Menu
▾
▴
[Firebird-devel] [ firebird-Bugs-730035 ] Firebird 1.5 integer handling not behaving correctly
|
From: SourceForge.net <no...@so...> - 2003-04-30 07:12:28
|
Bugs item #730035, was opened at 2003-04-30 10:44 Message generated for change (Comment added) made by skidder You can respond by visiting: https://sourceforge.net/tracker/?func=detail&atid=109028&aid=730035&group_id=9028 Category: Core Engine Group: Initial Bug Status: Open Resolution: None Priority: 9 Submitted By: Sean Preston (fireclaw) Assigned to: Nobody/Anonymous (nobody) Summary: Firebird 1.5 integer handling not behaving correctly Initial Comment: We have developed an application environment that has been in production for 2 years using Interbase as the backend. While testing the environment on Firebird 1.5 Alpha 5 and 1.5 RC1 we found there was a problem. On tables that have an integer ID as the first column that is filled by a generator value we noticed some strange behaviour. When the application is running Firebird crashes with the following errors: Windows - FB 1.5 RC1 - Application Error Linux - FB 1.5 RC1 - +++ killed by SIGSEGV +++ After about 11 man hours of investigation we found that there appears to be an incorrect handling of integers within complex transaction contexes when committing. We have tried to recreate some of our application environment in a standalone application to duplicate the bug but have failed so far but we are able to repeat the error without a problem within our environment. The application is developed in Delphi 5 Enterprise and IB Objects 3.6A Database software tested on and worked perfectly Linux - Interbase 6 (Open Source Version) Windows - Firebird 1.0.2 Database software tested om and crashed Linux - Firebird 1.5 Alpha 5 (CS) Linux - Firebird 1.5 RC1 (CS) Windows - Firebird 1.5 RC1 (SS) We have isolated a few integers that work and fail and upon looking at there binary representation it appears that there is a problem when the rolls over on the 16 bit. 101000001 0111111111111111 21069823 works 101000001 1000000000000000 21069824 fails 101110001 0111111111111111 24215551 works 101110001 1000000000000000 24215552 fails The problem does not appear if we change the datatype to be a numeric(13,2) or drop the column completely. We are more than willing to assist in testing the fix to this bug. We will continue to try and create a standalone test case but we don't too much hope of being able to recreate the exact conditions that are obviously in the environment. ---------------------------------------------------------------------- Comment By: Nickolay Samofatov (skidder) Date: 2003-04-30 11:12 Message: Logged In: YES user_id=495356 1) The best thing you could do (if you have any mean to reproduce the bug) is to perform your test with DEV_BUILD Firebird version and examine resulting "core" file with gdb. (stacktrace and related variables). 2) This is better to discuss problems of Firebird 1.5 (exceptionally regressions) in fb-devel first before posting them to tracker 3) your report is useless. It misses all important information. Database and connection dialect, statement that causes the problem, mean how you reproduce the problem, etc... It will have to de deleted as invalid if you don't add this information ---------------------------------------------------------------------- You can respond by visiting: https://sourceforge.net/tracker/?func=detail&atid=109028&aid=730035&group_id=9028 |
