[Xapian-discuss] Search performance issues and profiling/debugging search?

[Ron Kass]

Hi Olly

We have always been collapsing on a value (#4) in all tests till now.

so all valgrind output you so thus far were all with collapsing.


On the latest SVN, we ran  : ./configure CXXFLAGS='-O0 -g' 
--enable-assertions ; make; make install

Then ran search through gdb, got SEGV and ran full backtrace, here is 
the output..

http://www.pidgintech.com/other/fts/test/gdb_segfault_backtrace_full_noopt_assert.txt

Hope it helps.


Looking forward to hearing your insight on this..

Best regards,

Ron


Olly Betts wrote:

> On Sun, Oct 28, 2007 at 12:46:44PM +0200, Ron Kass wrote:
>   
>> Ok.. Good (sort of) news.. We managed to get the seg fault again..
>>
>> This time we will keep the DB and this time we got a gdb of such a run.
>>
>> http://www.pidgintech.com/other/fts/test/gdb_segfault_backtrace_full.txt
>>     
>
> This narrows done the error to MultiMatch::get_mset(), but the error
> location is reported as being in C++ headers (due to inlining I assume),
> so I only know it's during some std::string call in multimatch.cc.
>
> Can you rebuild without optimisation and with assertions and repeat
> this?
>
>   
>> (Note: this run is from a more complex test script we have, but the 
>> error still comes from the same action - getting the mset. We will try 
>> soon to recreate it with the simpler test script)
>>     
>
> It looks like you're collapsing on a value.  Was the previous time this
> occurred also when collapsing?
>
> Also, are the valgrind errors when collapsing?
>
> Cheers,
>     Olly
>