Re: seawasp failing, maybe in glibc allocator

Поиск
Список
Период
Сортировка
От Andres Freund
Тема Re: seawasp failing, maybe in glibc allocator
Дата
Msg-id 20210522002510.oorongxox362vgvd@alap3.anarazel.de
обсуждение исходный текст
Ответ на Re: seawasp failing, maybe in glibc allocator  (Andres Freund <andres@anarazel.de>)
Ответы Re: seawasp failing, maybe in glibc allocator  (Thomas Munro <thomas.munro@gmail.com>)
Список pgsql-hackers
Hi,

On 2021-05-21 15:57:01 -0700, Andres Freund wrote:
> I found the LLVM commit to blame (c8fc5e3ba942057d6c4cdcd1faeae69a28e7b671).
> Contacting the author and reading the change to see if I can spit the
> issue myself.

Hrmpf. It's a silent API breakage. The author intended to email us about
it, but apparently forgot. One now needs to increment a string-pool
refcount. The reason that didn't trigger a reliable crash is that
there's a path where the refcount of string-pool entries aren't asserted
to be above before decrementing the refcount... And that there
practically never are references to the pool entries after use.

Continuing to discusss whether there's a better way to deal with this.

Greetings,

Andres Freund



В списке pgsql-hackers по дате отправления:

Предыдущее
От: Tom Lane
Дата:
Сообщение: Re: Subscription tests fail under CLOBBER_CACHE_ALWAYS
Следующее
От: Peter Smith
Дата:
Сообщение: Re: Refactor "mutually exclusive options" error reporting code in parse_subscription_options