perl: capturing STDOUT- logfile ends up binary

Question

I wrote a script that feeds data into an exe program called ctffind. The exe program outputs a bunch of data to the screen which I'm trying to capture in a logfile.

Here's what I'm using right now

my $logout = logfile 
open STDOUT, '>>', "$logout";
open my $PIPE1, '|-', '/opt/ctf/ctffind3_mp.exe' or die $!;

It works outputting everything that appeared on the screen into logfile.

If I more $logfile the file displays as expected it's about 5000 lines long and has about 50 lines that contain the string "final values". In my next step I do grep "final values" logfile it thinks logfile is a binary file and doesn't work instead it returns:

Binary file logfile matches

How do set it so that logfile is properly encoded? ctffind.exe is also generating binary files while it runs, is this part of the problem?

Answer 1

I have found two different ways to resolve the problem, thanks to commenters' suggestions:

1. One can change the grep logfile to grep -a logfile .

2. One can use strings logfile logfile2 to make a usable version of the logfile.

Answer 2

An Ascii NUL (\\000) in a file's first block is enough for Perl to call the file a "binary" one.

echo "hello world\000" > myfile
perl -E '$f=shift;open $fh,"<",$f or die;say -e $f && -B $f ? "binary":"text"' myfile

Answer 3

It's possible that ctffind3_mp.exe has code in it to do screen formatting such as changing colors, bolding, clearing the screen, etc. There are a few ways you could eliminate those:

• You can pipe it through cat -v which will turn all nonprintables into ASCII representations (eg NUL becomes ^@ ).
• There may be specific utilities for stripping ANSI or VT100 sequences out of text--try asking superuser.
• You could try setting $TERM (or in perl $ENV{'TERM'} ) to unknown in the hopes that the program no longer emits any special sequences for color, bold, etc.