Re: [sdcc-devel] More on ppc uminus/muldiv problems

[Michael H. <mic...@ju...>]

Oh, and I just realised that 203 (not 209, oops) is 256-53.  The common
bugs with endianness are:
* De-referencing from the wrong pointer type.  Works on little endian, not
on big.
* Overlaying different types in a union and then referencing the wrong
type.

-- Michael

On Thu, 11 Oct 2001, Michael Hope wrote:

> We'll, I'm not very good at this but here is what I know sofar.  Attached
> is a simple test case and *.dump* outputs.  The code goes bad in the cse
> stage.  It seems that when the a = b == c line is optimised and the left
> hand value is corrupted.  When the evaluation in SDCCicode.c:984 happens
> it compares 209 with -53 instead of the real value with -53.
>
> It seems that the value is evaluated correctly earlier on, but during the
> comparison it isn't.
>
> Help.  Can anyone suggest where I should look?
>
> -- Michael
>