…er instead of a ByteSlice

In #30 we added the `ddprof_ffi_Buffer_from_byte_slice` function to
allow FFI users to provide their own data to be reported using
the `ProfileExporterV3` (via `ddprof_ffi_ProfileExporterV3_build`).

Thus, if one wanted to report some data, it first converted it
from a `ByteSlice` into a `Buffer`, and then invoke `libddprof` with
it.

Here's a (simplified) example from the Ruby profiler:

```c
  ddprof_ffi_File files[1];
  ddprof_ffi_Slice_file slice_files = {.ptr = files, .len = 1};

  ddprof_ffi_Buffer *pprof_buffer =
    ddprof_ffi_Buffer_from_byte_slice((ddprof_ffi_ByteSlice) {
      .ptr = /* byte array with data we want to send */,
      .len = /* ... */
    });

  files[0] = (ddprof_ffi_File) {.name = /* ... */, .file = pprof_buffer};

  ddprof_ffi_Request *request =
    ddprof_ffi_ProfileExporterV3_build(exporter, start, finish, slice_files, timeout_milliseconds);
```

This approach had a few downsides:

1. It copied the data to be reported twice. It would be first
  copied into a `Buffer`, and then inside
  `ddprof_ffi_ProfileExporterV3_build` it would be copied again.

2. **Callers manually needed to clean up the `Buffer` afterwards
  using `ddprof_ffi_Buffer_free`**.

After discussing this with @morrisonlevi, we decided to go the
other way: change the `File` to contain a `ByteSlice` directly.

This avoids the extra copy in (1.), as well as the caller needing
to manage the `Buffer` objects manually in (2.).

This is a breaking API change for libddprof FFI users.

…er instead of a ByteSlice

In #30 we added the `ddprof_ffi_Buffer_from_byte_slice` function to
allow FFI users to provide their own data to be reported using
the `ProfileExporterV3` (via `ddprof_ffi_ProfileExporterV3_build`).

Thus, if one wanted to report some data, it first converted it
from a `ByteSlice` into a `Buffer`, and then invoke `libddprof` with
it.

Here's a (simplified) example from the Ruby profiler:

```c
  ddprof_ffi_File files[1];
  ddprof_ffi_Slice_file slice_files = {.ptr = files, .len = 1};

  ddprof_ffi_Buffer *pprof_buffer =
    ddprof_ffi_Buffer_from_byte_slice((ddprof_ffi_ByteSlice) {
      .ptr = /* byte array with data we want to send */,
      .len = /* ... */
    });

  files[0] = (ddprof_ffi_File) {.name = /* ... */, .file = pprof_buffer};

  ddprof_ffi_Request *request =
    ddprof_ffi_ProfileExporterV3_build(exporter, start, finish, slice_files, timeout_milliseconds);

  ddprof_ffi_Buffer_free(pprof_buffer);
```

This approach had a few downsides:

1. It copied the data to be reported twice. It would be first
  copied into a `Buffer`, and then inside
  `ddprof_ffi_ProfileExporterV3_build` it would be copied again.

2. **Callers manually needed to clean up the `Buffer` afterwards
  using `ddprof_ffi_Buffer_free`**.

After discussing this with @morrisonlevi, we decided to go the
other way: change the `File` to contain a `ByteSlice` directly.

This avoids the extra copy in (1.), as well as the caller needing
to manage the `Buffer` objects manually in (2.).

This is a breaking API change for libddprof FFI users.

…er instead of a ByteSlice (#33)

* [PROF-4780] Breaking change: Change FFI File struct to contain a Buffer instead of a ByteSlice

In #30 we added the `ddprof_ffi_Buffer_from_byte_slice` function to
allow FFI users to provide their own data to be reported using
the `ProfileExporterV3` (via `ddprof_ffi_ProfileExporterV3_build`).

Thus, if one wanted to report some data, it first converted it
from a `ByteSlice` into a `Buffer`, and then invoke `libddprof` with
it.

Here's a (simplified) example from the Ruby profiler:

```c
  ddprof_ffi_File files[1];
  ddprof_ffi_Slice_file slice_files = {.ptr = files, .len = 1};

  ddprof_ffi_Buffer *pprof_buffer =
    ddprof_ffi_Buffer_from_byte_slice((ddprof_ffi_ByteSlice) {
      .ptr = /* byte array with data we want to send */,
      .len = /* ... */
    });

  files[0] = (ddprof_ffi_File) {.name = /* ... */, .file = pprof_buffer};

  ddprof_ffi_Request *request =
    ddprof_ffi_ProfileExporterV3_build(exporter, start, finish, slice_files, timeout_milliseconds);

  ddprof_ffi_Buffer_free(pprof_buffer);
```

This approach had a few downsides:

1. It copied the data to be reported twice. It would be first
  copied into a `Buffer`, and then inside
  `ddprof_ffi_ProfileExporterV3_build` it would be copied again.

2. **Callers manually needed to clean up the `Buffer` afterwards
  using `ddprof_ffi_Buffer_free`**.

After discussing this with @morrisonlevi, we decided to go the
other way: change the `File` to contain a `ByteSlice` directly.

This avoids the extra copy in (1.), as well as the caller needing
to manage the `Buffer` objects manually in (2.).

This is a breaking API change for libddprof FFI users.

* Fix exporter.cpp example compilation by enabling C++ 11

* Update exporter.cpp example with new `File` struct API

* Use target_compile_features instead of cmake_cxx_standard to enable c++11

* Update examples/ffi/exporter.cpp

Minor cleanup as suggested during review.

Co-authored-by: Nicolas Savoire <[email protected]>

Co-authored-by: Nicolas Savoire <[email protected]>