CVS log for src/usr.bin/systat/main.c

Up to [cvs.NetBSD.org] / src / usr.bin / systat

Request diff between arbitrary revisions

Keyword substitution: kv
Default branch: MAIN

Revert previous, and switch to using real arithmetic.

Ever since this code was first committed to the CSRG SCCS
tree, in Oct 1983, it has contained either
	int dellave;
or more recently:
	static int dellave;
(that change is irrelevant, as is the move of the declaration
from systat.h into main.c)

And then in main() initialised it as:
	dellave = 0.0;

It seems clear to me that the variable really should have been a
double since day one.   All its uses were in floating point contexts.

So, undo the change to use int constants, and instead just make
"dellave" (the load average delta) be a double.

This should make the load average bar which appears at the
top of most systat display screens more sensitive to showing
when the load average is increasing or decreasing (showing
it using '>' or '<' instead of '|' which is used when there
is not much change since last time).

systat: don't compare integer to floating point constants

No binary change.

PR/56331: Paul Goyette: Add -z option to display 0 instead of ' ' in vmstat.

Sync with HEAD

Pull up following revision(s) (requested by christos in ticket #1176):

	usr.bin/systat/main.c: revision 1.55
	usr.bin/systat/extern.h: revision 1.47
	usr.bin/systat/syscall.c: revision 1.10
	usr.bin/systat/iostat.c: revision 1.39
	usr.bin/systat/vmstat.c: revision 1.85
	usr.bin/systat/vmstat.c: revision 1.86

no need for curses.h

 -

PR/53906: Paul Ripke: systat(1) vmstat & syscall get
"alternate system clock has died" on terminal resize (SIGWINCH)

Sync with HEAD

PR/53906: Paul Ripke: systat(1) vmstat & syscall get
"alternate system clock has died" on terminal resize (SIGWINCH)

Sync with HEAD, resolve a few conflicts

Extend "systat vm" output to provide better insight about buffercache and
document it.

From Greg A. Woods in PR bin/36542

Pull up the following revisions(s) (requested by mrg in ticket #1639):
	usr.bin/systat/main.c:	revision 1.52 via patch

Add missing 'b' to the list of options.
Alpha-sort the option handling.

add -b to usage.

add missing 'b' to the optstring.  alpha-sort the option handling.

Pull up following revision(s) (requested by mrg in ticket #1491):
	usr.bin/systat/main.c: revision 1.51
	usr.bin/systat/systat.1: revision 1.47 via patch
add a basic "-b" option - that outputs one display and exits.
inspired by "top -b".  still requires a curses-capable output
without some fairly significant internal restructuring.

Sync with HEAD.  (Note that most of these changes are simply $NetBSD$
tag issues.)

add a basic "-b" option - that outputs one display and exits.
inspired by "top -b".  still requires a curses-capable output
without some fairly significant internal restructuring.

Sync with HEAD

allow fractional intervals

if you change display types, the "Showing %s, refresh every %d seconds."
message is retained, and you can't get rid of it.  since it's dispalyed
in inverse, it feels like there's always an error happening (which is
really what the code has -- the above string is displayed via the error()
function.)  for several types of key press that largely are just "redraw"
operation (space, ^L, etc.), clear the error string as well.

Rebase to HEAD as of a few days ago.

Rebase.

Always use_sysctl in preparation of removing all kvm code.

resync with head

sync with (a bit old) head

use more sysctl

Consistently use __dead and __printflike.

Allow display intervals > 25 seconds.
Fixes PR/36999
I did contemplate chaging curses - but the code is replicated and warped.

Don't assume that two identical-looking string literals will have the
same address and will therefore be comparable with the == operator.
Instead, use a const variable.
Inspired by PR 41255 from Kurt Lidl.

Sync with wrstuden-revivesa-base-2.

Remove the \n and tabs from the __COPYRIGHT() strings.
Tweak to use a consistent format.

Sync with netbsd-4.

Use sysctl to fetch network statistics unless a memory file is specified
on the command line.

Pull up following revision(s) (requested by christos in ticket #1031):
	usr.bin/systat/keyboard.c: revision 1.24
	usr.bin/systat/main.c: revision 1.41
	usr.bin/systat/extern.h: revision 1.38
don't call curses from a signal handler!

sync with HEAD

don't call curses from a signal handler!

use c99 initializers
consistently terminate nlist arrays with NULL not ""

WARNS=3

Add (unsigned char) cast to ctype functions
Although, IIRC, there is another PR on all commands being converted to
lower case.

Using timeout() to control refresh gets rid of a whole pile of signal-related
lossage.
Making ungetch() signal-safe, as ncurses seems to, would be better, though.

Use KEY_RESIZE.  Might as well have something in our source tree that does.

Move UCB-licensed code from 4-clause to 3-clause licence.

Patches provided by Joel Baker in PR 22365, verified by myself.

if kvm_nlist() fails, nlisterr() would be called before the screen
curses context would be initialized; just use errx() instead in this case
this fixes coredump for cases like 'systat -N /netbsd.gz', reported by Walt
on port-i386

in nlisterr(), wait a while (5 seconds currently) before exiting, so that
it would be actually possible to actually see the error

No need to resizeterm() if size not changed.
On 'fg' just restore old screen contents.

Reinstate our STOP handler before curses installs its one.
Ensures that both handlers are executed on subsequent stops.

Resize windows on SIGWINCH and restart from SIGTSTP
(approved by christos)

Change deprecated curses function to new equivalent.

file main.c was added on branch nathanw_sa on 2001-12-06 12:40:52 +0000

Change deprecated curses function to new equivalent.

Fetch maxslp from the kernel.  Only use USPACE if it's defined.

Remove extern variables and functions now declared in dkstats.h.
While here, remove 77 (!) unneed #includes.

Pull up the following revisions, approved by releng-1-5:

- Add "boot", "run", "time" and "zero" commands to all IP displays. These work
  as in the "vmstat" display mode. The default mode is now "time", not "boot".
- ANSIfy.
- s@kre\(@vmstat\(@.
- __RCSID() police.
- Tweak inet.icmp display visual a little.

icmp.c           1.5
iostat.c         1.16
ip.c             1.8
ip6.c            1.5, 1.7
ipsec.c          1.5
keyboard.c       1.12
main.c           1.26
mbufs.c          1.9
netcmds.c        1.16

Add "all" mode to systat to cycle between all modes,
sort of like VMS' "MONITOR ALL"-command.

Code was submitted by Jarkko Teppo <jate@uwasa.fi> in PR 10851,
cleaned up and added 'turns' handling by me.

- Add `boot', `run', `time' and `zero' commands to all IP displays. These work
  as in the `vmstat' display mode. The default mode is now `time', not `boot'.
- ANSIfy.
- s@kre\(@vmstat\(@.
- __RCSID() police.
- Tweak inet.icmp display visual a little.

Sync w/ netbsd-1-5-base.

Use hline() and variants in a bunch of places.

Two UI changes:

  a.) restore the capability to enter a substring of a mode or command
      name.  Pointed out by Soren Jorvang.

  b.) (capital) Q now exits systat immediately

add command line -n flag, which is same as "netstat numbers".
without it, we can't disable name resolution before the resolution starts.

Pull up to last week's -current.

Pull up revision 1.21 (requested by jwise):
  Guard against buffer overflow.

Block alarm signals while displaying -- there's no reason to allow recursive
display at all.

Knee-jerk snprintf()-ism.

Two things:

  1.) fix a bug which caused coredumps when starting with no mode arguments
  2.) move to table-lookup for mode-specific command parsing as well.
      all command matching is now table-driven.

Sigh...  this is the missing part of last night's final commit -- perform
the same simplification on mode parsing for the command line as for user
commands.

My apologies to anyone who tried to build systat this morning.

Rototill global command matching -- switch from ad-hoc tomfoolery to a
table lookup.  This will make prefix matching and so on easier.

Expect a similar change for mode-specific commands before too long.

While there, rename some structures with misleading names.

Watch this space for more changes soon.

Sync usage() with man page.

Pull up revision 1.16 (requested by he):
  LP64 and type sanity fixes, with visible effects of enabling echo
  on commands and correct cursor placement after update.

Functional LP64 and type sanity fixes.

Fix handling of privileges - the gid was given up before dkinit()
was called.
Pass the gid to use for KVM access explicitely to dkinit(), the caller
decides whether to call setgid(getgid()) or not.

- KNF
- use snprintf
- don't overwrite the keyboard buffer (and make this sized on
  the window size).

- use an array MAXHOSTNAMELEN+1 size to hold hostnames
- ensure hostname from gethostname() is nul-terminated in all cases
- minor KNF
- use MAXHOSTNAMELEN over various other values/defines
- be safe will buffers that hold hostnames

fix .Nm usage, deprecate register, getopt returns -1 not EOF, s/[br]*/[ms]*/

- some KNF.
- use __RCSID().
- WARNS=1

Message buffer passed to kvm_openfiles() should _POSIX2_LINE_MAX long; PR#3266

Freshen up a bit:

 - Replace deprecated sigblock() with sigprocmask()
 - Redraw the current display if SIGWINCH is caught
 - Style police

New userland code to read the statistics kept by the NetBSD generic
disk framework, from John M. Vinopal <banshee@gabriella.resort.com>

Be pedantic, main() returns int in C.

Print error message and exit if TERM is not set.
From PR #1015.

various includes, casts, and cleanups...

Use err*(3)/warn*(3).

Don't core dump on an illegal display type.  Display a more specific error
message for an ambiguous type.

Changed to conform to NetBSD's new RCS Id convention.

imported from 44lite

Initial revision