Mysterious range error

rbuser

A report that has been working in production for 6 months. Added a new
field to the TADOquery. Deleted all the field definitions in the
TADOquery and then added all.

Added the field to the report. Run the report. Get a range error
(overflow) on a field. For a different selection criteria, get a range
error on a different field. The range error happens when the report
itself is called.

Put exceptions around the couple of events with custom code called by
the report. They don't fire.

Deleted the data pipeline and added it back. Double-clicked it so it
could add the fields.

Added a grid so I could see the data. The largest number is 33. The
field is a TBCDField with a precision of 16 and size of 0.

Still getting a range error. I don't see how or why. With one set of
selection criteria, I can get the first page of the report to show up in
the preview, but the second page bombs with the range error.

Where do you think I should look next?

nardmoseley

If the TField.FieldType is ftFMTBcd, then RB internally converts the type to
Double.

As a test try the following:

1. Modify your Delphi library path from RBuilder\Lib to RBuilder\Source

2. Add a TButton to a form and add the following code to the OnClick
event-handler:

var
ldValue: Double;

myPipeline.Open;

while not myPipeline.EOF do
try
ldValue := myPipeline.GetFieldValue('myFieldName');

if (ldValue > 10000) then
ShowMessage('The code is to fool the optimizer');

myPipeline.Next;

except
ShowMessage('Exception!!!');

end;


--
Nard Moseley
Digital Metaphors
http://www.digital-metaphors.com

rbuser

The test flips through all the records without error, but the report
still has the exception. I'm tracing through the
GetFieldValue(dataField) in function TppDBText.GetFieldValue: Variant;
in ppCTrls right now.

rbuser

I *think* -maybe- I've pinned down two actual bugs in rb:

1) Based on where the debugger indicates a problem in the source: when
there's an exception in getfieldvalue, somehow the wrong field is
reported. The exception here looks like it's happening in the
getfieldtype. However, none of this ties in with what is really
happening. What was really happening is that one field that is set on
the groupfooterband beforegenerate event, and that field cannot hold the
data, hence the overflow. But it's the not the field that's coming up
in the exception! The field that's showing in the exception is the
immediately previous field.

2) when you add a new pipeline to an open ado query, there's an apparent
bug in how the pipeline fields get created for TBCDField's. For a
TBCDField, according to the Borland help file and observation, the size
of the field is in the PRECISION property, and the number of decimals
are in the SIZE property. But the pipeline creates a TppField with a
size = size of the TBCDField, which is much smaller than the TBCDField
as a whole.

rbuser

Scratch #2 - it makes no difference what the TppField size is.

rbuser

This one field will not allow the assignment of any decimal places,
period. If I round the data before assigning in the beforegenerate
event, no problem.

I can't seem to find any differences between the form variable in RB on
the report design screen and others very similar that don't have the
problem, nor between the TppVariable in the pipeline and others very
similar that don't have the problem, nor between the ado query field
and others very similar that don't have the problem.

rbuser

Now that's not working either (rounding) - it was working a few minutes ago.

There's something very weird and unstable happening and I think it's
with the pipeline. I got an AV when I clicked on it, so I closed the
project and reopened, and all my workarounds no longer work. So I'll
try deleting the pipeline and adding it back again.

Needless to say, this has been a very tedious trial-and-error process of
trying to figure out what is happening in the "black-box". Shows the
limits of "having source" - if you don't know the product intimately,
tracing is just a maze and has not gotten me anywhere so far.

I can't wait for this to go pop and be over...

rbuser

Okay, here's the solution:

1 set all the components on the report with a pipeline property to
parentpipeline.

2 delete the pipeline.

3 Delete all the fields from the TADOQuery, and set it inactive.

4 Change the TADOQuery's ENABLEBCD property on the TADOQuery to false.
Set the TADOQuery active. Open the fields editor and add all the fields
back. Now you will not nave any TBCDField - you will now have TFloatFields.

5 Add the pipeline back on the form and name it what it was named
before. Double-click it and it will come up with all the fields newly
created. Close it.

6 Select the report and set the pipeline property to the pipeline you
just added back. Save it.

This fixes everything. Looks like RB doesn't "like" BCDFields, (at
least for assigning new values in groupheaderevents).

Thanks for the suggestions - those put me on the path to finding the
problem.

nardmoseley

Good to hear that found a solution.

E-mail support@digital-metaphors.com with steps on how we can recreate this
issue and we'll check it out.

--
Nard Moseley
Digital Metaphors
http://www.digital-metaphors.com

rbuser

Thanks for the pipeline field iteration idea. That's a good technique to
debug external to the reporting events, where it's much easieer.