Re: [PATCH bpf-next v1 3/4] selftests/bpf: Test btf_dump rendering of decl_tags

[Eduard Zingerman <eddyz87@gmail.com>]

On Wed, 2026-09-16 at 18:20 -0700, Ihor Solodrai wrote:
> Cover the decl tag rendering the previous patch added. One subtest
> builds a BTF with every shape that renders, plus two that must not:
> 
>  - a named record with both forms of tag
>  - a typedef'd anonymous record, whose attribute precedes the declarator
>  - a union
>  - a record that also carries the derived packed attribute
>  - an anonymous record inlined at a member
>  - two records whose tags are added in reverse target order, which is
>    what pins the sort: without it the earlier record loses its attribute
>  - attributes on a var and on a func, which btf_dump never declares
> 
> A second subtest appends types and attributes to a dumper that has
> already run. Its last phase adds an attribute for a record emitted two
> dumps earlier and checks nothing is rendered for it: an attribute is
> part of the type, and btf_dump emits each definition once.
> 
> A third covers members. A member attribute and a record one share a key
> in the index, so it checks they do not leak into each other's position,
> over a function pointer, an array and a bit-field - declarators the
> attribute has to follow rather than precede.
> 
> A fourth covers typedefs, including a typedef of an anonymous record
> carrying three groups at once, binding to the member, the record and
> the typedef.
> 
> Switch test_ctx__dump_and_compare() to compare_text_to_expected()
> while we are here.
> 
> Signed-off-by: Ihor Solodrai <ihor.solodrai@linux.dev>
> ---

Acked-by: Eduard Zingerman <eddyz87@gmail.com>

>  .../selftests/bpf/prog_tests/btf_dump.c       | 343 +++++++++++++++++-
>  1 file changed, 342 insertions(+), 1 deletion(-)
> 
> diff --git a/tools/testing/selftests/bpf/prog_tests/btf_dump.c b/tools/testing/selftests/bpf/prog_tests/btf_dump.c
> index fe04a955d46c..6c031e83dad3 100644
> --- a/tools/testing/selftests/bpf/prog_tests/btf_dump.c
> +++ b/tools/testing/selftests/bpf/prog_tests/btf_dump.c
> @@ -224,7 +224,12 @@ static void test_ctx__dump_and_compare(struct test_ctx *t,
>  	fflush(t->dump_buf_file);
>  	t->dump_buf[t->dump_buf_sz] = 0; /* some libc implementations don't do this */
>  
> -	ASSERT_STREQ(t->dump_buf, expected_output, message);
> +	/*
> +	 * The mismatch has already been reported, so this only has to
> +	 * register the failure. ASSERT_OK() would append a stale errno to it.
> +	 */
> +	err = compare_text_to_expected(t->dump_buf, expected_output);
> +	ASSERT_EQ(err, 0, message);

Nit: Every caller of the compare_text_to_expected does such ASSERT,
     let's embed the assert in the function and avoid the above comment.

...

> +/*
> + * btf_dump indexes decl attributes incrementally, only walking types appended
> + * since the last dump. Check that attributes on types added to an existing
> + * dumper are picked up, and that one added for a record that has already been
> + * emitted is not - btf_dump emits each definition once.
> + */
> +static void test_btf_dump_decl_tags_incremental(void)

Nit: I'd drop this test or made it a part of an existing one (w/o the s3 part).

...