[PATCH] Renumber confusing value for GUC_UNIT_BYTE

Поиск
Список
Период
Сортировка
От Justin Pryzby
Тема [PATCH] Renumber confusing value for GUC_UNIT_BYTE
Дата
Msg-id 20220720145220.GJ12702@telsasoft.com
обсуждение исходный текст
Ответы Re: [PATCH] Renumber confusing value for GUC_UNIT_BYTE  (Peter Eisentraut <peter.eisentraut@enterprisedb.com>)
Список pgsql-hackers
The GUC units are currently defined like:

#define GUC_UNIT_KB                             0x1000  /* value is in kilobytes */
#define GUC_UNIT_BLOCKS                 0x2000  /* value is in blocks */
#define GUC_UNIT_XBLOCKS                0x3000  /* value is in xlog blocks */
#define GUC_UNIT_MB                             0x4000  /* value is in megabytes */
#define GUC_UNIT_BYTE                   0x8000  /* value is in bytes */
#define GUC_UNIT_MEMORY                 0xF000  /* mask for size-related units */

#define GUC_UNIT_MS                        0x10000      /* value is in milliseconds */
#define GUC_UNIT_S                         0x20000      /* value is in seconds */
#define GUC_UNIT_MIN               0x30000      /* value is in minutes */
#define GUC_UNIT_TIME              0xF0000      /* mask for time-related units */

0x3000 and 0x30000 seemed wrong, since they're a combination of other flags
rather than being an independant power of two.

But actually, these aren't flags:  they're tested in a "case" statement for
equality, not in a bitwise & test.

So the outlier is actually 0x8000, added at:
|commit 6e7baa322773ff8c79d4d8883c99fdeff5bfa679
|Author: Andres Freund <andres@anarazel.de>
|Date:   Tue Sep 12 12:13:12 2017 -0700
|
|    Introduce BYTES unit for GUCs.

It looks like that originated here:

https://www.postgresql.org/message-id/CAOG9ApEu8bXVwBxkOO9J7ZpM76TASK_vFMEEiCEjwhMmSLiaqQ%40mail.gmail.com

commit 162e4838103e7957cccfe7868fc28397b55ca1d7
Author: Justin Pryzby <pryzbyj@telsasoft.com>
Date:   Wed Jul 20 09:27:24 2022 -0500

    Renumber confusing value for GUC_UNIT_BYTE
    
    It had a power-of-two value, which looks right, and causes the other values
    which aren't powers-of-two to look wrong.  But this is tested for equality and
    not a bitwise test.
    
    See also:
    6e7baa322773ff8c79d4d8883c99fdeff5bfa679
    https://www.postgresql.org/message-id/CAOG9ApEu8bXVwBxkOO9J7ZpM76TASK_vFMEEiCEjwhMmSLiaqQ%40mail.gmail.com

diff --git a/src/include/utils/guc.h b/src/include/utils/guc.h
index 4d0920c42e2..be928fac881 100644
--- a/src/include/utils/guc.h
+++ b/src/include/utils/guc.h
@@ -219,11 +219,12 @@ typedef enum
 #define GUC_DISALLOW_IN_AUTO_FILE 0x0800    /* can't set in
                                              * PG_AUTOCONF_FILENAME */
 
+/* GUC_UNIT_* are not flags - they're tested for equality */
 #define GUC_UNIT_KB                0x1000    /* value is in kilobytes */
 #define GUC_UNIT_BLOCKS            0x2000    /* value is in blocks */
 #define GUC_UNIT_XBLOCKS        0x3000    /* value is in xlog blocks */
 #define GUC_UNIT_MB                0x4000    /* value is in megabytes */
-#define GUC_UNIT_BYTE            0x8000    /* value is in bytes */
+#define GUC_UNIT_BYTE            0x5000    /* value is in bytes */
 #define GUC_UNIT_MEMORY            0xF000    /* mask for size-related units */
 
 #define GUC_UNIT_MS               0x10000    /* value is in milliseconds */



В списке pgsql-hackers по дате отправления:

Предыдущее
От: Fujii Masao
Дата:
Сообщение: Remove useless arguments in ReadCheckpointRecord().
Следующее
От: Andres Freund
Дата:
Сообщение: Re: pgsql: Default to hidden visibility for extension libraries where possi