[postgis-tickets] [PostGIS] #2260: Benchmarking speed between built-in tiger normalizer and pagc_address_parser

[PostGIS]

#2260: Benchmarking speed between built-in tiger normalizer and
pagc_address_parser
---------------------------------+------------------------------------------
 Reporter:  robe                 |       Owner:  robe         
     Type:  task                 |      Status:  new          
 Priority:  medium               |   Milestone:  PostGIS 2.1.0
Component:  pagc_address_parser  |     Version:  trunk        
 Keywords:                       |  
---------------------------------+------------------------------------------

Comment(by woodbri):

 I'm at a loss on this one.

 This works fine on Linux.
 This crashes on Windows. I seems that the crash is not related to the code
 below, but rather something that has happened prior to this point and the
 server is just now recognizing the issue. I say this because it does not
 always crash at the same point just near this location. So sometimes a few
 statements before this. This is the only process using the database other
 than autovacuum.

 When I allocate memory for std, I copy the pointer to _gp_std_ just so I
 could check if it is getting changed and it does not look like it is. So
 the big picture is:

 {{{
 std = calloc()
 do stuff
 free(std) -- crash!
 }}}

 Here is the problem code:

 {{{
 void std_free(STANDARDIZER *std)
 {
     STANDARDIZER *tmp = std;

     DBG("In std_free -----------------");
     if (std != _gp_std_) {
         DBG("!!! std != _gp_std_ !!!");
     }

     if ( std == NULL ) return;
     DBG("Calling close_stand_process");
     if ( std -> pagc_p != NULL ) close_stand_process( std -> pagc_p ) ;
     if ( std -> pagc_p -> process_errors != NULL ) {
         DBG("Calling close_errors");
         close_errors( std -> pagc_p -> process_errors );
         DBG("Calling FREE_AND_NULL");
         FREE_AND_NULL( std -> pagc_p ) ;
     }
     DBG("Calling close_stand_context");
     close_stand_context( std -> misc_stand );
     DBG("Calling free");
     if (tmp != std) {
         DBG("std trashed");
     }
     if (std == NULL) {
         DBG("std is null");
     }
     free( std );
     std = NULL ;
 }
 }}}

 And here is my output

 {{{
 $ psql -U postgres -h localhost -f ../a test1
 psql:../a:1: NOTICE:  In std_free -----------------
 psql:../a:1: NOTICE:  Calling close_stand_process
 psql:../a:1: NOTICE:  remove_default_defs(__pagc_global__)
 psql:../a:1: NOTICE:  destroy_rules(__pagc_global__->rules) ;
 psql:../a:1: NOTICE:  destroy_rules 1
 psql:../a:1: NOTICE:  destroy_rules 2
 psql:../a:1: NOTICE:  destroy_rules 3
 psql:../a:1: NOTICE:  destroy_rules 4
 psql:../a:1: NOTICE:   destroy_rules 5
 psql:../a:1: NOTICE:  destroy_lexicon(__pagc_global__->addr_lexicon)
 psql:../a:1: NOTICE:  destroy_lexicon: i=7561
 psql:../a:1: NOTICE:  leaving destroy_lexicon
 psql:../a:1: NOTICE:  destroy_lexicon(__pagc_global__->poi_lexicon)
 psql:../a:1: NOTICE:  destroy_lexicon(__pagc_global__->gaz_lexicon)
 psql:../a:1: NOTICE:  destroy_lexicon: i=7561
 psql:../a:1: NOTICE:  leaving destroy_lexicon
 psql:../a:1: NOTICE:  Calling close_errors
 psql:../a:1: NOTICE:  Calling FREE_AND_NULL
 psql:../a:1: NOTICE:  Calling close_stand_context
 psql:../a:1: NOTICE:  Calling free
 psql:../a:1: server closed the connection unexpectedly
         This probably means the server terminated abnormally
         before or while processing the request.
 psql:../a:1: connection to server was lost
 }}}

 I also tried your idea of changing palloc calls to SPI_palloc because the
 later reverse to palloc if it is outside an SPI context which seems safe,
 but as you found out it did not help.

 I have also run this under valgrind on linux a while back and it reported
 no issues. Since I have made some changes since then maybe I should run
 that again.

 I'm at a loss here. Does anyone have any good debugging tools under window
 that they could check into this with?

-- 
Ticket URL: <http://trac.osgeo.org/postgis/ticket/2260#comment:6>
PostGIS <http://trac.osgeo.org/postgis/>
The PostGIS Trac is used for bug, enhancement & task tracking, a user and developer wiki, and a view into the subversion code repository of PostGIS project.