[gobo-eiffel-develop] Problems with DT_TIME_DURATION due to 32-bit integer arithmetic

[Colin P. A. <co...@co...>]

I came across what initially looked like a very strange bug, but is
easily explained by arithmetic overflow:

Given a DT_DATE_TIME_DURATION with the following components:

year: 0
month: 0
day: 1234567890
hour: 1234567890
minute: 1234567890
second: 1234567890
millisecond: 125

Then millisecond_count returns a value of 45919373 instead of the true
figure of 450435271316490125.

Obviously, the example is very contrived (it comes from the W3C XSLT
test suite), but it makes me wonder if anything can be done about
this. There is no post-condition violation against millisecond_count,
as the post-condition calculation suffers the same fate as the body.

This particular example could be solved by changing the result type to
INTEGER_64, but that doesn't change the principle - that the result is
only meaningfull if no arithmetic overflow occurs.
Is there a way to detect this?
-- 
Colin Adams
Preston Lancashire