rpms/time
7
0
Fork 0
Code Issues Pull Requests Releases Activity

Clarify RSS size as kibibytes in a documentation

Browse Source
This commit is contained in:
Petr Písař 2020-07-13 09:37:47 +02:00
parent e8939fc57d
commit b3f1910a0d
2 changed files with 197 additions and 1 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

189
time-1.9-Use-kibibytes-instead-of-kilobytes-in-a-documentatio.patch Normal file
View File

@ -0,0 +1,189 @@
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

9
time.spec
View File

@ -1,7 +1,7 @@
Summary: A GNU utility for monitoring a program's use of system resources
Name: time
Version: 1.9
Release: 8%{?dist}
Release: 9%{?dist}
# src/time.c: GPLv3+
# COPYING: GPLv3 text
# doc/time.texi: GFDL
@ -46,6 +46,9 @@ Source: ftp://ftp.gnu.org/gnu/%{name}/%{name}-%{version}.tar.gz
Patch0: time-1.8-Prefer-clock_gettime-CLOCK_MONOTONIC.patch
# Fix info directory entry
Patch1: time-1.9-Improve-info-directory-index-entry-description.patch
# Clarify RSS size as kibibytes in a documentation, proposed to an upstream,
# <https://lists.gnu.org/archive/html/bug-time/2020-07/msg00000.html>
Patch2: time-1.9-Use-kibibytes-instead-of-kilobytes-in-a-documentatio.patch
BuildRequires: autoconf
BuildRequires: automake
BuildRequires: bash
@ -65,6 +68,7 @@ the results.
%setup -q
%patch0 -p1
%patch1 -p1
%patch2 -p1
autoreconf -fi
%build
@ -87,6 +91,9 @@ make %{?_smp_mflags} check
# time(1) manual page lives in man-pages package, bug #1612294.
%changelog
* Mon Jul 13 2020 Petr Pisar <ppisar@redhat.com> - 1.9-9
- Clarify RSS size as kibibytes in a documentation
* Fri Jan 31 2020 Fedora Release Engineering <releng@fedoraproject.org> - 1.9-8
- Rebuilt for https://fedoraproject.org/wiki/Fedora_32_Mass_Rebuild