Mailing List Archive: Read past EOF ... Searcher.pm line 124

Read past EOF ... Searcher.pm line 124

Apr 30, 2007, 1:12 AM

Post #1 of 5 (1766 views)

Hi Marvin,

I know the subversion version is in a state of flux at the moment, but
just thought I'd give you a heads up:

Remember the bug which crashed the searcher on:

title2:test.co.us ?

Thanks for fixing that, but the following now happens:

title2:test.co --> fails

title2:test.co. --> fails

title2:test --> works

title2:test. --> works

The error from the log:

Read past EOF of uscon_invindex/_1.cf (start: 18446744073709551228 len
624) at
/usr/lib/perl5/site_perl/5.8.8/x86_64-linux-thread-multi/KinoSearch/Searcher.pm
line 124

\tKinoSearch::Searcher::collect('KinoSearch::Searcher=HASH(0xfc0840)',
'collector', 'KinoSearch::Search::TopDocCollector=SCALAR(0xfb9c20)',
'query', 'KinoSearch::Search::PhraseQuery=HASH(0x101a0d0)', 'filter',
'undef', 'num_wanted', 10, ...) called at
/usr/lib/perl5/site_perl/5.8.8/x86_64-linux-thread-multi/KinoSearch/Searcher.pm
line 75

\tKinoSearch::Searcher::top_docs('KinoSearch::Searcher=HASH(0xfc0840)',
'num_wanted', 10, 'query',
'KinoSearch::Search::PhraseQuery=HASH(0x101a0d0)', 'filter', 'undef',
'sort_spec', 'undef', ...) called at
/usr/lib/perl5/site_perl/5.8.8/x86_64-linux-thread-multi/KinoSearch/Search/Hits.pm
line 44

\tKinoSearch::Search::Hits::seek(3) called at
/usr/lib/perl5/site_perl/5.8.8/x86_64-linux-thread-multi/KinoSearch/Search/Searchable.pm
line 44

\tKinoSearch::Search::Searchable::search('KinoSearch::Searcher=HASH(0xfc0840)',
'query', 'title2:test.co.', 'offset', 0, 'num_wanted', 10) called at...

...

Regards
Henry

Read past EOF ... Searcher.pm line 124 [ In reply to ]

marvin at rectangular

Apr 30, 2007, 10:06 PM

Post #2 of 5 (1704 views)

Permalink

On Apr 30, 2007, at 12:57 AM, Henka wrote:

> Remember the bug which crashed the searcher on:
>
> title2:test.co.us ?

Thanks for the tip. It's now fixed, along with a slew of other bugs
that showed up with Sunday night's massive commit.

We're still failing one test, and we've still got one warning...

Marvin Humphrey
Rectangular Research
http://www.rectangular.com/

Read past EOF ... Searcher.pm line 124 [ In reply to ]

henka at cityweb

May 1, 2007, 3:45 AM

Post #3 of 5 (1705 views)

Permalink

On Mon, 30 Apr 2007, Marvin Humphrey wrote:

>
> On Apr 30, 2007, at 12:57 AM, Henka wrote:
>
> > Remember the bug which crashed the searcher on:
> >
> > title2:test.co.us ?
>
> Thanks for the tip. It's now fixed, along with a slew of other bugs
> that showed up with Sunday night's massive commit.

Excellent, thanks Marvin. Works like a charm.

Read past EOF ... Searcher.pm line 124 [ In reply to ]

dpavlin at rot13

May 1, 2007, 8:51 AM

Post #4 of 5 (1708 views)

Permalink

On Mon, Apr 30, 2007 at 09:47:56PM -0700, Marvin Humphrey wrote:
>
> On Apr 30, 2007, at 12:57 AM, Henka wrote:
>
>> Remember the bug which crashed the searcher on:
>>
>> title2:test.co.us ?
>
> Thanks for the tip. It's now fixed, along with a slew of other bugs that
> showed up with Sunday night's massive commit.
>
> We're still failing one test, and we've still got one warning...

I have similar problem, but my error message is something like:

WARN - Error in function refill at ../c_src/KinoSearch/Store/InStream.c:83: Read past EOF of /home/dpavlin/Grep/var/invindex/_7.cf (start: 51 len 0) at /usr/local/lib/perl/5.8.8/KinoSearch/Searcher.pm line 123

This is current svn checkout, r2382. Should I try to find smallest
possible test case?

--
Dobrica Pavlinusic 2share!2flame dpavlin@rot13.org
Unix addict. Internet consultant. http://www.rot13.org/~dpavlin

Read past EOF ... Searcher.pm line 124 [ In reply to ]

marvin at rectangular

May 1, 2007, 6:28 PM

Post #5 of 5 (1709 views)

Permalink

On May 1, 2007, at 8:32 AM, Dobrica Pavlinusic wrote:

> I have similar problem, but my error message is something like:
>
> WARN - Error in function refill at ../c_src/KinoSearch/Store/
> InStream.c:83: Read past EOF of /home/dpavlin/Grep/var/invindex/
> _7.cf (start: 51 len 0) at /usr/local/lib/perl/5.8.8/KinoSearch/
> Searcher.pm line 123
>
> This is current svn checkout, r2382. Should I try to find smallest
> possible test case?

Yes, please.

You might consider updating to revision 2388 first. It will not
solve the problem, but it will produce an improved error message,
specifying the name of the subfile ("_7.p3" or the like") instead of
the name of the compound file ("_7.cf" above).

If you do that, though, you will need to regenerate your index, as
revisions 2385 and 2386 both changed the file format.

Incidentally, this is what the big commit on Sunday did.

* Complete the process of isolating posting file format specs within
individual subclasses of Posting. Now we can treat them as plugins
to add, archive, improve, etc, in isolation.
* Improve best-case indexing speed by another 15%, by integrating
some
custom memory management techniques. KS now benchmarks more than
twice as fast for indexing the Reuters test collection in
comparison
with version 0.09 from a year ago (c. 29 seconds versus 68
seconds).

Marvin Humphrey
Rectangular Research
http://www.rectangular.com/