[developers] cheap bug: bool is_leaftype(type_t): Assertion `is_type(s)' failed.

Francis Bond fcbond at gmail.com
Sun Mar 12 15:14:04 CET 2006


G'Day,

with cheap version 0.99.11 I get a new and exciting failure, only when
run in -tsdb mode:

[t40008] cheap: ../common/types.cpp:82: bool is_leaftype(type_t):
Assertion `is_type(s)' failed.
[t40008] Error in format: No more arguments.
[t40008]   Serious signal ~D caught.
[t40008]                   ^
[t40008] while processing indirect format string:
[t40008]   ~?
[t40008]    ^
[t40008] Broken at NIL.

This occurs once every 500 or so sentences with a recent JACY and the
kinou1 test set (now distributed with the grammar).   The first
sentence it is is failing on is:
 `この 七 億 円 は 所得 隠し に あたる '.

If I kill cheap, then tsdb pciks up and keeps going, which is very nice:
[t40008] MRS>
[t40008] EOF
[gin] (287) `この 七 億 円 は 所得 隠し に あたる 。 ' [10000] <1> --- client exit <40008>.
[t40009] BEGIN
[t40009] reading `/home/bond/delphin/grammars/japanese/pet/japanese.set'...
[t40009] loading `/home/bond/delphin/grammars/japanese/japanese.grm'
(Jacy (2006-02-06))
[t40009] Could not open SM file "hinoki.mem"Ignoring entry with
unknown rule `$-*-' in irregular forms
[t40009] Ignoring entry with unknown rule `$this' in irregular forms
[t40009] Ignoring entry with unknown rule `$the' in irregular forms
[t40009] Ignoring entry with unknown rule `$currently' in irregular forms
[t40009] Ignoring entry with unknown rule `$c2stem' in irregular forms
[t40009] Ignoring entry with unknown rule `$---' in irregular forms
[t40009]
[t40009] 132376 types in 4.6 s
wait-for-clients(): `gin' registered as tid <40009> [0:7].
(287) `この 七 億 円 は 所得 隠し に あたる 。 ' [10000] <1> --- error: maximum
number of strikes exhausted (1).

(1) could someone have a look at why cheap is doing this?
(2) does anyone know how to stop ECL falling into the debugger?  It
would be nice if it would just die, not hang there forever.

I would be happy to provide more debugging information.

--
Francis Bond  <www.kecl.ntt.co.jp/icl/mtg/members/bond/>
NTT Communication Science Laboratories | Machine Translation Research Group



More information about the developers mailing list