<snip>
>As I previously state, Even I put that statement at VFP command prompt when just starting VFP! No one know the command request is or not over the OS nor VFP error! 2nd, as previous state:
>ww(_screen) && no error
>ww(NVL(_screen, .NULL.)) && no error
>ww(NVL(.NULL., _screen)) && OS Error....
>
>I think it would be the internal VFP500.DLL handling problem!
>If only OS Error, why 2nd statement work fine??!!
>
>>When you call this test function with an argument of NVL( .NULL., _SCREEN), you get a program error (usually on the
>> second time you call it). If you make the same call, except with a form object (that is, o = createobject( "form")), it works fine.
>> An argument of NVL( _SCREEN, .NULL.) works fine too! That's wierd.
>>
>> - Mark
>
>
>I agree with Mark state, it seem NVL() got problem on requesting OS to task.
>For programmer view, sometime afraid if huge error occur with only very simply
>command/function call because the reliability become uncertain now! >_<'
>
Actually, after a quick run through the debugger, the function fails on the RETURN, not on evaluation of the NVL(), which had to occur before your UDF code got control. The result of the NVL() is visible in your first parameter within your procedure.
The details of a page fault are something outside of the scope of an application error handler; what has happened is that an operating system service was unable to safely return control to the application environment - to protect the operating system and any other tasks active on the machine at the time, the application's environment was terminated by the operating system kernal. There simply is nothing to return control to in the application; the task is dead and buried before it could react to a fault. VFP died when the fault occured, and no error handler within the VFP, outside of the protected mode environment of the kernal, could possibly get handed the result, any more than it could be (safely) expected to deal with other events like parity errors.
This doesn't explain why the fault occurs; it only is meant to describe why a VFP error routine can't deal with the error.
BTW, if you think this type of error is scary, don't write in C, ever. It's infinitely easier to break things in an unrecoverable fashion with simple code in C. And even easier with assembler.
Ed
