This is the mail archive of the cygwin mailing list for the Cygwin project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: Buffered output/Forked processes


Nuno Leitao schrieb:
Sorry, I should have stated my question better -- look at the following command on cygwin:

% vmstat 1 > somefile.txt &
(...time passes...)
% ls -la somefile.txt
-rw-r--r-- 1 leitaon mkgroup- 0 Oct 13 13:48 somefile.txt
(...time passes...)
% ls -la somefile.txt
-rw-r--r-- 1 leitaon mkgroup- 4096 Oct 13 13:52 somefile.txt
What is going on is that vmstat/cygwin/glibc/whatever is just buffering the output until the buffer is full (4096 bytes) at which point it flushes it. The effect I want is for each line outputed from vmstat to go straight into the file (or in my original example, into the pipe and the Perl script) -- the problem is *not* on the Perl script but the way vmstat is buffering the output.

yes, our glibc (newlib) buffers.
most libc implementations of printf do buffer. Use might want to use syswrite instead.


So you have to fix vmstat to remove the output buffering to stdout.
Can we see this script? Is vmstat perl?
Are you sure the output from vmstat is not buffered?
Or do say the cygwin pipe buffers implicitly?
  open( VMSTAT, "vmstat 5|" );

I've developed some tail -f like perl scripts like this to redirect live log data to a webbrowser and it worked ok on cygwin as on linux.

But I did it like this: (you'll get the idea)
check size, wait some sec, and print interim buffer if size changed.

open(LOG, $log) or die "can't open $log: $!\n$usage";
seek(LOG, 0, SEEK_END);
for (;;) {
  $_ = readline(*LOG);
  if ($grep) {
	# die at syntax errors.
        # highlight?
	eval {
	  if (/$grep/i) {
	    s|($grep)|<span class="high">$1</span>|i;
	    print $_, "<br />";
	  }
	};
	die $@ if $@;
    } else {
	print $_, "<br />";
    }
    for (my $size = -s LOG; $size == -s LOG; $i+=$sleep, sleep $sleep) {
      if ($i > $timeout) {
	print "\n<br /><font color=\"blue\">timeout</font><br />\n";
	exit;
      }
    }
    LOG->clearerr();
}

-----Original Message-----
From: Reini Urban [mailto:rurban@x-ray.at]
Sent: 13 October 2004 13:12
To: Nuno Leitao
Cc: 'cygwin@cygwin.com'
Subject: Re: Buffered output/Forked processes


Nuno Leitao schrieb:
I have a Perl script which looks like:

 open( VMSTAT, "vmstat 5|" );
 VMSTAT->autoflush( 1 );
 while( <VMSTAT>) {
    print $_;
 }

 Now, under Linux and other UNIX OS's this works fine, and "print $_" will
print the vmstat output every 5 seconds since the output from vmstat is not
buffered. Under Cygwin however, it seems vmstat will always buffer its
stdout with the undesirable effect that lines come in batches as the output
buffer gets full and is flushed by the Cygwin C libraries.

 Is there a way to go around this without having to patch and recompile
vmstat or other binaries I might want to use in this way ?


basic perl question, not cygwin related.
$ perldoc -q buffer
$ perldoc -f select

cygwin perl doesn't honor $| ? not true.
Because you told VMSTAT not to buffer.
But you shold have told it to STDOUT.

open( VMSTAT, "vmstat 5|" );
VMSTAT->autoflush( 1 );
$| = 1; # STDOUT is selected so STDOUT will get unbuffered.
while( <VMSTAT>) {
     print $_;
}
--
Reini Urban
http://xarch.tu-graz.ac.at/home/rurban/

--
Unsubscribe info:      http://cygwin.com/ml/#unsubscribe-simple
Problem reports:       http://cygwin.com/problems.html
Documentation:         http://cygwin.com/docs.html
FAQ:                   http://cygwin.com/faq/


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]