Re: [PATCH bpf-next] selftests/bpf: add --json-summary option to test_progs

[Eduard Zingerman <eddyz87@gmail.com>]

On Thu, 2023-03-16 at 09:38 -0700, Manu Bretelle wrote:
> [...]
>
> I was originally going to do a nested structure similar to this too
> (minus the repeat of test_* entries for subtests. But while discussing this
> offline with Andrii, a flatter structured seemed to be easier to parse/manage
> with tools such as `jq`. I am also very probably missing the right
> incantation for `jq`.
> 
> Finding whether a test has subtests (currently only adding failed ones,
> but this could change in the future) would be easier (essentially checking
> length(subtests)). But neither is it difficult to reconstruct using
> higher level language.
> 

`jq` query is a bit more complicated with nested structure, but not terribly so:

  $ cat query.jq
  .results | map([
                  .test_name,
                  (.subtests | map([([.test_name, .subtest_name] | join("/")) ]))
                 ])
           | flatten

  $ jq -f query.jq test.json
  [
    "test_global_funcs",
    "test_global_funcs/global_func16"
  ]

Test data for reference:

  $ cat test.json | sed -r 's/"[^"]{20,}"/"..."/g'
  {
      "success": 1,
      "success_subtest": 24,
      "skipped": 0,
      "failed": 1,
      "results": [{
              "test_name": "test_global_funcs",
              "test_number": 223,
              "message": "...",
              "failed": true,
              "subtests": [{
                      "test_name": "test_global_funcs",
                      "subtest_name": "global_func16",
                      "test_number": 223,
                      "subtest_number": 16,
                      "message": "...",
                      "is_subtest": true,
                      "failed": true
                  }
              ]
          }
      ]
  }

> In term of logical structure and maybe extensibility, this is more appropriate,
> in term of pragmatism maybe less.
> 
> I don't have strong opinions and can see benefit for both.

idk, I don't have a strong opinion either.

> [...]