On May 19, 2006, at 6:45 PM, John Baxter wrote:

> * I find it hard to believe that Apple would use a different format  
> in Intel-land, but, since we don't know the format, that is possible.

The underlying cause of the problem has been revealed in the post  
from the Studio list.  Which reminded me that the bytes inside a  
signed 64-bit integer can look very different in Intel-land than they  
did in PPC-land.  And there are at least two rational ways to handle  
them, probably more.  There are several possibilities as to what  
Chris meant in the post as the instruction-level cause of the  
problem, but an unhappy mixture of two of those ways seems likely.

It does seem to be workaround time.

I think my question about "where did the reference date come from?"  
remains valid.

Did the script travel to the client with the problem as a compiled  
script containing the reference date?  If so, one thing to try would  
be to send it in text form and have it compiled on the client's  
machine (or compile it on a (borrowed if necessary, as it would be  
here) Intel machine and send that complied version).  The fact  
scripts created on Intel machines are running sensibly somewhat  
encourages these ideas.

   --John (who sees a scramble among suppliers of compiled scripts  
that deal with dates and have date values compiled in, in the near  
future if the final paragraph above leads to a solution)