Bug 2438 - hfst compilation fails
Summary: hfst compilation fails
Status: NEW
Alias: None
Product: Kven fkv
Classification: Unclassified
Component: Lexicon (show other bugs)
Version: unspecified
Hardware: Macintosh Linux
: P5 - Later enhancement
Assignee: Trond Trosterud
URL:
Keywords:
Depends on:
Blocks:
 
Reported: 2017-10-19 20:44 CEST by Børre Gaup
Modified: 2017-11-20 12:39 CET (History)
2 users (show)

See Also:

Attachments
Add an attachment (proposed patch, testcase, etc.)

Note You need to log in before you can comment on or make changes to this bug.
Description Børre Gaup 2017-10-19 20:44:53 CEST 
Using these configure options:
./configure --with-hfst --without-xfst --enable-tokenisers --enable-reversed-intersect --enable-alignment

Gives this result:
make[2]: Entering directory '/home/boerre/repos/langtech/langs/fkv/src/phonology'
  HTWOLC   fkv-phon.hfst
Exception: HfstException in file: htwolcpre1-parser.yy on line: 541
Comment 1 Trond Trosterud 2017-10-25 13:44:48 CEST 
I now repeat this result,w ith the settings:
./configure --with-hfst --enable-oahpa
and the svn version of the twolc file.

Also, nn gemination (sano ==> sannoot) fails for the .xfst analyser.
Comment 2 Trond Trosterud 2017-11-20 11:43:52 CET 
I found the error: A % missing in front of a  ^HMETA (should be %^HMETA).

Unfortunately the case is not quite closed: The code now compiles:

- in twolc, without the error message
*** Warning: make_num_re: Positive integer expected, got 0. ***

- in hfst, without the error message

But the test form sannoot still fails in xfst, we get no nn gemination:
 
sannoot
sannoot	sannoot	+?

sannoot+V+Inf
sannoot+V+Inf	sanoot
sannoot+V+Inf	sanoa

Interestingly, it **succeeds** in hfst:

tf4-hsl-m0024:fkv trond$ hufkv
sannoot
sannoot	sannoot+V+Inf	0.000000

So this calls for some afterthought.
Comment 3 Trond Trosterud 2017-11-20 12:38:18 CET 
New testing. Here is the situation:
- Everythings work compiled with hfst (so the bug name is now misleading).
- With the xerox tools, we have this situation:
-- dfkv generates "sannoot+V+Inf" ok, but it cannot analyse it, instead it analyses "sanoot"
-- testing in twolc, we get mixes results: lex-test gives non-geminated n (as does the analyser), whereas the pair-test behaves as it should:

twolc> lex-test

Lexical string ('q' = quit): san0o>^Vt
                             san0o>^Vt
s
a
n
0
o
>
^V
t

Lexical string ('q' = quit): q

twolc> pair-test

Lexical string ('q' = quit): san0o>^Vt
Surface string ('q' = quit): sanno>ot
s
a
n
0:n
o
>
^V:o
t
ACCEPTED

Lexical string ('q' = quit): sano>^Vt
Surface string ('q' = quit): sano>ot
s
a
n
o
>
^V:o
REJECTED: "Gemination Vo" fails in state 113.
Comment 4 Trond Trosterud 2017-11-20 12:39:21 CET 
New testing. Here is the situation:
- Everythings work compiled with hfst (so the bug name is now misleading).
- With the xerox tools, we have this situation:
-- dfkv generates "sannoot+V+Inf" ok, but it cannot analyse it, instead it analyses "sanoot"
-- testing in twolc, we get mixes results: lex-test gives non-geminated n (as does the analyser), whereas the pair-test behaves as it should:

twolc> lex-test

Lexical string ('q' = quit): san0o>^Vt
                             san0o>^Vt
s
a
n
0
o
>
^V
t

Lexical string ('q' = quit): q

twolc> pair-test

Lexical string ('q' = quit): san0o>^Vt
Surface string ('q' = quit): sanno>ot
s
a
n
0:n
o
>
^V:o
t
ACCEPTED

Lexical string ('q' = quit): sano>^Vt
Surface string ('q' = quit): sano>ot
s
a
n
o
>
^V:o
REJECTED: "Gemination Vo" fails in state 113.