[#CORE-2582] Infinity from UDF should be trapped by the engine

Firebird Core

Infinity from UDF should be trapped by the engine

Created: 05/Aug/09 04:02 AM Updated: 28/Jul/11 09:43 AM

Component/s:

None

Affects Version/s:

2.5 Beta 1, 2.5 Beta 2

Fix Version/s:

2.5 RC1

Time Tracking:

Not Specified

Environment:

Windows XP

Issue Links:

Relate

This issue is related to:

~~QA-410~~ Test for CORE-2582

Planning Status:

Unspecified

Description

« Hide

This is related to ~~CORE-2581~~ Infinity should not escape from the engine
In only tested FB2.5, but maybe FB2.1 behaves the same.
In Windows, there's a __try/__catch block around UDF execution to trap several special conditions. Normally the answer from these conditions is engine shutdown.
I think that in case of overflow, we don't need to panic. We simply need to convert it to an exception about FP overflow.
But I wonder why the protection block used by the Windows build is not working: it should invoke ISC_exception_post() in isc_sync.cpp.

SQL> declare external function dPower
CON> double precision by descriptor, double precision by descriptor,
CON> double precision by descriptor
CON> returns parameter 3
CON> entry_point 'power' module_name 'fbudf';
SQL> select dpower(10e+110, 3) from rdb$database;

DPOWER
=======================
Infinity

C.

All

Comments

Work Log

Change History

Version Control

Subversion Commits

Sort Order:

[ Permalink | « Hide ]

Dmitry Yemanov added a comment - 05/Aug/09 05:35 AM

As far as I can see, the CRT function POW() used in that UDF simply doesn't throw any exception, so there's nothing to catch.

[ Permalink | « Hide ]

Claudio Valderrama C. added a comment - 05/Aug/09 06:00 AM

In this case, Infinity will have to be detected in fun.epp directly.

[ Permalink | « Hide ]

Claudio Valderrama C. added a comment - 06/Aug/09 06:58 AM

Given these declarations:

DECLARE EXTERNAL FUNCTION DPOWER
DOUBLE PRECISION BY DESCRIPTOR, DOUBLE PRECISION BY DESCRIPTOR, DOUBLE PRECISION BY DESCRIPTOR
RETURNS PARAMETER 3
ENTRY_POINT 'power' MODULE_NAME 'fbudf';

DECLARE EXTERNAL FUNCTION XASIN
DOUBLE PRECISION
RETURNS DOUBLE PRECISION BY VALUE
ENTRY_POINT 'IB_UDF_asin' MODULE_NAME 'ib_udf';

DECLARE EXTERNAL FUNCTION XDIV
INTEGER, INTEGER
RETURNS DOUBLE PRECISION BY VALUE
ENTRY_POINT 'IB_UDF_div' MODULE_NAME 'ib_udf';

the expected answer from the engine is:

SQL> select dpower(1e120, 3) from rdb$database;

                 DPOWER
=======================
Statement failed, SQLSTATE = 39000
expression evaluation not supported
-Floating point overflow in result from UDF DPOWER
SQL> select xasin(2) from rdb$database;

                  XASIN
=======================
Statement failed, SQLSTATE = 39000
expression evaluation not supported
-Invalid floating point value returned by UDF XASIN
SQL> show functions;
       DPOWER XASIN
       XDIV
SQL> select xdiv(10, 0) from rdb$database;

                   XDIV
=======================
Statement failed, SQLSTATE = 39000
expression evaluation not supported
-Floating point overflow in result from UDF XDIV

[ Permalink | « Hide ]

Pavel Cisar added a comment - 28/Jul/11 09:43 AM

QA test added.