bin/164535: ps(1) truncates command to screen size even when stdout is not a tty

Bruce Evans brde at optusnet.com.au
Fri Jan 27 11:49:42 UTC 2012


On Fri, 27 Jan 2012, Marcus Reid wrote:

>> Description:
> ps(1) truncates long commands to the size of the screen even when stdout is not a terminal.  This is counter-intuitive and differs from another implementation I looked at.  Output of ps | grep differs depending on how big your terminal window is for example.
>> How-To-Repeat:
> ps aux | grep
> ps auxww | grep
>> Fix:
> Patch included.  Tested; behavior remains consistent with docs: COLUMNS variable retains effect, and -w limits to 132 characters still.
>
>
> Patch attached with submission follows:
>
> --- bin/ps/ps.c.orig	2012-01-27 01:24:10.519024952 -0800
> +++ bin/ps/ps.c	2012-01-27 01:24:20.350023629 -0800
> @@ -187,6 +187,8 @@
>
> 	if ((cols = getenv("COLUMNS")) != NULL && *cols != '\0')
> 		termwidth = atoi(cols);
> +	else if (isatty(STDOUT_FILENO) == 0)
> +		termwidth = UNLIMITED;

This change defeats the else clause, which checks all of STDOUT_FILENO,
STDERR_FILENO and STDIN_FILENO and uses the terminal width iff any if
these is a terminal according to the TIOCGWINSZ test for being a
terminal, and otherwise gives a default of 79 columns (not UNLIMITED,
which is only documented for ps -ww and only reachable using that and
via the missing sanity checking for COLUMNS=0).

> 	else if ((ioctl(STDOUT_FILENO, TIOCGWINSZ, (char *)&ws) == -1 &&
> 	     ioctl(STDERR_FILENO, TIOCGWINSZ, (char *)&ws) == -1 &&
> 	     ioctl(STDIN_FILENO,  TIOCGWINSZ, (char *)&ws) == -1) ||

So the correct way to avoid terminalness when outputting to a file
is "ps </dev/null >foo 2>&1".  This is too strange for stdin, so
apparently checking all 3 fd's is a feature, precisely to get the
terminal width from somewhere even when stdout and stderr are
redirected to a file.  "COLUMNS=0 ps" is an easier way.  If this
feature is considered a bug, then remove the checks of STDERR_FILENO
and STDIN_FILENO.

Bruce


More information about the freebsd-bugs mailing list