I'm running a C executable from Python and this executable sometimes segfaults. When it segfaults the subprocess module does not return anything in stdout or stderr.
Sample code:
import subprocess
proc = subprocess.Popen(['./a.out'], stdin=subprocess.PIPE,
stdout=subprocess.PIPE, stderr=subprocess.PIPE)
out, err = proc.communicate()
print 'out: "%s"' % out
print 'err: "%s"' % err
print proc.returncode
The source for a.out
is:
int main() {
printf("hello world\n");
int x = 1 / 0;
}
The output of ./a.out
is:
hello world
Floating point exception
The output of the Python code is (in Linux, python 2.7):
out: ""
err: ""
-8
Is there a way to get the output of the executable even if it crashes?
A generic method to translate returncode to a string message, would also be nice.
The application is almost certainly not flushing its output buffers before it encounters the segmentation fault.
By default, the C stdio library configures buffering for
stdout
so that the buffers are flushed at every end of line, but only isstdout
is a tty. That's why you do see the output when you runa.out
from the command line.You cannot directly make the C program change this from the Python side. But what you can do is give it a tty to use for its
stdout
, namely a pseudo tty. The details of opening and setting up a pseudo terminal are much more complicated than setting up a regular pipe, but there are some modules that will help you: See Pexpect and this SO question, among others.Your C program is not flushing its output buffer. The easiest way around this problem is to change the output mode of STDOUT to unbuffered:
Now your program (which I edited to fix a typo) produces the correct output:
The return code is 0 on my Mac, confusingly, because programs that are terminated for a signal don't have a return code.