From 9fd52d5705fad70c0cb4ad8d508596a488262acf Mon Sep 17 00:00:00 2001

From: =?UTF-8?q?Petr=20P=C3=ADsa=C5=99?= <ppisar@redhat.com>

Date: Mon, 13 Jul 2020 09:24:35 +0200

Subject: [PATCH] Use kibibytes instead of kilobytes in a documentation

MIME-Version: 1.0

Content-Type: text/plain; charset=UTF-8

Content-Transfer-Encoding: 8bit

This patch does not changes any output of the time program. It only

clarifies a manual and the comments.

<https://lists.gnu.org/archive/html/bug-time/2020-07/msg00000.html>

Signed-off-by: Petr Písař <ppisar@redhat.com>

---

 configure.ac          |  6 +++---

 doc/time.texi         | 12 ++++++------

 src/rusage-kb.c       |  4 ++--

 src/rusage-kb.h       | 10 +++++-----

 src/time.c            |  6 +++---

 tests/time-max-rss.sh |  4 ++--

 6 files changed, 21 insertions(+), 21 deletions(-)

diff --git a/configure.ac b/configure.ac

index d2950bd..67738b5 100644

--- a/configure.ac

+++ b/configure.ac

@@ -90,7 +90,7 @@ if test -z "$time_getrusage_mem_units" ; then

     solaris*)        time_getrusage_mem_units=pages ;;

-    # As a fallback, assume KB (the most common value).

+    # As a fallback, assume KiB (the most common value).

     # Set the 'warn' variable to warn the user at the end

     # of ./configure

     *) time_getrusage_mem_units=kb

@@ -104,7 +104,7 @@ case $time_getrusage_mem_units in

   kb)

      AC_DEFINE([GETRUSAGE_RETURNS_KB],[1],

                 [Define to 1 if getrusage(2) on this systems returns

-                 ru_maxrss in kilobytes])

+                 ru_maxrss in kibibytes])

      ;;

   bytes)

@@ -140,7 +140,7 @@ AC_OUTPUT

 # Warn the user if getrusage(2) behaviour on this OS is unknown

 if test "$warn_getrusage_mem_units" ; then

   AC_MSG_WARN([unknown getrusage behavior on operating system '$host_os'.

-               Assuming Kilobytes.

+               Assuming kibibytes.

                please report this with the output of 'uname -a' to

                bug-time@gnu.org])

 fi

diff --git a/doc/time.texi b/doc/time.texi

index dac65b4..3a05ed9 100644

--- a/doc/time.texi

+++ b/doc/time.texi

@@ -241,22 +241,22 @@ times divied by the total running time.

 @table @code

 @item M

 Maximum resident set size of the process during its lifetime, in

-Kilobytes.

+kibibytes.

 @item t

-Average resident set size of the process, in Kilobytes.

+Average resident set size of the process, in kibibytes.

 @item K

-Average total (data+stack+text) memory use of the process, in Kilobytes.

+Average total (data+stack+text) memory use of the process, in kibibytes.

 @item D

-Average size of the process's unshared data area, in Kilobytes.

+Average size of the process's unshared data area, in kibibytes.

 @item p

-Average size of the process's unshared stack, in Kilobytes.

+Average size of the process's unshared stack, in kibibytes.

 @item X

-Average size of the process's shared text, in Kilobytes.

+Average size of the process's shared text, in kibibytes.

 @item Z

 System's page size, in bytes.  This is a per-system constant, but

diff --git a/src/rusage-kb.c b/src/rusage-kb.c

index aad06b8..04352cd 100644

--- a/src/rusage-kb.c

+++ b/src/rusage-kb.c

@@ -25,8 +25,8 @@

 #include <limits.h>

 #include <unistd.h>

-/* Return the number of kilobytes corresponding to a number of pages PAGES.

-   (Actually, we use it to convert pages*ticks into kilobytes*ticks.)

+/* Return the number of kibibytes corresponding to a number of pages PAGES.

+   (Actually, we use it to convert pages*ticks into kibibytes*ticks.)

    Try to do arithmetic so that the risk of overflow errors is minimized.

    This is funky since the pagesize could be less than 1K.

diff --git a/src/rusage-kb.h b/src/rusage-kb.h

index 4a53515..b7b1155 100644

--- a/src/rusage-kb.h

+++ b/src/rusage-kb.h

@@ -19,7 +19,7 @@

 #ifndef _RUSAGE_KB_

 #define _RUSAGE_KB_

-/* As of 2017, most kernels' getrusage(2) returns ru_maxrss in kilobytes:

+/* As of 2017, most kernels' getrusage(2) returns ru_maxrss in kibibytes:

       Linux, Hurd, Free/Open/Net-BSD, MINIX, AIX7

    OpenSolaris's getrusage(2) documents a return value in pages,

@@ -32,20 +32,20 @@

 #if GETRUSAGE_RETURNS_KB

-/* define as no-op, as RUSAGE values are already in KB */

+/* define as no-op, as RUSAGE values are already in KiB */

 #define RUSAGE_MEM_TO_KB(x) (x)

 #elif GETRUSAGE_RETURNS_BYTES

-/* Convert bytes to kilobytes */

+/* Convert bytes to kibibytes */

 #define RUSAGE_MEM_TO_KB(x) ((x)/1024)

 #elif GETRUSAGE_RETURNS_PAGES

-/* Convert bytes to kilobytes */

+/* Convert bytes to kibibytes */

 #define RUSAGE_MEM_TO_KB(x) (ptok (x))

-/* A function to get the system's page size and convert pages to KB */

+/* A function to get the system's page size and convert pages to KiB */

 unsigned long

 ptok (unsigned long pages);

diff --git a/src/time.c b/src/time.c

index 7e07995..f76265a 100644

--- a/src/time.c

+++ b/src/time.c

@@ -108,8 +108,8 @@ typedef RETSIGTYPE (*sighandler) ();

    and 100 on the sun4.

    Some manuals have an apparent error, claiming that units for average

-   sizes are kb*sec.  Judging by the contents of `struct rusage', it

-   looks like it should be kb*ticks, like on SunOS.  Ticks/sec seems

+   sizes are KiB*sec.  Judging by the contents of `struct rusage', it

+   looks like it should be KiB*ticks, like on SunOS.  Ticks/sec seems

    to be (empirically):

    50 Mt. Xinu

    250 Ultrix (mips)

@@ -412,7 +412,7 @@ linear_argv (argv)

    x == exit status of command

    Various memory usages are found by converting from page-seconds

-   to kbytes by multiplying by the page size, dividing by 1024,

+   to kibibytes by multiplying by the page size, dividing by 1024,

    and dividing by elapsed real time.

    FP is the stream to print to.

diff --git a/tests/time-max-rss.sh b/tests/time-max-rss.sh

index 0adda5c..5ecd3f2 100755

--- a/tests/time-max-rss.sh

+++ b/tests/time-max-rss.sh

@@ -27,7 +27,7 @@ fail=

 # The auxiliary program should be built and runnable

 time-aux || framework_failure_ "time-aux is missing/not runnable"

-# Get the baseline number of MAX-RSS kilobytes

+# Get the baseline number of MAX-RSS kibibytes

 # use by the program when not allocating any extra memory

 env time -o mem-baseline -f "%M" time-aux \

   || framework_failure_ "failed to run time/time-aux (baseline max-rss)"

@@ -49,7 +49,7 @@ test "$b" -eq "0" && test "$c" -eq 0 \

 # There could be alot of variation between each invocation,

 # accept a reasonable range

 if test "$d" -ge 5000 && test "$d" -le 6000 ; then

-    : # acceptable values: 5000-6000 KB

+    : # acceptable values: 5000-6000 KiB

 else

     cat<<EOF>&2

 time(1) failed to detect 5MB allcoation.

-- 

2.25.4