Smalltalk › Pharo › Pharo Smalltalk Developers

Possible FFI issue

_‹ Previous Topic Next Topic _›

Classic

List

Threaded

2 messages Options

Diego Lont-2

Possible FFI issue

Hi all,

I am working with Friedrich on the ODBC issue, that the doubles aren't read correctly, and this is the point where I got by following the code:
in the class: 'ByteArray' the function "doubleAt:" is implemented as a FFI-Kernel primitive.
<primitive:'primitiveFFIDoubleAt' module:'<SqeakFFIPrims>'>
In a version based on the seaside one click (Pharo v 1.3) this results in the correct double, but in the Pharo one click v1.4, this results in a double that is way too small.

Writing a C program that makes the "same" api calls in ODBC also results in the correct values, although it gives an funny value for sqlTypLen (53). But we are kinda stuck here.

Can someone give us the code of FFI, so we can continue debugging here, and possibly making a fix here?
Also: does someone know what has changed between v1.3 and v1.4 in FFI, because both are loaded with the same Configuration.

Thanks in advance for all the help,

Diego Lont

DiegoLont

Re: Possible FFI issue

Hi all,

Ok, an update. We tried a different access driver, and this also solves the problem. So for now, we are helped enough.

Diego

On Jul 31, 2012, at 2:10 PM, Diego Lont wrote:

> Hi all,
>
> I am working with Friedrich on the ODBC issue, that the doubles aren't read correctly, and this is the point where I got by following the code:
> in the class: 'ByteArray' the function "doubleAt:" is implemented as a FFI-Kernel primitive.
> <primitive:'primitiveFFIDoubleAt' module:'<SqeakFFIPrims>'>
> In a version based on the seaside one click (Pharo v 1.3) this results in the correct double, but in the Pharo one click v1.4, this results in a double that is way too small.
>
> Writing a C program that makes the "same" api calls in ODBC also results in the correct values, although it gives an funny value for sqlTypLen (53). But we are kinda stuck here.
>
> Can someone give us the code of FFI, so we can continue debugging here, and possibly making a fix here?
> Also: does someone know what has changed between v1.3 and v1.4 in FFI, because both are loaded with the same Configuration.
>
> Thanks in advance for all the help,
>
> Diego Lont