[developers] Newer 64-bit builds cause Non-structure argument error

[Stephan Oepen]

hi mike,

thanks for the bug report.  to clarify, do you need to
issue an [incr tsdb()] command to trigger the issue,
or does it just happen as part of the batch parsing
command?

it looks to me as if (a) your results include MRSs
(not a bad thing per se :-), and (b) [incr tsdb()] tries
to operate on those MRSs but fails to unserialize.
i would not expect (b) to be part of batch parsing
(but commands like `Browse|Results' or `Trees|
Annotate' should trigger it), and further expect the
unserializing of `simple' MRSs to be enabled by
default.

any more details on how you are running things
will be appreciated ...

cheers, oe


On Tue, Aug 7, 2012 at 10:00 AM, Michael Wayne Goodman <goodmami at uw.edu> wrote:
> Hello,
>
> I started noticing that my up-to-date versions of LOGON (r11666-11670)
> and the Matrix repositories were giving errors after running a batch
> job. Specifically, I was running the Grammar Matrix's regression
> tests, and after successfully parsing the profile, I would get this
> error:
>
> flush-cache(): flushing `current/tiniest' cache ... done.
> close-connection(): `current/tiniest' expiry.
> NIL
> TSNLP(10): Error: Non-structure argument
>        "[ LTOP: h1 INDEX: e2 [ e SF: PROP-OR-QUES E.TENSE: TENSE E.ASPECT: ASPEC
> T E.MOOD: MOOD ] RELS: < [ \"_dog_n_rel\"<> LBL: h3 ARG0: x4 [ x SPECI: BOOL COG
> -ST: COG-ST PNG: PNG ] ] [ \"exist_q_rel\"<> LBL: h5 ARG0: x4 RSTR: h6 BODY: h7
> ] [ \"_sleep_v_rel\"<> LBL: h8 ARG0: e2 ARG1: x4 ] > HCONS: < h6 qeq h3 > ]"
>        passed to ref of structure slot 1
>   [condition type: SIMPLE-ERROR]
>
> I reproduced the error on 3 different machines, all 64-bit. Fellow
> matrix-developer Joshua said he did not see the error on his 32-bit
> machine. The error is similar to one reported by Bart in 2009
> (http://lists.delph-in.net/archives/developers/2009/001137.html), but
> in my case it was not fixed by changing the discriminant mode (see the
> reply to the above message).
>
> I pinpointed the erroneous behavior to the following LOGON revision:
>
> ----------------------
> r11551 | oe | 2012-05-21 18:13:06 +0800 (Mon, 21 May 2012) | 1 line
>
> new 64-bit build (on mt.ifi, for a change)
> ----------------------
>
> I do not see the error at revision 11550, but I do see it from 11551
> and after. Perhaps a regression was introduced in that build? If not,
> any advice on how to proceed?
>
> --
> -Michael Wayne Goodman