FW: [mpich-discuss] mpdtrace cannot connect (cygwin & mpd), previous smpd remnants?

[Jayesh Krishna]

 Hi,
  Did you install Cygwin after logging in as "ibojak" ? Is user "ibojak" in
the Administrators Group (If you recently added ibojak to the Administrators
group did you do a "mkpasswd -l > /etc/passwd" "mkgroup -l > /etc/group" ?
Make sure that you save a backup of /etc/passwd and /etc/group before
overwriting them.)?

(PS: During Cygwin installation I think there is an option to select whether
the installation is for "only the current user" or "all users". I usually
select "all users" to get out of permission troubles later on  :) The
permission pblms will kill you if you try to setup ssh in your machine.)

Regards,
Jayesh
-----Original Message-----
From: owner-mpich-discuss at mcs.anl.gov
[mailto:owner-mpich-discuss at mcs.anl.gov] On Behalf Of Ingo Bojak
Sent: Wednesday, April 02, 2008 11:51 AM
To: mpich-discuss at mcs.anl.gov
Subject: Re: FW: [mpich-discuss] mpdtrace cannot connect (cygwin & mpd),
previous smpd remnants?

Jayesh Krishna wrote:
> A little more info may gained by changing this line in the installed 
> copy of
> mpdlib.py:
>          mpd_print(1111,"PTC2\n")
> to this:
>          mpd_print(1111,"PTC2 %s\n",errmsg) I should have done it this 
> way but didn't think of it.
>   
Actually, after some scanning of the Python code, the correct update appears
to be rather

mpd_print(1111,"PTC2 %s\n" % errmsg)

(not a comma like in C, but a %). With this I get the following output:

---
 > mpdtrace
mpdtrace (__init__ 1194): PTB

mpdtrace (__init__ 1208): PTC

mpdtrace (__init__ 1219): PTC2 (9, 'Bad file descriptor')

PTF

mpdtrace: cannot connect to local mpd (/tmp/mpd2.console_ibojak); possible
causes:
  1. no mpd is running on this host
  2. an mpd is running but was started without a "console" (-n option) In
case 1, you can start an mpd on this host with:
    mpd &
and you will be able to run jobs just on this host.
For more details on starting mpds on a set of hosts, see the MPICH2
Installation Guide.
---

> If mpd is running, and you do an "ls -l" of /tmp, e.g.:
>          ls -l /tmp/mpd*
> you should see a file whose type is socket.  For example on my MacOSX 
> box, I
> see:
>          srwxr-xr-x  1 rbutler  wheel  0 Apr  2 09:36 /tmp/ 
> mpd2.console_rbutler This is mpd's socket that mpiexec tries by 
> default to connect to.
>   
This is the output I get:
---
 > ls -l /tmp
total 2048
-rw-r--r-- 1 ibojak None 1983 Apr  2 15:39 XWin.log
-rw-r--r-- 1 ibojak None   53 Apr  2 17:56 mpd2.console_ibojak
---
I think we are getting close to the solution here though, because the "x"
for executable is missing. However, now it gets weird: my user owns these
files, and the directory

drwxr-xr-x   1 ibojak None     0 Apr  2 18:32 tmp

but I'm not in fact able to change permissions:
---
 > chmod a+x mpd2.console_ibojak
 > ls -l
total 2048
-rw-r--r-- 1 ibojak None 2111 Apr  2 18:31 XWin.log
-rw-r--r-- 1 ibojak None   53 Apr  2 18:32 mpd2.console_ibojak
---
Note that there is no error, but nothing happens. On the other hand, in my
home directory

drwxrwxrwx+ 7 ibojak None 0 Apr  2 16:14 ibojak

I can change permissions
---
 > touch entity
 > ls -l entity
-rw-r--r-- 1 ibojak None 0 Apr  2 18:44 entity

 >chmod a+x entity
 >ls -l entity
-rwxr-xr-x 1 ibojak None 0 Apr  2 18:44 entity
---
I note that the /tmp directory has no og+w permissions, but I can't chmod
that either! I have no clue what's going on there or how to remedy it.

> Did you intentionally mask the username (ibojak) in your previous 
> emails (see below) ?
>   
Indeed, I did. I also masked the computername (by replacing it with
"computername") and its IP (by "x"-ing most numbers).

>   Query the file permissions of the socket using the getfacl command, eg:
> getfacl /tmp/mpd2.console_ibojak .
---
 > getfacl /tmp/mpd2.console_ibojak
# file: /tmp/mpd2.console_ibojak
# owner: ibojak
# group: None
user::rw-
group::r--
other:r--
mask:rwx
---

> (PS: Use "whoami" to list the current user. Use "mkpasswd -l" to get a 
> list of users on the system.)
>
>   
Output of the following stripped of all but user names (and "x"-ed some
numbers):
---
 > whoami
ibojak

 > mkpasswd -l
SYSTEM
LocalService
NetworkService
Administrators
Administrator
Guest
HelpAssistant
ibojak
SUPPORT_xxxxxxxx
---

> Also make sure that your cygwin user has enough privileges to create a 
> socket (Is the cygwin user an admin ?).
>   
My understanding is that within cygwin the installing user (ibojak) is also
root for all intents and purposes. If you try
---
 > su
su: user root does not exist
---
I'm also logged in as a the Computer Administrator on XP when starting
cygwin. As mentioned, turning the Windows firewall on or off has no effect
on the output within cygwin, e.g., for the mpdtrace above.

Cheers,
Ingo