In Cil.truncateInteger64, the variable 'chopped', which is used to see if truncation has occurred, is incorrectly set. It is currently set to
Int64.shift_right_logical i (64 - nrBits),
but it should be
Int64.shift_right_logical i nrBits.
In addition, the test for 'choppedness' seems a bit odd. For example, why should converting 0xffffffff80000000 and 0x80000000 to a C 'int' *both* return 'false' for whether truncation occurred? Here's a similar example: converting 0x80 to a signed char would return 'false' for whether truncation occurred, even though the input represented 128 and the output is -128.
I am using CIL 1.3.6.
Log in to post a comment.