Re: Avoid unecessary MemSet call (src/backend/utils/cache/relcache.c)

Поиск
Список
Период
Сортировка
От Ranier Vilela
Тема Re: Avoid unecessary MemSet call (src/backend/utils/cache/relcache.c)
Дата
Msg-id CAEudQArc+NxrubSKdz0khByqkyUT0-h9APtRgOYAntL1an4jDg@mail.gmail.com
обсуждение исходный текст
Ответ на Re: Avoid unecessary MemSet call (src/backend/utils/cache/relcache.c)  (Ranier Vilela <ranier.vf@gmail.com>)
Ответы Re: Avoid unecessary MemSet call (src/backend/utils/cache/relcache.c)  (Justin Pryzby <pryzby@telsasoft.com>)
Re: Avoid unecessary MemSet call (src/backend/utils/cache/relcache.c)  (Tom Lane <tgl@sss.pgh.pa.us>)
Re: Avoid unecessary MemSet call (src/backend/utils/cache/relcache.c)  (Alvaro Herrera <alvherre@alvh.no-ip.org>)
Список pgsql-hackers
Дерево обсуждения
Em ter., 17 de mai. de 2022 às 10:33, Ranier Vilela <ranier.vf@gmail.com> escreveu:
Em seg., 16 de mai. de 2022 às 20:26, David Rowley <dgrowleyml@gmail.com> escreveu:
On Sun, 15 May 2022 at 09:47, Ranier Vilela <ranier.vf@gmail.com> wrote:
> At function load_relcache_init_file, there is an unnecessary function call,
> to initialize pgstat_info pointer to NULL.
>
> MemSet(&rel->pgstat_info, 0, sizeof(rel->pgstat_info));

What seems to have happened here is the field was changed to become a
pointer in 77947c51c.  It's not incorrect to use MemSet() to zero out
the pointer field.  What it does probably do is confuse the casual
reader into thinking the field is a struct rather than a pointer to
one.   It's probably worth making that consistent with the other
fields so nobody gets confused.

Can you add a CF entry for PG16 for this so we come back to it after we branch?
Of course.
I will add it.
However, I would like to add more.
I found, I believe, a serious problem of incorrect usage of the memset api.
Historically, people have relied on using memset or MemSet, using the variable name as an argument for the sizeof.
While it works correctly, for arrays, when it comes to pointers to structures, things go awry.

#include <stdio.h>

struct test_t
{
     double b;
     int a;
     char c;
};

typedef struct test_t Test;

int main()
{
     Test * my_test;
   
     printf("Sizeof pointer=%u\n", sizeof(my_test));
     printf("Sizeof struct=%u\n", sizeof(Test));
}

Output:
Sizeof pointer=8
Sizeof struct=16
 
So throughout the code there are these misuses.

So, taking advantage of this CF I'm going to add one more big patch, with suggestions to fix the calls.
This pass vcregress check.

regards,
Ranier Vilela
Вложения

В списке pgsql-hackers по дате отправления:

Предыдущее
От: Tom Lane
Дата:
Сообщение: Re: Limiting memory allocation
Следующее
От: Justin Pryzby
Дата:
Сообщение: Re: Avoid unecessary MemSet call (src/backend/utils/cache/relcache.c)