Xapian devel - Nov 2007 - Problem indexing text with spelling enabled in Perl

Hi All,

I'm using the TermGenerator::index_text() on version 1.0.4 with the  
FLAG_SPELLING turned on, because the new spelling suggestion stuff  
seems awesome, but I'm getting a segv.

(gdb) bt
#0  0xb7ae153c in Xapian::WritableDatabase::add_spelling  
(this=0xa553988, word=@0xbff97724, freqinc=1) at ./include/xapian/ 
base.h:154
#1  0xb7becf47 in Xapian::TermGenerator::Internal::index_text  
(this=0xa553970, itor       {p = 0xab2d69f " North Face Windwall 1 Jacket
boys", end =
0xab2d6c1 "", seqlen = 1}, weight=3, prefix=@0xbff977ac,  
with_positions=true)
     at queryparser/termgenerator_internal.cc:207
#2  0xb7bebf0c in Xapian::TermGenerator::index_text (this=0x9b12d68,  
itor=@0xbff9779c, weight=3, prefix=@0xbff977ac) at queryparser/ 
termgenerator.cc:90
#3  0xb7c6b6d6 in XS_Search__Xapian__TermGenerator_index_text ()  
from /usr/local/lib/perl5/site_perl/5.8.8/i686-linux/auto/Search/ 
Xapian/Xapian.so
#4  0x080acc85 in Perl_pp_entersub ()
#5  0x080ab9ae in Perl_runops_standard ()
#6  0x08060ed0 in perl_run ()
#7  0x0805dc19 in main ()

It seems like there's a ref counting problem with the term, but its  
in C++ templates and I can't quite figure out what's wrong.

Any ideas to try?

I'm not quite yet re-subscribed to the list yet, but I'm working on  
it, so please reply directly for now.

Thanks,

Rusty

Please don't Cc: me when mailing the lists - I do read the lists, and it
won't get you a reply any quicker.

On Tue, Nov 13, 2007 at 05:38:43PM -0700, Rusty Conover
wrote:
> I'm using the TermGenerator::index_text() on version 1.0.4 with the  
> FLAG_SPELLING turned on, because the new spelling suggestion stuff  
> seems awesome, but I'm getting a segv.
Hmm, TermGenerator::set_flags() doesn't appear to be wrapped for Perl
yet, so how are you doing this?  Have you patched Search::Xapian?  If
so, please contribute the patch!
> (gdb) bt
> #0  0xb7ae153c in Xapian::WritableDatabase::add_spelling  
> (this=0xa553988, word=@0xbff97724, freqinc=1) at ./include/xapian/ 
> base.h:154
> #1  0xb7becf47 in Xapian::TermGenerator::Internal::index_text  
> (this=0xa553970, itor>       {p = 0xab2d69f " North Face Windwall 1
Jacket boys", end =
> 0xab2d6c1 "", seqlen = 1}, weight=3, prefix=@0xbff977ac,  
> with_positions=true)
>     at queryparser/termgenerator_internal.cc:207
> #2  0xb7bebf0c in Xapian::TermGenerator::index_text (this=0x9b12d68,  
> itor=@0xbff9779c, weight=3, prefix=@0xbff977ac) at queryparser/ 
> termgenerator.cc:90
> #3  0xb7c6b6d6 in XS_Search__Xapian__TermGenerator_index_text ()  
> from /usr/local/lib/perl5/site_perl/5.8.8/i686-linux/auto/Search/ 
> Xapian/Xapian.so
> #4  0x080acc85 in Perl_pp_entersub ()
> #5  0x080ab9ae in Perl_runops_standard ()
> #6  0x08060ed0 in perl_run ()
> #7  0x0805dc19 in main ()
I can't really say from that backtrace as inlining has muddied the
location of the SEGV.  Do you have an example which reproduces this?

Cheers,
    Olly