Re: how to run shell command like "<<EOT .... EOT"

[Kushal Kumaran]

On Fri, Sep 28, 2012 at 1:15 PM, 叶佑群 <ye.y...@eisoo.com> wrote:
> Hi, all,
>
> I have the shell command like this:
>
> sfdisk -uM /dev/sdb << EOT
> ,1000,83
> ,,83
> EOT
>
>
> I have tried subprocess.Popen, pexpect.spawn and os.popen, but none of
> these works, but when I type this shell command in shell, it is works fine.
> I wonder how to emulate this type of behavior in python , and if someone can
> figure out the reason why?
>
> The sample code of subprocess.Popen is:
>
> command = ["sfdisk", "-uM", target, "<<EOT", "\r\n",
> ",", 1000, ",", "83", "\r\n",
> ",", ",", "83", "\r\n", "EOT", "\r\n"]
>
> pobj = subprocess.Popen (command, bufsize=1, \
> stderr=subprocess.PIPE, stdout=subprocess.PIPE)
>
> res = pobj.stderr.readline ()
> if res is not None and pobj.returncode != 0:
> observer.ShowProgress (u"对设备 %s 分区失败！" % target)
> return False
>

The "<<EOT" syntax (called a here-document) just provides input to the
command. If you use the communicate method, you can provide input as
an argument:

command = ["sfdisk", "-uM", target ]
instructions = """
,1000,83
,,83
"""
pobj = subprocess.Popen(command, stdin=subprocess.PIPE,
stdout=subprocess.PIPE, stderr=subprocess.PIPE)
(output, errors) = pobj.communicate(instructions)

> and pexpect code is:
>
> child = pexpect.spawn ("sfdisk -uM /dev/sdb <<EOT")
> child.sendline (....)
> child.sendline (....)
> child.sendline (....)
>
> and os.popen like this:
>
> os.popen ("sfdisk -uM /dev/sdb <<EOT\n,1000,83\n,,83\nEOT\n")
>
> I tried "\r\n", and it doesn't work either.
>

--
regards,
kushal

[叶佑群]

I tried this, but it is still not work.

[叶佑群]

If I want to read the output line by line and not put all output to
memory buffer in one time, how to write the code?

[Kushal Kumaran]

> I tried this, but it is still not work.
>

What do you mean by "not work"?

- If you get an exception, copy the entire traceback into an email

- If you do not get an exception, print out the value of the "errors"
variable to see why the command failed. You can also check
pobj.returncode for the exit status of the subprocess.

A possibility is that you have to replace "sfdisk" with the full path
to the binary, if it cannot be located on the PATH. So you will
replace it with "/usr/sbin/sfdisk", or "/sbin/sfdisk", or wherever the
file actually is.

<from your other email>

> If I want to read the output line by line and not put all output to memory buffer in one
> time, how to write the code?

You can read line by line by calling pobj.stdout.readline() and
pobj.stderr.readline(). You can send input to the process by calling
pobj.stdin.write(). If you manage this interaction "by hand", you
should not call communicate(). Also, you should be aware of the
problem mentioned in the subprocess documentation:

"Use communicate() rather than .stdin.write, .stdout.read or
.stderr.read to avoid deadlocks due to any of the other OS pipe
buffers filling up and blocking the child process."

Is there any reason why you need to read line-by-line? You could use
communicate(), and then call stdout.splitlines() to get a list of
lines, if that's all you need.

--
regards,
kushal

[叶佑群]

Sorry for replying so late, these days are long vocation in china.

If I type command in shell line by line, the command will run as
expected, but when I invoke the command in python, it is always failed.
Which is what I mean "not work".

>
> - If you get an exception, copy the entire traceback into an email

No exception occured.

>
> - If you do not get an exception, print out the value of the "errors"
> variable to see why the command failed. You can also check
> pobj.returncode for the exit status of the subprocess.

I solved this problem as below:

fop = os.popen ("sfdisk -uM %s <<EOT\n,%d,83\n,,83\nEOT\n" % (target, fps))

But when I run it with subprocess.Popen (), it doesn't work as os.popen
even use the code as:

command = ["sfdisk", "-uM", target ]
instructions = """
,1000,83
,,83
"""

pobj = subprocess.Popen(command, stdin=subprocess.PIPE,
stdout=subprocess.PIPE, stderr=subprocess.PIPE)
(output, errors) = pobj.communicate(instructions)

stderr.readline () always return the error message means "sfdisk won't accept more than one device parameter except -s or -l is specified", it seems that some parameters were incorrectly treated