Or if your OS supports it, you could make a pipe and use /dev/fd/# (or the like) as the filename.

Thanks :)

This kind of works:

my @args=();
@args=("$pipeline/EXE/searchXY", "-5", "UEB2K005I", "locate", "234", "156", "4", "Print"); system(@args);

where earlier system "/usr/bin/mkfifo $pipeline/OUT/locate";

this sends it to the right place, but it gets the stuck at the process. Namely, the pipe is open for writing, and waits to be read, the C program gets paused until something reads the pipe... no problem so far :) This is as it should be from what I managed to understand about named pipes.

One more minor problem is that my whole script gets stuck at the point of when the pipe is open, waiting to be read... as the reading part is after the pipe open (the call to the C program opens it).

I suspect I need to start off the C program as some independent process, NOT use system(). I'll post the solutions when I fully work it out. The KEY here were named pipes :)

THANKS ALL :) The suggestions were RIGHT what I needed.

You just need to fork before you call system, so that your system is running in its own process.

---

Caution: Contents may have been coded under pressure.

If your C programs follow this sort of convention, do a pipeline "open()" to launch the sub-process, and put "-" where the output file name should be. Then read from the pipeline file handle in the normal way, just like reading from any data file.

(If the C programs don't follow this convention, and you don't have / can't update the C source code, and you can't create a named pipe as suggested in an earlier reply, then you'll just have to put up with saving output to a named file, and reading that file as a separate step in your perl script.)

but i would suggest: change the C program to output to plain old STDOUT if you want to catch the output in a pipe ( for instance with a normal printf)

Plus Perl is used to "glue" other programs together, it is a well known approach, and this is how I use it in large parts... so it is 100% Perl related. I'm just curious if enybody has ever managed to do it? Or if they even heard of a way to do it?

#include <stdio.h>
#include <stdlib.h>
#define RETBUFF 512

int main(int argc, char** argv) {
  char* input[RETBUFF]; 
  while (fscanf(stdin,"%s",&input) != EOF) {   
    printf("%s",&input);
    fflush(stdout);
  } 
  return 0;
}
[download]

unless (defined ($pid = fork)) 
{
    die "cannot fork: $!";
}

unless ($pid)  
{
    my @args=();
    @args=("$pipeline/EXE/searchXY", "-5", "UEB2K005I", "locate", "234
+", "156", "4", "Print");
    system(@args);
    exit;                        # the child stops here
}

open(FIFO, "<$pipeline/OUT/locateXY");

my $line;
while ($line = <FIFO>)
{
    print "rep $line";
    #or whatever here
}
close(FIFO);
waitpid($pid, 0); #clean up
[download]

Now I just need a good reference/tutorial on fork to understand this properly :)

P.S. At some points you may want to add the two following lines, at appropriate places, before and after you use the above code, respectively:

system "/usr/bin/mkfifo $pipeline/OUT/locateXY";

system "chmod 777 $pipeline/OUT/locateXY"; #as needed

system "rm $pipeline/OUT/locateXY"; #to clean up if required
[download]