[itsdb] [incr tsdb()] and PVM problems

Michael Wayne Goodman goodmami at u.washington.edu
Wed Aug 24 22:23:59 CEST 2011


Hi,

I'm getting PVM errors when I try to use PET for parsing with [incr
tsdb()]. I am running Fedora 15 instead of Ubuntu, but I noticed this
problem on an Ubuntu install as well. I am also using LOGON revision
9746 from http://svn.emmtee.net/trunk.

The problem does not appear to be related to Stefan Müller's PVM issue
from May 2008 (http://lists.delph-in.net/archive/itsdb/2008-May/000110.html),
but I could be wrong.

Here are the errors (from STDERR of the $LOGONROOT/bin/logon script):

====
libpvm [pid27468] /tmp/.pvm.socket.goodmami: No such file or directory
libpvm [pid27468] /tmp/.pvm.socket.goodmami: No such file or directory
libpvm [pid27468] /tmp/.pvm.socket.goodmami: No such file or directory
libpvm [pid27468]: pvm_mytid(): Can't contact local daemon
libpvm [pid27468]: pvm_register(): Can't contact local daemon
pvm_register(): unable to initialize virtual machine.
libpvm [pid27468] /tmp/.pvm.socket.goodmami: No such file or directory
libpvm [pid27468] /tmp/.pvm.socket.goodmami: No such file or directory
libpvm [pid27468] /tmp/.pvm.socket.goodmami: No such file or directory
libpvm [pid27468] /tmp/.pvm.socket.goodmami: No such file or directory
libpvm [pid27468]: pvm_mytid(): Can't contact local daemon
====

...and later during parsing:

====
(1) `dog slept' [100000] --- error: lexicon is not open for reading.
(2) `cat slept' [100000] --- error: lexicon is not open for reading.
(3) `dog cat chased' [100000] --- error: lexicon is not open for reading.
(4) `cat dog chased' [100000] --- error: lexicon is not open for reading.
(5) `*slept dog' [100000] --- error: lexicon is not open for reading.
(6) `*slept cat' [100000] --- error: lexicon is not open for reading.
(7) `*dog chased cat' [100000] --- error: lexicon is not open for reading.
(8) `*cat chased dog' [100000] --- error: lexicon is not open for reading.
(9) `*chased dog cat' [100000] --- error: lexicon is not open for reading.
====

(I doubt it matters, but I am using a small grammar in the Matrix
regression tests)

These errors only occur when I use PET for parsing. It works fine when
the LKB is used. The only difference between the commands for using
PET vs the LKB are in the CPU definition. The list commands below are
used for PET, and when we use the LKB we comment out the first block
uncomment the last line below:

====
          # Use the following for PET parsing
          echo "(setf *tsdb-cache-connections-p* t)"
          echo "(setf *pvm-encoding* :utf-8)"
          echo "(setf *pvm-cpus* (list (make-cpu"
          echo "  :host (short-site-name)"
          echo "  :spawn \"${LOGONROOT}/bin/cheap\""
          echo "  :options (list \"-tsdb\" \"-packing\" \"-mrs\" \"$grm_file\")"
          echo "  :class :$lgname :name \"$lgname\""
          echo "  :grammar \"$lgname (current)\""
          echo "  :encoding :utf-8"
          echo "  :task '(:parse) :wait 300 :quantum 180)))"
          echo "(tsdb:tsdb :cpu :$lgname :task :parse :file t)"

          # Use the following for LKB parsing
          #echo "(lkb::read-script-file-aux \"$grammar\")"
====

These commands have worked before, so it could be something in my
local configuration that is causing the issue.

I have tried clearing any /tmp/.pvm.* files, but the issue remains.
Any ideas? Let me know if more information is needed.

Thanks

-- 
-Michael Wayne Goodman




More information about the itsdb mailing list