bug#15683: [critical] ERROR: ... close-pipe: pipe not in table

[David Pirotte]

Hello guilers,

        GNU Guile 2.0.9.20-10454

[cfr irc chat of October the 17th, 2013]

I am facing a bug that only occurs on extremely powerful servers:

        in ice-9/popen.scm:
         106: 1 [close-pipe #<input: #{read pipe}# 69>]
        In unknown file:
         ?: 0 [scm-error misc-error #f "~A" ("close-pipe: pipe not in table") 
#f]

I can not reproduce the bug on my personal computer [i5, 4 cores], neither on 
the
lab most powerful server we have [i7 12 cores], but on this customer's server [2
Xeons E5-2687W, 32 cores total], the bug is not random anymore, it _always_ 
raises, 
which is critical to us.

Thank you for debugging this asap,
Cheers,
David

in case it might help, here is an extract of the code that raises the error.  
in this code,
rg-ergbd1 is an octave [heavy] script [that will be called between 1000 to 
62000 depending on
other factors...]

...

(define (ergbd path im-name im-type seeds-dir im-ones x y threshold 
connectivity mutex log-port)
  (with-mutex mutex ;; (write-log-filename (format #f "(~A, ~A) " x y) log-port)
              (write-log "." log-port))
  (let* ((cmd (format #f "rg-ergbd1 ~A ~A ~A ~A ~A ~A ~A ~A ~A" path im-name 
im-type seeds-dir im-ones x y threshold connectivity))
         (s (open-input-pipe cmd))
         (results (read-line s)))
    (unless (zero? (status:exit-val (close-pipe s)))
      (error "subprocess returned non-zero result code" cmd))
    results))
...
...
        ...
        (par-map (lambda (coord)
               (ergbd target-dir im-cpol-norm-name im-type seeds-dir im-ones 
(car coord) (cdr coord) threshold connectivity mutex log-port))
          coords)
        ...