fix build break
* Merge branch 'pr/1' into fix-issue4283
* Fix buildbreak
* Build monster_test.bfbs with --bfbs-builtins
Attribute flexbuffer has be included in bfbs. Only with this attribute test
will run. By initialization a parser by a bfbs the attribute has to be known
for this filed. monsterdata_test.golden has a flexbuffer field so parse would
fail.
* Fix generate_code.sh
* Revert automatic indent changes by IDE
* Auto detect size prefixed binary schema files
* Use identifier (bfbs) to detect schema files
This runs a script in TravisCI that executes a bunch of small Docker image
scripts to test the language ports in isolated environments. This allows us to
test multiple language versions with little additional complexity.
Covers:
+ Java OpenJDK 10.0.2
+ Java OpenJDK 11.0.1
+ Node 10.13.0
+ Node 11.2.0
+ Python CPython 2.7.15
+ Python CPython 3.7.1
+ Rust 1.30.1
Multiple calls of e.g. CreateString inside a call to a CreateTable
could cause those strings to end up in different locations in the
wire format, since order or argument evaluation is undefined.
This is allowed by the FlatBuffer format, but it is not helpful,
especially when debugging the contents of binaries, or comparing
against a "golden" binary for tests etc.
Now making sure that all the CreateTableDirect calls first serialize
sub strings/vectors before calling CreateTable.
Also made similar changes to the serialization of "binary schemas".
Change-Id: I5747c4038b37a0d400aca2bc592bec751cf5c172
* Make the Parser independent from the global C-locale
* Set a specific test locale using the environment variable FLATBUFFERS_TEST_LOCALE
* Remove redundant static qualifiers
Tests for third_party code are run out of the main workspace. This
isn't an issue when the main workspace is the
com_github_google_flatbuffers workspace, but is an issue when you are
running the tests from another repository.
To reproduce, use "git_repository" to add flatbuffers to a project and
then run:
bazel test @com_github_google_flatbuffers//:flatbuffers_test
Some generic C++ and Rust code is not generated when unions use type
aliases because of potential ambiguity. Actually check for this
ambiguity and only disable offending code only if it is found.
This is because they are incompatible with C++ and possibly other
languages that make them minimum size 1 (to make sure multiple
such objects don't reside at the same address). Forcing them to size
1 was also not practical, as that is requires updating the logic
of a lot of implementations and thus possibly backwards incompatible.
More here: https://github.com/google/flatbuffers/issues/4122
Change-Id: I2bfdc8597b7cfd2235bb4074bb2ae06f81f8e57d
* Add '-fsanitize' optional flags to flattests and flatc targets
Control: -DFLATBUFFERS_CODE_SANITIZE=(ON | OFF | "=memory,undefined")
Travis-CI: building with -DFLATBUFFERS_CODE_SANITIZE=ON
* Fix -pie flag
* Cleanup
tymcauley