Re: [PATCH 00/10] Mildly better refcount safety for Python

[Tom de Vries <tdevries@suse.de>]

On 2/21/26 3:31 AM, Simon Marchi wrote:
> 
> 
> On 2026-02-20 16:03, Tom Tromey wrote:
>> This series converts a number of gdb-to-Python converion functions to
>> return a gdbpy_ref<> rather than returning a PyObject* and then simply
>> documenting that a new reference is returned.
>>
>> The hope is that this will make refcount mistakes less likely.
>>
>> Regression tested on x86-64 Fedora 43.
>>
>> Signed-off-by: Tom Tromey <tromey@adacore.com>
> 
> I don't know if someone else wants to review this, but I think this is
> good.  I think it's good if the references are managed from the point
> where we get them from the Python API to the point where we release them
> to the Python API.
> 
> I gave it a quick look, since it's very mechanical and I probably
> wouldn't spot mistakes anyway.  But I asked Claude to review and it said
> it looked good.  It even noted that you fixed a ref leak in the symbol
> patch.
> 
> Approved-By: Simon Marchi <simon.marchi@efficios.com>

Hi,

I'd like to ask, is there anything holding back committing this patch 
series?

I'm working on a v3 of this ( 
https://sourceware.org/pipermail/gdb-patches/2026-May/227374.html ) 
patches series, and realized I'm touching the same code.

FWIW, I asked Gemini Pro 3.1 to review this patch series, and it found 
the fixed ref leak Simon mentions, and no problems.

[ It also reported another fixed ref leak, which turned out to be 
incorrect. ]

Thanks,
- Tom