Only one first row is returned (Lazarus + Firebird + Oracle)
lazarus / 16-Databases

----------------------------------------------------------------------
krzynio · 2018-12-20 22:39

Hello!
I have been working with Lazarus trunk 2.1 on Debian testing and on
Windows 10
plus Zeoslib 7.2.1-rc1 and Firebird 3.0 and Oracle
Recently I updated to Zeoslib 7.2.4-stable.
After this some queries on Lazarus return only one first row.
I went back to 7.2.3-rc and the same, only first row is returned.
The same query on Zeoslib 7.2.1-rc1 or under Flamerobin returns all
rows.
Where can be the problem? Is it a bug?
Any suggestions, please?

----------------------------------------------------------------------
madref · 2018-12-24 11:02

Do you have Booleans stored as 0 and -1?
If so then look if the new one stores them as 1.


I had that problem once

----------------------------------------------------------------------
Thaddy · 2018-12-24 11:48

Indeed, but this Boolean issue should be fixed in trunk.

----------------------------------------------------------------------
Zoran · 2018-12-24 18:09

I experienced this with some query couple of days ago, and worked
around it by redisigning the query.
I don't think I can reproduce the original query, but I am quite sure
I didn't have any boolean fields at all.

I think it is a Zeos bug. Please report it in their bugtracker or in
their forum.
I don't have time now, but if noone reports it, I will see tomorrow if
I can make an example and report it there.

----------------------------------------------------------------------
madref · 2018-12-27 12:26

Then use in ALL your queries where you use boolean the function ABS.


You get something like
Code: [Select]SELECT * FROM Some_Table WHERE ABS(MyBoolen) = 1

----------------------------------------------------------------------
Thaddy · 2018-12-27 14:34

The bug is that a C Bool is not false which is defined as any value
except false ( = 0 ) and a Pascal Boolean (all types) were defined as
either  0 or 1. This has been corrected. There is now a correct Bool
type to interface with C code in trunk. This was an FPC bug ( for a
relatively short - some months, a year - period only) and not a Zeos
bug. Maybe Zeos needs to switch to the new Bool type, though.
A SQL query with not false instead of true should always work, though.
But  FPC booleans (pascal booleans) are not suited for interfacing
with C code as most database engines are written in C. The Bool type
now adheres to that, as was also previously the case. It was a rather
serious regression that made me at some point very angry. (Actually
real anger, not grumpy, I seldom have such strong negative vibes...
to paraphrase my fellow tank commander from Kelly's heroes. (odd
ball))