[developers] More problems with PET

[Christopher Rupp]

Hi,

I'm still having some problems with the current PET version (now revision 509).
(Thanks to Bernd for installing the fix for tokeniser setting.) I think this 
could
be more basic than that but I don't think it is determined by the SMAF 
tokeniser,
as I get a similar behaviour with 'tok=string' and 'tok=fsr'. I think what I'm
not getting is parse results, but there could be something very simple 
determining
that.

The best indication I can present is the -verbose=3 output for a functioning
version (up to derivation output) and the current version. This is working with
an export copy of the /lingo/cvs erg, as of today plus local SMAF interface
definitions, but I assume they don't affect string input.

This is what the old (revision 486) version produces:
id: 1
trees: 4
readings: 1
words: 34
words_pruned: 0
mtcpu: 0
first: 80
tcpu: 100
utcpu: 0
ftasks_fi: 80080
ftasks_qc: 20634
fsubs_fi: 4068
fsubs_qc: 0
etasks: 1868
stasks: 1187
aedges: 673
pedges: 500
upedges: 19
raedges: 2
rpedges: 10
medges: 0
unifications_succ: 1710
unifications_fail: 681
subsumptions_succ: 68
subsumptions_fail: 1311
copies: 1258
dyn_bytes: 6493440
stat_bytes: 9503424
p_dyn_bytes: 140760
p_nstat_bytes: 9643632
cycles: 0
fssize: 1
unify_cost_succ: 138
unify_cost_fail: 51
equivalent: 29
proactive: 29
retroactive: 4
frozen: 27
failures: 0
hypotheses: 146

And the new current one:

id: 1
trees: 4
readings: 0
words: 34
words_pruned: 0
mtcpu: 0
first: 70
tcpu: 100
utcpu: 0
ftasks_fi: 80080
ftasks_qc: 20634
fsubs_fi: 4068
fsubs_qc: 0
etasks: 1868
stasks: 1187
aedges: 673
pedges: 500
upedges: 0
raedges: 2
rpedges: 10
medges: 0
unifications_succ: 1682
unifications_fail: 681
subsumptions_succ: 68
subsumptions_fail: 1311
copies: 1231
dyn_bytes: 6493440
stat_bytes: 9503424
p_dyn_bytes: 0
p_nstat_bytes: 0
cycles: 0
fssize: 1
unify_cost_succ: 138
unify_cost_fail: 51
equivalent: 29
proactive: 29
retroactive: 4
frozen: 27
failures: 0
hypotheses: 0

On casual inspection upedges, p_dyn_bytes, p_nstat_bytes and hypotheses seem 
to be
missing.

Any ideas would be helpful. (I have until Friday to solve this.

Cheers,

C.J.

-- 
Dr. C.J. Rupp

University of Cambridge Computer Laboratory
William Gates Building
15 JJ Thomson Avenue
Cambridge CB3 0FD, UK

Tel: +44 1223 767025 
Fax: +44 1223 334678
Mobile: +44 795-8496 916
Email: cr351 at cl.cam.ac.uk