On 2025-05-02 09:02, Fujii Masao wrote:
> On 2025/05/02 2:27, Fujii Masao wrote:
>>
>>
>> On 2025/05/01 21:42, Robert Haas wrote:
>>> On Thu, May 1, 2025 at 3:53 AM Fujii Masao
>>> <masao.fujii@oss.nttdata.com> wrote:
>>>> Just idea, what do you think about adding a flag to indicate whether
>>>> ProcessLogMemoryContextInterrupt() is currently running? Then,
>>>> when a backend receives a signal and
>>>> ProcessLogMemoryContextInterrupt()
>>>> is invoked, it can simply return immediately if the flag is already
>>>> set
>>>> like this:
>>>
>>> I think that something like this could work, but you would need more
>>> than this. Otherwise, if the function errors out, the flag would
>>> remain permanently set.
>>
>> Yes, we need to either use PG_TRY()/PG_FINALLY() or handle the flag as
>> a global variable and reset it in the error handling path. I think
>> using
>> PG_TRY()/PG_FINALLY() is the simpler option.
>
> I've implemented the patch in that way. Patch attached.
Thank you for the patch!
I confirmed that with this patch applied, the process no longer crashes
even after applying the change Robert used to trigger the crash.
a small nitpick:
+ * requested repeatedly and rapidly, potentially
leading to infinite
It looks like there are two spaces between "requested" and "repeatedly".
--
Regards,
--
Atsushi Torikoshi
Seconded from NTT DATA GROUP CORPORATION to SRA OSS K.K.
