Re: ARM Updates
- Posted by mattlewis (admin) Apr 08, 2013
- 2510 views
The poke / peek issue goes deeper than that, though, since it's a critical part of how the front end of the interpreter communicates with the back end.
Based on the fact that he was able to get to the point of failing with std/map.eissues during a translation attempt, I suspect that this part is working fine. That is, peek()/poke() doesn't work from Euphoria code run by the eubin, but the front-backend communication is functional.
It only fails for some values. For instance, if you have a pointer that's too big for a signed integer of the right size, you'll get a bad value. So if you have a big literal value in your code, it will show up as zero. This was why the tests for text:format() were failing.
I don't think this was the cause of the map issues, but it's hard to tell. Anyways, it's a serious problem, especially since it's not obvious why some bits of code fail.
Matt
