error wth script

Greg Hennessy wrote on Oct 17, 2006

I'm trying to reduce some data from the CTIO smart data. I'm using some scripts written by Phil Massey, which ought to mean they work perfectly :), but I'm having a problem that neither I nor Phil can figure out.

The symptom is that a task is being reported as having no param file, i.e,
cl> y4kphot
Py051109.0006
Py051109.0007
Py051109.0017
[snip bunch of lines]
Py051109.0096
Py051109.0097
Py051109.0098
ERROR on line 35: task `y4kphotit' has no param file
y4kphot ()
y4kphot ()
cl>

However, I belive I have the tasks defined properly in my login.cl file,
[gsh@libra ~/iraf]$ grep phot loginuser.cl
task $y4kphot=/opt/massey/y4kphot.cl
task y4kphotit=/opt/massey/y4kphotit2.cl
$ ls -al /opt/massey/y4kphot*.cl
-rw-r--r-- 1 gsh gsh 813 Oct 7 14:25 /opt/massey/y4kphot.cl
-rw-r--r-- 1 gsh gsh 674 Oct 7 15:05 /opt/massey/y4kphotit2.cl
[gsh@libra ~/iraf]$

I'm running iraf PC-IRAF V2.12 on a 64 bit dual opteron with Fedora Core 5.

Does anyone have any sugguestions?

---

Mike Fitzpatrick wrote on Oct 17, 2006

Two possible issues: The "snip bunch of lines" makes me think this is a file descriptor leak, i.e. as the script (??is it a script??) processes each file it uses new FD for each file until it runs out. This most often happens in a construct like

list = myfile
while (fscan(list,x,y) != EOF) {
   .....
}
list = ""

The last line effectively "closes" the file and may be what's missing. Check also your underlying environment with a "limit descriptors" (csh version) command, the iraf limit is something like 4096.

A second issue may be an underlying 64-bit filesystem in which the inode for the file isn't 'seen' because of 64-bit support issues. Since the error is on the uparm file I don't think this is the case however. Hope this helps.

Cheers,
-Mike

---

Greg Hennessy wrote on Oct 17, 2006

I added a line
flist = "" to the script, but the behavior did not change.

My environment is
[gsh@libra massey]$ limit
cputime unlimited
filesize unlimited
datasize unlimited
stacksize 10240 kbytes
coredumpsize 0 kbytes
memoryuse unlimited
vmemoryuse unlimited
descriptors 1024
memorylocked 32 kbytes
maxproc 16379

If it were the 64 bit issue you referred to, how might I test that possibility?

---

Greg Hennessy wrote on Oct 17, 2006

I just discovered the weirdest thing. I made a copy of the script so I could endit it down to try to find the smallest script that worked, thinking that might help isolate the problem, and having defined a new task in my loginuser.cl file
I found that the copy runs.

I thought computers were *supposed* to be repeatable.

This is probably some subtle bug that will be almost impossible to debug properly.

---

Last post on Oct 17, 2006