[net.bugs.4bsd] awk doesn't wait for subordinate processes to complete

jeff@fluke.UUCP (Jeff Stearns) (11/28/84)

Index:	 usr.bin/awk 4.2BSD

Description:
	If one redirects awk output through a filter, awk does not wait
	for the filter to terminate before awk itself terminates.  The
	output therefore appears asynchronously while subsequent commands
	are being processed.  In a shell script, this can be fatal if the
	filter is, for example, sort(1).

Repeat-By:
	% awk 'END { print "echo hi" | "sh" }' </dev/null
	
	Note that the result is (usually):
	    % awk 'END { print "echo hi" | "sh" }' </dev/null
	    % hi

	rather than:
	    % awk 'END { print "echo hi" | "sh" }' </dev/null
	    hi
	    %

	Try this, and note the changing size of the sorted-output file:	
	    % awk '{ print $0 | "sort -o sorted-output" }' <big-file
	    % wc sorted-output
	    % wc sorted-output
	    % wc sorted-output
	    % wc sorted-output
Fix:
	(Suggestion only)
	Before exiting, awk should call pclose(3) on each stream pointer
	which was created via a popen(3) in run.c.  The pclose(3) will
	wait(3) on the subordinate child before returning.
-- 
	Jeff Stearns       (206) 356-5064
	John Fluke Mfg. Co.
	P.O. Box C9090  Everett WA  98043  
	{uw-beaver,decvax!microsof,ucbvax!lbl-csam,allegra,ssc-vax}!fluke!jeff

jeff@fluke.UUCP (Jeff Stearns) (01/15/85)

Index: 	 usr.bin/awk/main.c 4.2BSD

Description:
	If one redirects awk output through a filter, awk does not wait
	for the filter to terminate before awk itself terminates.  The
	output therefore appears asynchronously while subsequent commands
	are being processed.  In a shell script, this can be fatal if the
	filter is, for example, sort(1).

Repeat-By:
	% awk 'END { print "echo hi" | "sh" }' </dev/null
	
	Note that the result is (usually):
	    % awk 'END { print "echo hi" | "sh" }' </dev/null
	    % hi

	rather than:
	    % awk 'END { print "echo hi" | "sh" }' </dev/null
	    hi
	    %

	Try this, and note the changing size of the sorted-output file:	
	    % awk '{ print $0 | "sort -o sorted-output" }' <big-file
	    % wc sorted-output
	    % wc sorted-output
	    % wc sorted-output
	    % wc sorted-output
Fix:
	The following context diff shows the additions made to the file
	/usr/src/usr.bin/awk/main.c:

*** /usr/src/usr.bin/awk/main.old	Mon Jan 14 15:51:47 1985
--- /usr/src/usr.bin/awk/main.c	Mon Jan 14 15:08:44 1985
***************
*** 91,93
  	run();
  	exit(errorflag);
  }
  
--- 91,102 -----
  	run();
+ 	/*
+ 	 *  Awk may have spawned some children (printf "....." | sort).
+ 	 *  The children may still be alive (sort can be quite slow).
+ 	 *  It is a bad idea to exit before our children have completed,
+ 	 *  as the next command to be executed may depend on ALL of our
+ 	 *  processing being complete.  Thus we wait for our kids.
+ 	 */
+ 	while (wait (0) != -1)
+ 		;
  	exit(errorflag);
  }

jeff@fluke.UUCP (Jeff Stearns) (01/15/85)

Index: 	 bin/awk/main.c 4.2BSD

Description:
	If one redirects awk output through a filter, awk does not wait
	for the filter to terminate before awk itself terminates.  The
	output therefore appears asynchronously while subsequent commands
	are being processed.  In a shell script, this can be fatal if the
	filter is, for example, sort(1).

Repeat-By:
	% awk 'END { print "echo hi" | "sh" }' </dev/null
	
	Note that the result is (usually):
	    % awk 'END { print "echo hi" | "sh" }' </dev/null
	    % hi

	rather than:
	    % awk 'END { print "echo hi" | "sh" }' </dev/null
	    hi
	    %

	Try this, and note the changing size of the sorted-output file:	
	    % awk '{ print $0 | "sort -o sorted-output" }' <big-file
	    % wc sorted-output
	    % wc sorted-output
	    % wc sorted-output
	    % wc sorted-output
Fix:
	The following context diff shows the additions made to the file
	/usr/src/bin/awk/main.c:

*** /usr/src/bin/awk/main.old	Mon Jan 14 15:51:47 1985
--- /usr/src/bin/awk/main.c	Mon Jan 14 17:15:56 1985
***************
*** 8,9
  #include "awk.h"
  #define TOLOWER(c)	(isupper(c) ? tolower(c) : c) /* ugh!!! */

--- 8,10 -----
  #include "awk.h"
+ #include <sys/param.h>
  #define TOLOWER(c)	(isupper(c) ? tolower(c) : c) /* ugh!!! */
***************
*** 89,91
  	run();
  	exit(errorflag);
  }

--- 92,107 -----
  	run();
+ 	/*
+ 	 *  Awk may have spawned some children (printf "....." | sort).
+ 	 *  The children may still be alive (sort can be quite slow).
+ 	 *  It is a bad idea to exit before our children have completed,
+ 	 *  as the next command to be executed may depend on ALL of our
+ 	 *  processing being complete.  Thus we wait for our kids.
+ 	 */
+ 	{
+ 		int i; for (i = 0; i < NOFILE; i++)
+ 			(void) close (i);
+ 		while (wait (0) != -1)
+ 			;
+ 	}
  	exit(errorflag);
  }
-- 
	Jeff Stearns       (206) 356-5064
	John Fluke Mfg. Co.
	P.O. Box C9090  Everett WA  98043  
	{uw-beaver,decvax!microsof,ucbvax!lbl-csam,allegra,ssc-vax}!fluke!jeff