use ALIGN for Push::alignment in struct types

[bkietz]

This PR adds an override of the default Push::alignment(), which is incorrectly 1 for structs because they are byte arrays and this doesn't accurately reflect the type's alignment requirements.

fixes #8150

[google-cla]

Thanks for your pull request! It looks like this may be your first contribution to a Google open source project. Before we can look at your pull request, you'll need to sign a Contributor License Agreement (CLA).

View this failed invocation of the CLA check for more information.

For the most up to date status, view the checks section at the bottom of the pull request.

[bkietz]

@evgenyx00 @tustvold

[alamb]

I wonder if we should add a test to this PR to make sure there are no regressions

[bkietz]

I was not sure where to add such a test, but I'd be happy to. I opened apache/arrow-rs#6426 since that seemed like a good way to check for regressions (I didn't see any). @aardappel @CasperN any recommendations?

[aardappel]

@bkietz yes we generally should have tests :) Not too familiar with the Rust tests though, have a look.

[bkietz]

I've regenerated and added a test.

• struct_vec3_is_written_with_correct_alignment_in_table did not catch this bug before; adding a struct to a Table also goes through the Push trait. I could extend it (for example using a loop over multiple preceding strings to ensure at least one iteration will require padding for alignment)
• The pull request template suggests that building should also regenerate. However, AFAICT cmake will only regenerate for C++. Perhaps cmake should instead invoke generate_code.py?
• Could a reviewer please approve the CI workflows so we can see if I've missed anything?

[alamb]

Thank you @bkietz

I am not familiar with this codebase, but I did verify that the test you added fails without the code in this PR

To test I did:

cd flatbuffers/tests/rust_usage_test
cargo test --test integration_test

The test fails like this without the PR

---- flatbuffers_tests::push_impls::push_u8_generated_struct_alignment stdout ----
thread 'flatbuffers_tests::push_impls::push_u8_generated_struct_alignment' panicked at tests/integration_test.rs:2360:9:
assertion `left == right` failed
  left: [10, 0, 20, 0, 0, 1]
 right: [10, 0, 20, 0, 1]

Seem like an improvement to me (the 1u8 is now aligned at a byte boundary rather than a u16 boundary)

[bkietz]

@dbaileychess could you review and/or approve the CI workflows?

[bkietz]

@aardappel whom can we ask to review this?

[aardappel]

I'd say either @CasperN or @dbaileychess

[bkietz]

Thanks!

[bkietz]

@CasperN could you review please?

[alamb]

Any update here @aardappel @CasperN or @dbaileychess ? We are waiting on this fix to enable some cross implementation testing in Apache Arrow

[aardappel]

This looks fine to me and can probably be merged, but I am not qualified to review Rust really.. is going to need someone else to review.

[RuidongWu]

Well, I have also encountered a similar aligned problem of struct. The generated data from rust cannot be correctly verified by C/C++, which reports unaligend data error.
If this struct is filled into a table, the generated data may be unaligned in binary data when using rust builder.
such as:

struct Something {
value: uint64;
}

table Anything {
name: string;
}

table OneTable {
names: [Anything];
values: Something;
}

[alamb]

Well, I have also encountered a similar aligned problem of struct. The generated data from rust cannot be correctly verified by C/C++, which reports unaligend data error.

Hi @RuidongWu - thank you for the report. Does this PR fix the problem you are seeing? Or are you reporting a different problem?

[RuidongWu]

@alamb Thans for your reply. Yes, this PR solves the unaligned problem of data-generated by rust. The key change is the file of src/idl_gen_rust.cpp. Thanks for the contribution of @bkietz

[aardappel]

Ok, I will merge this for now, any Rustaceans with complaints please follow up.

[alamb]

Thank you so much @aardappel 🙏