Mark Cave-Ayland wrote:
At least now if I find similar errors, I have an idea of how to fix them :)
In terms of related errors, it now looks as if backwards branches (b<mark)...(b?branch) are broken in a similar way. I'm still struggling at the moment to write some code that works (I can't seem to get setup-tmp-comp/execute-tmp-comp to nest correctly at the moment), so I may have some more questions related to this coming up over the next few days.
However, the one thing I did find is that fcode-offset is broken for backwards branches, since it currently treats the 8-bit or 16-bit offset as unsigned - which is not helpful when trying to execute backwards branches. Please find attached a patch that does the right thing here - can someone review and commit this to SVN?
Many thanks,
Mark.