I've just returned to a project after a length absence (which will include one or two versions of ElevateDB) and I have a problem. I have a field in a table defined as DECIMAL(19,2). It shows up as a defined field in D2007 as TBCDField. Prior to leaving off the project I had things working pretty well. The form in which I updated the table in question worked (I'm returning to do some documentation - shudder). Now it doesn't . The statement things break down on is

 dm.Ledger_Value.AsCurrency :=  ietValue.AsFloat;

and the error is one I've never seen before and don't really understand.

Debugger Fault Notification
---------------------------
Project E:\MOA - DEV\MOA.exe faulted with message: 'datatype misalignment at 0x76f98658'. Process Stopped. Use Step or Run to continue.
---------------------------
OK
---------------------------

I've disabled triggers on the table in case they were what was causing the problem but no success. Tried installing a 2019 version of ElevateDB - no success

HELP


Roy Lambert

I've now tried repairing the table - no joy, deleting & re-creating again no joy

Roy Lambert

Roy Lambert wrote:

I've now tried repairing the table - no joy, deleting & re-creating again no joy


Hi Roy,

if I remember right, we had some problems with Decimal fields too.

I could check...
dm.Ledger_Value.AsBCD :=  ietValue.AsFloat;
or
dm.Ledger_Value.Value :=  ietValue.Value;

gripsware


Thanks for the suggestion but still not working. The puzzling thing is that this did work, even if I hadn't tried all functions in the program I couldn't miss this one because its the basic transaction insert / edit form.

Roy Lambert

I have  a 2 drive bay laptop and keep my W7 OS on a 120GB SSD - this gets cloned occasionally, the last time was 12/5/2021. I just briefly swapped the clone into my laptop  and the problem goes away. My quest now is what amongst the bits and pieces I've installed has caused the problem. I have only 17 items in programs & features installed after 12/5/2021 but I'm not sure where else I might need to look.


If anyone has any thoughts I'd appreciate them. Then again I might just revert back to the old clone


Roy Lambert

Finally, after c8 hours, I finally traced down the culprit (I think). I wrote code to load a custom font (shown below my sig), taking that out and the problem has gone.

Next to figure out why


Roy Lambert

unit NLHCharLoader;

interface

uses

Classes, Windows;

{$R nlhdings.res}

const

nlhCharClock = 33; // Time lookup
nlhCharTick = 34; // A tick
nlhCharDropDown = 35; // General dropdown / lookup
nlhCharTwoForms = 36; // Image lookup
nlhCharBlackCircle = 37; // Colour lookup
nlhCharOpenBook = 38; //Date & time lookup (its a diary)
nlhCharLightening = 39; // QuickLookup
nlhCharSquare = 40; // Unknown Date/Time type
nlhCharCalendar = 41; //Date lookup
nlhCharCalculator = 42; // calculator
nlhCharKeyboard = 43; // On screen calculator (the calculator symbol is naff)
nlhCharPhone = 44; // Dial phone number

nlhDingFont = 'NLHDings';

implementation

function LoadResourceFont(const ResourceName: string): boolean;
var
ResStream: TResourceStream;
FontsCount: integer;
HFont: THandle;
begin
ResStream := tResourceStream.Create(hInstance, ResourceName, RT_RCDATA);
HFont := AddFontMemResourceEx(ResStream.Memory, ResStream.Size, nil, @FontsCount);
Result := (HFont <> 0);
ResStream.Free();
end;

initialization
LoadResourceFont('NLHDings');
{
Note: This will run only once since the various instances of the unit are shared
     I'm not bothered about removing from memory since I'll want these characters
     available for the life of the program - they are part of the GUI
}

end.

Roy,

<< Finally, after c8 hours, I finally traced down the culprit (I think). I wrote code to load a custom font (shown below my sig), taking that out and the problem has gone. >>

That could just be a coincidence.  The font loading may just change the memory enough that it changes the symptoms of the problem or makes it go away.

My suggestion would be to run the application with FastMM4 and FullDebugMode turned on.  That will usually catch any memory overwrites/misbehaviors pretty close to when they happen.  Or, at the very least, they will give you some idea of where an issue may occur.





Roy Lambert

unit NLHCharLoader;

interface

uses

Classes, Windows;

{$R nlhdings.res}

const

nlhCharClock = 33; // Time lookup
nlhCharTick = 34; // A tick
nlhCharDropDown = 35; // General dropdown / lookup
nlhCharTwoForms = 36; // Image lookup
nlhCharBlackCircle = 37; // Colour lookup
nlhCharOpenBook = 38; //Date & time lookup (its a diary)
nlhCharLightening = 39; // QuickLookup
nlhCharSquare = 40; // Unknown Date/Time type
nlhCharCalendar = 41; //Date lookup
nlhCharCalculator = 42; // calculator
nlhCharKeyboard = 43; // On screen calculator (the calculator symbol is naff)
nlhCharPhone = 44; // Dial phone number

nlhDingFont = 'NLHDings';

implementation

function LoadResourceFont(const ResourceName: string): boolean;
var
ResStream: TResourceStream;
FontsCount: integer;
HFont: THandle;
begin
ResStream := tResourceStream.Create(hInstance, ResourceName, RT_RCDATA);
HFont := AddFontMemResourceEx(ResStream.Memory, ResStream.Size, nil, @FontsCount);
Result := (HFont <> 0);
ResStream.Free();
end;

initialization
LoadResourceFont('NLHDings');
{
Note: This will run only once since the various instances of the unit are shared
     I'm not bothered about removing from memory since I'll want these characters
     available for the life of the program - they are part of the GUI
}

end.

Tim Young
Elevate Software
www.elevatesoft.com

Got that wrong - after much messing about its not the culprit - the search continues

Roy Lambert

Tim


<<That could just be a coincidence.  The font loading may just change the memory enough that it changes the symptoms of the problem or makes it go away.>>

Yup I think you're right

<<My suggestion would be to run the application with FastMM4 and FullDebugMode turned on.  That will usually catch any memory overwrites/misbehaviors pretty close to when they happen.  Or, at the very least, they will give you some idea of where an issue may occur.>>

I run with FastMM normally but I don't have FullDebugMode on yet - will soon though. I do have another candidate for the cause as well.


Roy Lambert

Finally tracked it down to an alteration in a subclassed table component of mine - at least I hope so

Roy Lambert