Determine process environment variables

[giampaolo]

From billiej...@gmail.com on May 12, 2009 17:46:04

Proposal  
for addition of a new Process.environ property returning the
environment variables registered by a process, possibly as a dictionary.

Original issue: http://code.google.com/p/psutil/issues/detail?id=52

[giampaolo]

From jlo...@gmail.com on May 12, 2009 13:22:58

Status: Accepted
Labels: Milestone-0.1.3

[giampaolo]

From billiej...@gmail.com on May 12, 2009 17:22:30

On Windows this seems to be painful to determine although still possible: http://www.codeguru.com/forum/archive/index.php/t-416841.html

[giampaolo]

From jlo...@gmail.com on July 13, 2009 18:18:59

Ugh, this is just as ugly as determining the cmdline was, it uses basically the same
method and will likely suffer from the same reliability problems we're seeing there.

[giampaolo]

From billiej...@gmail.com on September 17, 2009 02:06:06

Removing 0.1.3 milestone

Labels: -Milestone-0.1.3

[giampaolo]

From billiej...@gmail.com on February 04, 2010 05:40:13

Labels: Milestone-0.2.0

[giampaolo]

From wj32...@gmail.com on February 26, 2010 02:04:27

I'm the guy who helped you with the command line retrieval on Windows. Please ignore
those CreateRemoteThread idiots. Take a look at this code (function
PhGetProcessEnvironmentVariables): 
http://processhacker.svn.sourceforge.net/viewvc/processhacker/2.x/trunk/ProcessHacker/native.c?revision=HEAD&view=markup

[giampaolo]

From billiej...@gmail.com on March 01, 2010 07:56:12

Labels: -Milestone-0.2.0

[giampaolo]

From g.rodola on July 09, 2010 13:10:26

Just for the record when we'll reconsider this, Linux code was committed as r406 and removed in r464 and r599 .

Labels: -Progress-0in4 Progress-1in4

[giampaolo]

From g.rodola on September 24, 2010 16:40:43

Status: PostPoned

[giampaolo]

From g.rodola on March 01, 2011 09:51:11

On FreeBSD we can use kvm_getenvv(): http://fuse4bsd.creo.hu/localcgi/man-cgi.cgi?kvm_getenvv+3

[giampaolo]

From g.rodola on January 24, 2012 13:23:25

Issue 247 has been merged into this issue.

[giampaolo]

From g.rodola on January 31, 2012 08:42:15

Another article about Windows: 
http://www.codeproject.com/Articles/25647/Read-Environment-Strings-of-Remote-Process

[giampaolo]

From g.rodola on February 01, 2012 10:22:40

Reopening as I have a partial working patch for Windows.
I'd like to have this ready for 0.5.0.

Status: ReOpened
Labels: Milestone-0.5.0

[giampaolo]

From g.rodola on February 01, 2012 11:06:08

Linux implementation readded as r1259 .
Windows implementation added as r1260 .

Labels: -Progress-1in4 Progress-2in4

[giampaolo]

From g.rodola on February 01, 2012 14:53:24

kvm_getenvv() on FreeBSD can only be used if /proc filesystem is mounted (and it's not by default).
Partial implementation is below but I guess we better off not supporting FreeBSD at all.

static PyObject*
get_process_environ(PyObject* self, PyObject* args)
{
    long pid;
    char **env;
    struct kinfo_proc *pinfo;
    int num;
    static kvm_t **kd;

    if (! PyArg_ParseTuple(args, "l", &pid)) {
        return NULL;
    }

    kd = kvm_open(NULL, NULL, NULL, O_RDONLY, NULL);

    pinfo = kvm_getprocs(kd, KERN_PROC_PID, pid, &num);
    if (pinfo == NULL) {
        return PyErr_SetFromErrno(PyExc_OSError);
    }

    env = kvm_getenvv(kd, pinfo, 9086);
    if (env == NULL) {
        return PyErr_SetFromErrno(PyExc_OSError);
    }

    return Py_BuildValue("i", 33);
}

[giampaolo]

From david.da...@gmail.com on April 24, 2012 21:45:37

I fixed the heap corruption issue happening with the get_process_environ() call on windows in r1297

Cc: g.rodola

[giampaolo]

From g.rodola on April 25, 2012 04:57:59

Great! Thanks.

[giampaolo]

From g.rodola on April 25, 2012 05:57:13

Looking back at this I start to doubt the usefulness for two reasons:

- on Linux this seems to work just fine but if the environment of the process is modified the change is not propagated:

>>> p = psutil.Process(os.getpid())
>>> os.putenv('XXX', '1')
>>>'XXX' in p.get_environ()
False
>>>

- on Windows this can be used for the current process only; for all other 
processes, including the ones owned by the current user, we get an AccessDenied 
exception - also, if the environment of the current process is changed we also get AccessDenied.


That said, it probably make sense to give up on this.
I'm going to leave this open for a while a take a definitive decision later.

Status: WaitingForReview

[giampaolo]

From g.rodola on June 26, 2012 10:08:43

More info about Linux not updating /proc/PID/environ: 
http://serverfault.com/questions/66363/environment-variables-of-a-running-process-on-unix/79463#79463

[giampaolo]

From g.rodola on June 26, 2012 10:31:22

Ok, I'm still inclined to remove this due to the unreliabilities explained 
above for  both Windows and Linux. Removal done in r1366 and closing this out as WontFix.

Status: WontFix
Labels: -Milestone-0.5.0

[giampaolo]

From g.rodola on March 02, 2013 03:49:26

Updated csets after the SVN -> Mercurial migration: r406 == revision 
3e6905e0c623 r1260 == revision a299e2431879 r1297 == revision e278f365986c r1366 == revision b1dd9bd9ba25