Google Groups no longer supports new Usenet posts or subscriptions. Historical content remains viewable.
Dismiss

IPC::Open3 : Why can't I catch program output here?

1 view
Skip to first unread message

Ronny

unread,
May 5, 2008, 9:12:36 AM5/5/08
to
I try to use the program below to catch standard output and standard
error
of a process. The program itself runs well, and it *does* write at
least to
standard output, but in the code below, the line saying "### READING"
is never printed and @msg is still empty afterwards. This means that
$selector->can_read() must have returned false already the first time
it was called. Any suggestion on what could I have done wrong here?

my @msg;
my $pid=open3(*CMD_IN,*CMD_OUT,*CMD_ERR,@cmd)
or confess "Open3 error: $!\n";;
$SIG{CHLD} = sub {
waitpid($pid,0)>0 &&
print "SIGCHILD status=$? pid=$pid\n";
};
my $selector=IO::Select->new();
$selector->add(*CMD_ERR,*CMD_OUT);
while(my @ready=$selector->can_read()) {
foreach my $fh (@ready) {
print "### READING\n";
push @msg,scalar(<$fh>);
$selector->remove($fh) if eof($fh);
}
}
close CMD_OUT;
close CMD_ERR;

Ben Morrow

unread,
May 5, 2008, 6:55:27 PM5/5/08
to

Quoth Ronny <ro.nald...@gmail.com>:

> I try to use the program below to catch standard output and standard
> error of a process. The program itself runs well, and it *does* write
> at least to standard output, but in the code below, the line saying
> "### READING" is never printed and @msg is still empty afterwards.
> This means that $selector->can_read() must have returned false already
> the first time it was called. Any suggestion on what could I have done
> wrong here?

Works for me here (i686-freebsd). What platform are you on, and what
command are you trying to run? In general, if you can provide a complete
program people can copy/paste/run, you're more likely to get help: in
this case, you omitted the 'use' lines and the contents of @cmd.

> my @msg;
> my $pid=open3(*CMD_IN,*CMD_OUT,*CMD_ERR,@cmd)
> or confess "Open3 error: $!\n";;

I would normally say 'Use lexical filehandles!' at this point;
unfortunately, IPC::Open3 was written before they existed and the
obvious way

my $pid = open3(my $CMD_IN, my $CMD_OUT, my $CMD_ERR, @cmd)...

doesn't work (and can't be made to since undef is already meaningful).
So you have to fall back to the old method of using Symbol::gensym:

use Symbol qw/gensym/;

my ($CMD_IN, $CMD_OUT, $CMD_ERR) = (gensym, gensym, gensym);
my $pid = open3($CMD_IN, $CMD_OUT, $CMD_ERR, @cmd)...

which is definitely ugly, but IMHO worth it (if you must use IPC::Open3
at all: I'd use IPC::Run instead). For one thing, these gensymmed
filehandles close when they go out of scope, just like more usual
lexical filehandles. Of course, I'd use a hash to hold the filehandles:

my %CMD = map {($_, gensym)} qw/IN OUT ERR/;
my $pid = open3(@CMD{qw/IN OUT ERR/}, @cmd)...

> $SIG{CHLD} = sub {
> waitpid($pid,0)>0 &&
> print "SIGCHILD status=$? pid=$pid\n";
> };
> my $selector=IO::Select->new();
> $selector->add(*CMD_ERR,*CMD_OUT);
> while(my @ready=$selector->can_read()) {
> foreach my $fh (@ready) {
> print "### READING\n";
> push @msg,scalar(<$fh>);
> $selector->remove($fh) if eof($fh);

Using eof is nearly always a mistake: it's almost always better to check
the return value of your read statement instead. Something like

my $msg = <$fh>;
defined $msg
? push @msg, $msg
: $selector->remove($fh);

This also means you don't get a spurious undef on the end of @msg.

Ben

--
I must not fear. Fear is the mind-killer. I will face my fear and
I will let it pass through me. When the fear is gone there will be
nothing. Only I will remain.
b...@morrow.me.uk Frank Herbert, 'Dune'

Ilya Zakharevich

unread,
May 5, 2008, 8:26:34 PM5/5/08
to
[A complimentary Cc of this posting was NOT [per weedlist] sent to
Ben Morrow
<b...@morrow.me.uk>], who wrote in article <vrf4f5-...@osiris.mauzo.dyndns.org>:

> I would normally say 'Use lexical filehandles!' at this point;
> unfortunately, IPC::Open3 was written before they existed and the
> obvious way
>
> my $pid = open3(my $CMD_IN, my $CMD_OUT, my $CMD_ERR, @cmd)...
>
> doesn't work (and can't be made to since undef is already meaningful).

I have no idea what you mean here.

sub open3 ($$$$@) {
$_[0] = IO::Handle->new unless defined $_[0]; # Or whatever is THE initializer
...
}

Yours,
Ilya

Darren Dunham

unread,
May 6, 2008, 12:01:19 PM5/6/08
to
Ben Morrow <b...@morrow.me.uk> wrote:
> I would normally say 'Use lexical filehandles!' at this point;
> unfortunately, IPC::Open3 was written before they existed and the
> obvious way
>
> my $pid = open3(my $CMD_IN, my $CMD_OUT, my $CMD_ERR, @cmd)...
>
> doesn't work (and can't be made to since undef is already meaningful).

What's wrong with the above?

The only issue I see is that Open3 will not autogenerate a filehandle
for stderr (instead it will combine stdout and stderr and place them
both on the $CMD_OUT filehandle). I actually use that method so I don't
have to select between two filehandles and can just read one at a time.

# capture STDOUT and STDERR, on a single filehandle
use IPC::Open3;
open3(my $cmd_in, my $cmd_out, undef, @cmd);
while (<$cmd_out>)
{
...
}

--
Darren Dunham ddu...@taos.com
Senior Technical Consultant TAOS http://www.taos.com/
Got some Dr Pepper? San Francisco, CA bay area
< This line left intentionally blank to confuse you. >

Ben Morrow

unread,
May 6, 2008, 12:00:19 PM5/6/08
to

Quoth Ilya Zakharevich <nospam...@ilyaz.org>:

Yeah, in general you can do that; however, for the specific case of
open3, the docs say

If CHLD_ERR is false, or the same file descriptor as CHLD_OUT, then
STDOUT and STDERR of the child are on the same filehandle.

so at least the CHLD_ERR filehandle can't be autovivified. Presumably it
is felt that this behaviour can't be changed.

Ben

--
"If a book is worth reading when you are six, * b...@morrow.me.uk
it is worth reading when you are sixty." [C.S.Lewis]

xho...@gmail.com

unread,
May 6, 2008, 12:07:52 PM5/6/08
to
ddu...@taos.com (Darren Dunham) wrote:
> Ben Morrow <b...@morrow.me.uk> wrote:
> > I would normally say 'Use lexical filehandles!' at this point;
> > unfortunately, IPC::Open3 was written before they existed and the
> > obvious way
> >
> > my $pid = open3(my $CMD_IN, my $CMD_OUT, my $CMD_ERR, @cmd)...
> >
> > doesn't work (and can't be made to since undef is already meaningful).
>
> What's wrong with the above?

It causes cmd's stderr to go to $CMD_OUT rather than the obviously
intended $CMD_ERR.

>
> The only issue I see is that Open3 will not autogenerate a filehandle
> for stderr (instead it will combine stdout and stderr and place them
> both on the $CMD_OUT filehandle).

And this is obviously a problem if that is not what you want to happen.

Xho

--
-------------------- http://NewsReader.Com/ --------------------
The costs of publication of this article were defrayed in part by the
payment of page charges. This article must therefore be hereby marked
advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate
this fact.

Darren Dunham

unread,
May 6, 2008, 5:37:56 PM5/6/08
to
xho...@gmail.com wrote:
> ddu...@taos.com (Darren Dunham) wrote:
>> Ben Morrow <b...@morrow.me.uk> wrote:
>> > I would normally say 'Use lexical filehandles!' at this point;
>> > unfortunately, IPC::Open3 was written before they existed and the
>> > obvious way
>> >
>> > my $pid = open3(my $CMD_IN, my $CMD_OUT, my $CMD_ERR, @cmd)...
>> >
>> > doesn't work (and can't be made to since undef is already meaningful).
>>
>> What's wrong with the above?
>
> It causes cmd's stderr to go to $CMD_OUT rather than the obviously
> intended $CMD_ERR.

I don't know that it's obviously intended. It took me quite a while to
realize that I even had an option to return output and error on a single
filehandle. I mention that explicitly because the OP might have been
having issues with selecting between the two, when reading them combined
may have been preferable.

However, if you do want them separated, then the other choices offered
seem quite verbose. Instead of them, I would pass in a (defined)
filehandle. That takes one more line of code (two if you count bringing
in the FileHandle module.

my $CMD_ERR = FileHandle->new();
my $pid = open3(my $CMD_IN, my $CMD_OUT, $CMD_ERR, @cmd)...

Ben Morrow

unread,
May 6, 2008, 6:59:27 PM5/6/08
to

Quoth ddu...@taos.com (Darren Dunham):

> xho...@gmail.com wrote:
> > ddu...@taos.com (Darren Dunham) wrote:
> >> Ben Morrow <b...@morrow.me.uk> wrote:
> >> > I would normally say 'Use lexical filehandles!' at this point;
> >> > unfortunately, IPC::Open3 was written before they existed and the
> >> > obvious way
> >> >
> >> > my $pid = open3(my $CMD_IN, my $CMD_OUT, my $CMD_ERR, @cmd)...
> >> >
> >> > doesn't work (and can't be made to since undef is already meaningful).
> >>
> >> What's wrong with the above?
> >
> > It causes cmd's stderr to go to $CMD_OUT rather than the obviously
> > intended $CMD_ERR.
>
> I don't know that it's obviously intended. It took me quite a while to
> realize that I even had an option to return output and error on a single
> filehandle. I mention that explicitly because the OP might have been
> having issues with selecting between the two, when reading them combined
> may have been preferable.
>
> However, if you do want them separated, then the other choices offered
> seem quite verbose. Instead of them, I would pass in a (defined)
> filehandle. That takes one more line of code (two if you count bringing
> in the FileHandle module.
>
> my $CMD_ERR = FileHandle->new();
> my $pid = open3(my $CMD_IN, my $CMD_OUT, $CMD_ERR, @cmd)...

...which is essentially the same as what I proposed, except that I
prefer to avoid FileHandle (and IO::Handle) and use Symbol::gensym
directly. That's just a matter of taste, of course.

Ben

Ilya Zakharevich

unread,
May 6, 2008, 8:04:49 PM5/6/08
to
[A complimentary Cc of this posting was NOT [per weedlist] sent to
Ben Morrow
<b...@morrow.me.uk>], who wrote in article <jtb6f5-...@osiris.mauzo.dyndns.org>:

> > sub open3 ($$$$@) {
> > $_[0] = IO::Handle->new unless defined $_[0]; # Or whatever is THE
> > initializer
> > ...
> > }

> Yeah, in general you can do that; however, for the specific case of
> open3, the docs say
>
> If CHLD_ERR is false, or the same file descriptor as CHLD_OUT, then
> STDOUT and STDERR of the child are on the same filehandle.

Thanks, I forgot about this semantic. Which does not mean that one
could not write ALSO open3_autovivify... ;-)

Yours,
Ilya

0 new messages