[Tarantool-patches] [PATCH luajit v2] dbg: remove wrong GC state

[Sergey Kaplun]

Hi, Mikhail!
Thanks for the patch!
LGTM, after adding the comment for the testing snippet, see below.

On 16.10.26, Mikhail Elhimov wrote:
> It has been copied by mistake from the LuaVela fork of LuaJIT.
> ---
> Changes in v2:
>   * Rebased onto https://lists.tarantool.org/pipermail/tarantool-patches/2026-October/030859.html to avoid conflicts
> 
> This patch is to be applied after https://lists.tarantool.org/pipermail/tarantool-patches/2026-October/030859.html
> 
> Branch: https://github.com/tarantool/luajit/tree/elhimov/gh-noticket-dbg-remove-wrong-gcstate
> 
>  src/luajit_dbg.py                             |  3 +-
>  .../debug-extension-tests.py                  | 96 +++++++++++++++++--
>  2 files changed, 87 insertions(+), 12 deletions(-)
> 
> diff --git a/src/luajit_dbg.py b/src/luajit_dbg.py
> index a685ecfa..8f7249ab 100644
> --- a/src/luajit_dbg.py
> +++ b/src/luajit_dbg.py

<snipped>

> diff --git a/test/tarantool-debugger-tests/debug-extension-tests.py b/test/tarantool-debugger-tests/debug-extension-tests.py
> index bbd53dcc..735de2cb 100644
> --- a/test/tarantool-debugger-tests/debug-extension-tests.py
> +++ b/test/tarantool-debugger-tests/debug-extension-tests.py

<snipped>

> +class TestGCStateBase(TestCaseBase):
> +    extension_cmds = (
> +        'lj-state\n'
> +        'lj-gc\n'
> +    )
> +    lua_script = (
> +        'local obj = newproxy(true)\n'
> +        'getmetatable(obj).__gc = function() end\n'
> +        'obj = nil\n'
> +        'collectgarbage()\n'
> +    )

It would be nice to add the comment here explaining why __all__ GC
states are reachable through this Lua code:
1) `gc_mark_start()` (GCpause) is always present in the full GC cycle.
2) `propagatemark()` (GCSpropagate) is reachable because our build has
   FFI and the finalizer table is marked as a weak table, i.e. it is
   grey. Hence, the g->gc.gray list is not empty.
3) `atomic()` (GCSatomic) is always present in the full GC cycle.
4) `gc_sweep_str_chain()` (GCSsweepstring) is always present in the full
   GC cycle.
5) `gc_sweep()` (GCSsweep) is always present in the full GC cycle.
6) `gc_finalize()` is met due to the finalizer to userdata in the
   script.

> +
> +    @classmethod
> +    def setUpClass(cls):
> +        cls.pattern = (
> +            compose_lj_state_rx(cls.gc_state) +
> +            compose_gc_stats_rx(cls.gc_state)
> +        )
> +        super(TestGCStateBase, cls).setUpClass()
> +
> +
> +class TestGCStatePause(TestGCStateBase):
> +    gc_state = GCS_PAUSE
> +    location = 'gc_mark_start'
> +
> +
> +class TestGCStatePropagate(TestGCStateBase):
> +    gc_state = GCS_PROPAGATE
> +    location = 'propagatemark'
> +
> +
> +class TestGCStateAtomic(TestGCStateBase):
> +    gc_state = GCS_ATOMIC
> +    location = 'atomic'
> +
> +
> +class TestGCStateSweepString(TestGCStateBase):
> +    gc_state = GCS_SWEEPSTRING
> +    location = 'gc_sweep_str_chain'
> +
> +
> +class TestGCStateSweep(TestGCStateBase):
> +    gc_state = GCS_SWEEP
> +    location = 'gc_sweep'
> +
> +
> +class TestGCStateFinalize(TestGCStateBase):
> +    gc_state = 'FINALIZE'
> +    location = 'gc_finalize'
> +
> +
>  STACK_RX = (
>      r'-+ Red zone:\s+\d+ slots -+\n'
>      r'(' + RX_ADDR + r'\s+' + RX_FRAME + r' VALUE: nil\n?)*\n'
> -- 
> 2.43.0
> 

-- 
Best regards,
Sergey Kaplun