有没有办法找出哪个 function 正在打印到终端？

Question

I was debugging a really complex function (python function that calls some C++ modules from Boost/Python), and there's a print or cout or another similar function that I forgot to delete (can't seem to find it with grep), and now the program just keeps printing a bunch of 0s. Is there a way to tell which function was printing to the terminal, or is deleting the print statement one by one the only way? I'm on Fedora linux if that helps

Thanks

Answer 1

I usually augment my complex designs with various forms of output tracing. The particular item that applies here is the trace_prefix. For instance:

TRACE = True
# TRACE = False    # Uncomment this line for production runs.

def func(param):
    trace_prefix = "TRACE func" if TRACE else ""
    if TRACE:
        print("ENTER", func, "param =", param)

    ...
    print(trace_prefix, func_data)
    ...

This produces output I can identify by "grep TRACE", as well as picking out the stream from a given routine or package. This is easy enough to parameterize (grab the function name); I've done it with a decorator.

Answer 2

If printing happens from Python, setting up logger with elaborate formatters can help you precisely view the module, line number, func name, timestamp, etc.

Here's a sample formatter that I use in my project.

%(asctime)-s.%(msecs)03d %(levelname)-6.6s %(filename)-18.18s line:%(lineno)-4.4d %(funcName)-18s %(message)s

And the results:

2020-08-05 00:28:36.232 DEBUG  connectionpool.py  line:0959 _new_conn          The fancy thing that gets printed.
2020-08-05 00:28:38.177 DEBUG  connectionpool.py  line:0437 _make_request      This is your message but you have more info before this.
2020-08-05 00:28:38.180 DEBUG  item.py            line:0159 start_indexing     Loggers with proper formatters can help you see
2020-08-05 00:28:38.180 DEBUG  indexer.py         line:0036 start              Extensive information

Code:

import logging
import sys

LOGGER = logging.getLogger(__name__)
HANDLER = logging.StreamHandler(sys.stdout)
HANDLER.setFormatter(logging.Formatter("%(asctime)-s.%(msecs)03d %(levelname)-6.6s %(filename)-18.18s line:%(lineno)-4.4d %(funcName)-18s %(message)s"))
LOGGER.addHandler(HANDLER)

LOGGER.error("Here we go!")

Answer 3

1. If your program runs in multi-process, you need to know which process writes "0" to terminal.

#!/usr/bin/env python3
'''
a simple program to stimulate writing terminal in multiple processes.
'''
prog name: ttywriter.py
import time
import sys
import multiprocessing
def print_stderr(s):
    sys.stderr.write(f'{s}\n')

while True:
    print("I'm writing to my terminal")
    p = multiprocessing.Process(target=print_stderr,args=("I am a CPP program, writing to termianl",))
    p.start()
    time.sleep(1)

Pseudo terminal is usually opened as file descriptor 1 and 2. you can strace them.

# tty
/dev/pts/0

# ps -ef
UID          PID    PPID  C STIME TTY          TIME CMD
root        2640    2427  0 07:07 pts/0    00:00:00 python3 ./ttywriter.py

# lsof -p `pgrep -f ttywriter` | grep pts
python3 2640 root    0u   CHR  136,0      0t0     3 /dev/pts/0
python3 2640 root    1u   CHR  136,0      0t0     3 /dev/pts/0
python3 2640 root    2u   CHR  136,0      0t0     3 /dev/pts/0

# strace -s 65536 -e trace=write -e write=1,2 -fp `pgrep -f ttywriter`
strace: Process 2640 attached
write(1, "I'm writing to my terminal\n", 27) = 27
 | 00000  49 27 6d 20 77 72 69 74  69 6e 67 20 74 6f 20 6d  I'm writing to m |
 | 00010  79 20 74 65 72 6d 69 6e  61 6c 0a                 y terminal.      |
strace: Process 2695 attached
[pid  2695] write(2, "I am a CPP program, writing to termianl\n", 40) = 40
 | 00000  49 20 61 6d 20 61 20 43  50 50 20 70 72 6f 67 72  I am a CPP progr |
 | 00010  61 6d 2c 20 77 72 69 74  69 6e 67 20 74 6f 20 74  am, writing to t |
 | 00020  65 72 6d 69 61 6e 6c 0a                           ermianl.         |
[pid  2695] +++ exited with 0 +++
--- SIGCHLD {si_signo=SIGCHLD, si_code=CLD_EXITED, si_pid=2695, si_uid=0, si_status=0, si_utime=0, si_stime=0} ---

2. if it's Python, you can hijack sys.out or print to print call stack.

3. if it's C/C++, you can set a conditional breakpoint for a specific string being written to terminal .