English is currently available.
Perl DISCUSSION
My Perl script calls an external flashing tool and then parses its output. I have seen three ways to run a program: system("tool --port $port"), backticks, and exec. With system I get the output on the screen but cannot capture it, and my check system(...) or die "failed"; dies when the tool succeeds.
What is the difference between the three, and how do I get both the output and the exit code reliably?
system runs the command, waits for it to finish and lets it write directly to your script's standard output and error. Its return value is the wait status, not the text. Backticks, or the equivalent qx{...}, also wait, but they capture the command's standard output and return it: one string in scalar context, a list of lines in list context. exec does not come back at all: it replaces the running Perl process with the new program, so any code after it runs only if the exec itself failed.
The inverted check is because success is 0. system returns the same value that it leaves in $?, a 16-bit status: the exit code is in the high byte, $? >> 8, and the low seven bits, $? & 127, hold the signal number if the program was killed. A tool that exits with code 2 gives a return value of 512, and 512 >> 8 is 2. A return of -1 means the program could not be started, with the reason in $!.
So write system(@cmd) == 0 or die "tool failed: exit " . ($? >> 8);. Backticks set $? the same way, so check it right after the capture.
Avoid building one command string from variables. A single string containing shell metacharacters is passed to the shell, so a port name or file name with a space, a semicolon or a quote changes the command. The list form system("tool", "--port", $port, $file) starts the program directly with exactly those arguments and no shell parsing.
Backticks have no list form. To capture output safely, open a pipe with a list: open(my $fh, "-|", "tool", "--port", $port) or die $!;, read lines from $fh, then close($fh) and look at $?.
Backticks capture only standard output. Error messages still go to the terminal unless you redirect them in the command, as in my $out = `tool --version 2>&1`;, which merges both streams into the result. If you need them separately, the core module IPC::Open3 gives you separate handles for the child's output and error streams.
exec is mainly for wrapper scripts that set up environment variables and then hand over to the real program, or for the child side after a fork. For "run it and parse the result", use the pipe open or backticks.
