The way AnyMetadata works is by holding two pointers to the fields of
google.protobuf.Any, and its methods `PackFrom` and `UnpackTo` mutates these
pointers directly.
This works but is a bit overcomplicated. A whole new object is created to
mutate members of Any directly, breaking isolation. Each Any object will also
need to unnecessarily create room to hold an AnyMetadata object, which is at
minimum the size of two pointers.
By removing AnyMetadata and using free functions, the call site passes
`const Value& foo()` and `Value* mutable_foo()` directly into `PackFrom` and
`UnpackTo`. This is more idiomatic and will play well with hasbits.
The only catch here is that in some templated implementations of `PackFrom` and
`UnpackTo`, caller will need access to T::FullMessageName(), which is in
private scope. The workaround here is to have a single templated helper
function that is made into a friend of google.protobuf.Any, so that it has the
right visibility.
PiperOrigin-RevId: 670722407
When the user adds `option deprecated = true` for an `enum` or `message`, we
should add the `[[deprecated]]` attribute to the generated C++ `enum` or
`class`.
PiperOrigin-RevId: 670583253
We need this type to be trivial and standard layout because we directly read
its contents in Rust. This change removes its user-defined constructors so that
it becomes a trivial type.
PiperOrigin-RevId: 670572557
The symbol renaming support has assumed folks would define a
custom version of the macro, but if a developer does renaming
just by #defining the class, the extension singleton names
required manual handing, by changing to just wrap the class
in the macro and using the compiler string concat support
it becomes easier as just the class name remapping can do
the work for extensions.
PiperOrigin-RevId: 670536928
On macOS 15, the test suite was crashing due to `pipe()` failing with `EMFILE`. On inspecting `lsof`, it appears the file descriptor table was full of duplicated `tty` file descriptors.
I'm not entirely sure what's changed in macOS 15 to tip it over the limit, but there was nevertheless clearly a file descriptor leak and this PR fixes that.
We were calling `dup(1)` and `dup(2)` to copy stdout/stderr to new temporary file descriptors, storing it `original_stdout_` and `original_stderr_`, later moving it back to FD 1 and 2 but never actually closing the temporary file descriptor.
Closes#18021
COPYBARA_INTEGRATE_REVIEW=https://github.com/protocolbuffers/protobuf/pull/18021 from Bo98:fd-leak 3df847287e
PiperOrigin-RevId: 669528160
Fixed some shadowing between local variables
Pretty simple fix, just renaming a couple local variables so they don't shadow other variables in scope. Just trying to reduce verbosity of warnings in users' own builds with higher warning levels (as is my case ;) ).
Obviously, no change in behavior, and hopefully the new names are reasonable.
Closes#17971
COPYBARA_INTEGRATE_REVIEW=https://github.com/protocolbuffers/protobuf/pull/17971 from mikael-s-persson:fix/shadowing_vars ad454e7a2a
PiperOrigin-RevId: 669393734
For some reason, Clang 19 appears to raise an error in some situations unless
we explicitly inline the implementation of `StrongPointer()` here.
PiperOrigin-RevId: 669373421
This relates to:
Issues:
- https://github.com/protocolbuffers/protobuf/issues/17036
- https://github.com/grpc/grpc/issues/36518
PRs:
- https://github.com/protocolbuffers/protobuf/pull/15519
The fix in https://github.com/protocolbuffers/protobuf/pull/15519 has one line missing that breaks `Grpc.Tools` and doesn't actually fix the problem it proports to fix.
This PR fixes this.
This fix needs to also be applied to the version of protobuf that is used by gRPC.
There are various scenarios that need to be tested on Windows:
**A. Non-ascii command line arguments, e.g.**
```
protoc.exe --csharp_out=out --proto_path=E:\work\grpc\protoctest\test-Dré E:\work\grpc\protoctest\test-Dré\helloworld.proto
```
Failed output:
```
E:\work\grpc\protoctest\test-DrΘ: warning: directory does not exist.
Could not make proto path relative: E:\work\grpc\protoctest\test-DrΘ\helloworld.proto: No such file or directory
```
Success output:
- no errors printed
- generated `.cs` file in the `out` directory
**B. Non-ascii arguments in a file containing the protoc arguments (no path to file) e.g.:**
```
protoc.exe @test.rsp
```
where `test.rsp` contains:
```
--plugin=protoc-gen-grpc=E:\work\grpc\protoctest\tools\grpc_csharp_plugin.exe
--csharp_out=E:\work\grpc\protoctest\test-Dré\out
--grpc_out=E:\work\grpc\protoctest\test-Dré\out
--proto_path=E:\work\grpc\protoctest\test-Dré
helloworld.proto
```
Success output for both old and fixed code:
- no errors printed
- generated `.cs` file in the `out` directory
**C. Non-ascii arguments in a file containing the protoc arguments (with non-ascii path to file).**
(This is what `Grpc.Tools` uses.) e.g.
```
protoc.exe @E:\work\grpc\protoctest\test-Dré\test.rsp
```
Failed output:
```
Failed to open argument file: E:\work\grpc\protoctest\test-DrΘ\test.rsp
```
Success output:
- no errors printed
- generated `.cs` file in the `out` directory
Closes#17854
COPYBARA_INTEGRATE_REVIEW=https://github.com/protocolbuffers/protobuf/pull/17854 from tonydnewell:windows-utf8-command-fix de9ec5401e
PiperOrigin-RevId: 669083804
This test case attempts to pack an Any with a message over 2 GiB in size, but
the allocation is failing in our 32-bit Windows test. This change should fix
the problem by skipping the test case on 32-bit platforms.
PiperOrigin-RevId: 669082000
Now 'thunk' is only for cpp kernel we don't need lots of special casing to replicate the upb accessor names (we basically have free choice of name instead).
Since the ThunkName() function should only be used from cpp kernel, it now check-fails if its called from upb kernel, which requires pushing down the thunkname() calls to the cpp-kernel specific paths in some cases.
PiperOrigin-RevId: 668958849
This being unused in some cases is causing a warning at head, just suppressing the warning rather than conditionally emitting it only when we need it.
PiperOrigin-RevId: 668937899
The extern "C" block in messages is now:
- Cpp kernel: All of the message/accessor/oneof externs
- Upb kernel: Only the upb_MiniTable extern
PiperOrigin-RevId: 668518308
instantiate objects if permitted by the compiler.
We can use memset for zero initialized objects, and memcpy for ones cloned from
the prototype.
This permits creating objects from the parser without calling virtual
functions.
For the cases where the efficient implementation can't be used, we generate a
"placement new" style function to offload the memory allocation out of the code
generation. This reduces code bloat even when we can't use the more efficient
implementation.
Migrate many callers of `New` and similar to the new functionality. In
particular, the parsing paths will use this.
Finally, make `New` non-virtual now that `MessageLite` can handle it directly.
It reduces binary size.
PiperOrigin-RevId: 668077355
`unsafe trait` means that the _implementation_ has constraints they need to meet for the behavior to be safe (which is the intent here, namely that it should always return the same value and always be a safe to deref value), unsafe on the fn means that the _caller_ of that fn has some constraints they need to meet for it to be safe (of which there are none in this case).
PiperOrigin-RevId: 668028628
This unifies the duplicated logic that we had for each different non-repeated field to emit these same 3 methods for each of the different field types, and moves the has/clear off of upb-accessors-C-codegen and onto the upb-C-API for all cases (matching the singular scalars case which moved in a previous change).
PiperOrigin-RevId: 667953223
These types are effectively two ways to spell the same type, but MapView/MapMut is more terse, especially where the unnamed lifetime was used which can now be implied (`View<'_, Map<K, V>>` can just be `MapView<K,V>`)
PiperOrigin-RevId: 667636945
`absl::string_view`.
This is a modernization of the API and performance improvement for all callers.
Currently gated behind opt-in macro
`PROTOBUF_TEMPORARY_ENABLE_STRING_VIEW_RETURN_TYPE` for users to test and prevent
issues when the breaking change happens.
PiperOrigin-RevId: 667616632
These are very old unit-tests that appear to have been written to provide some conformance and fuzzing capabilities before we had more comprehensive solutions. Today, they should be already covered by our conformance tests and fuzz tests and create unnecessary noise.
The goal of this change is to remove the checked-in binary golden data that we no longer know how to regenerate. There is a lot of nearby code that could likely also be cleaned up in a follow-up
PiperOrigin-RevId: 667604310
We were really inconsistent on where we put Private or not and this tries to make a sensible consistent state of:
- For types that are exposed to application code, any pub methods which are only pub so they can be used by gencode (which is mostly anything that has any internal/runtime type anywhere on the parameters or return type list), have those methods have both a `Private` arg and doc(hidden)
- For structs that are only inside __runtime / __internal to begin with, put doc(hidden) on the types, and don't put Private on any of their methods since callers can't reach those types regardless.
Note that for exposed functions which also _accept_ another internal/runtime type in a parameter, the additional `Private` arg is superfluous since application code shouldn't ever be able to reach one of those internal types to be able to pass one in, but this keeps the pattern of keeping Private on it in those cases as well (the `Private` would still be the only guard on methods which only _return_ an internal type).
PiperOrigin-RevId: 667547566
Putting it into BUILD files unintentionally forces it on all our downstream users. Instead, we just want to enable this during testing and let them choose for themselves in their builds.
Note, that this expands the scope of -Werror to our entire repo for CI, so a bunch of fixes and opt-outs had to be applied to get this change passing.
Closed#14714
PiperOrigin-RevId: 666903224
This is useful/necessary when the caller wants to tie the View to a lifetime other than a stack frame, including 'static.
The documentation is clarified that the preexisting one should be preferred in most situations, and only use the new one where tying the View lifetime to a stack frame isn't suitable.
PiperOrigin-RevId: 666396427