StackOverflow in weblogic.jdbc.base.BaseSQLParser

Question

Actually the method parseAlmostAnything throws a StackOverflow exception
when a sql statement containing many ? is given.
For example, using the weblogic 8.1SP2 informix driver, the following SQL request
generates a StackOverflow :
select snwcservic0_.X1624_0_cmMo as x0_0_, snwcservic0_.X1624_1_cmMo as x0_1_,
snwcservic0_.X1607_17_argOs as x1_0_, snwcservic0_.X1607_18_argLayer as x1_1_,
snwcservic0_.X1607_19_argClass as x1_2_, snwcservic0_.X1607_20_sistentId as x1_3_
from snwcservicevers snwcservic0_ where (snwcservic0_.X1624_1_cmMo in(? , ? ,
.. (there are exactly
3386 ? in the request) ...))
There are exacly 3386 ? in this request. The IBM informix driver correctly handles
this request,
as long as there are fewer than 16000 or so ? in the request (think it's a SQL-I
message length
limitation of 64 K or something like), then it clearly tells us that the request
is too long.
didn't find how to log it as a bug, didn't have permissions in Create Case area
(I'm currently using an evaluation copy to test whether weblogic is the right
way to go).
thank you in advance for any advice (patch, workaround...). Using temp tables
instead of
in may work, but needs quite a rewrite of the part of the framework we are developing
which
deals with sql generation, and not sure of the comportment of informix if many
threads
concurrently create/delete table in temp space.
sincerely yours
Emmanuel Ligne

Answer 1

I generated a simple reproducer for this.
I will enter a bug report on it.
This seems to happen for all of the new type 4 drivers except for SQL
Server.
"Emmanuel Ligné" <[email protected]> wrote in message
news:4003d39b$[email protected]..
>
Actually the method parseAlmostAnything throws a StackOverflow exception
when a sql statement containing many ? is given.
For example, using the weblogic 8.1SP2 informix driver, the following SQLrequest
generates a StackOverflow :
select snwcservic0_.X1624_0_cmMo as x0_0_, snwcservic0_.X1624_1_cmMo asx0_1_,
snwcservic0_.X1607_17_argOs as x1_0_, snwcservic0_.X1607_18_argLayer asx1_1_,
snwcservic0_.X1607_19_argClass as x1_2_, snwcservic0_.X1607_20_sistentIdas x1_3_
from snwcservicevers snwcservic0_ where (snwcservic0_.X1624_1_cmMo in(? ,? ,
.. (there are exactly
3386 ? in the request) ...))
There are exacly 3386 ? in this request. The IBM informix driver correctlyhandles
this request,
as long as there are fewer than 16000 or so ? in the request (think it's aSQL-I
message length
limitation of 64 K or something like), then it clearly tells us that therequest
is too long.
didn't find how to log it as a bug, didn't have permissions in Create Casearea
>
(I'm currently using an evaluation copy to test whether weblogic is theright
way to go).
thank you in advance for any advice (patch, workaround...). Using temptables
instead of
in may work, but needs quite a rewrite of the part of the framework we aredeveloping
which
deals with sql generation, and not sure of the comportment of informix ifmany
threads
concurrently create/delete table in temp space.
sincerely yours
Emmanuel Ligne