1. Changes to casting / OS X fixes
- Posted by jaygade Jul 11, 2013
- 1284 views
Hopefully I didn't break anything much. I've only tested 32-bit on OS X but now I'm only failing three tests:
t_io.e
t_net_http.e
t_socket.e
They fail eui, eub, and euc.
I should have tested some more, and I will. I need to work better on committing my updates to my working directory and working on one problem at a time.
If I DID break anything, it might be the ARM port, although I think I'm on the right track for my fixes. I really tried to leave the ARM sections alone, but I'm really hoping that the track I'm on will eliminate some (or all) ARM-specific hacks for converting between objects and doubles.
Next I will compile with -mno-sse turned off for OS X 32-bit. After that, I have ideas for changing the build system and refactoring a lot of things, to make them cleaner and more correct.
Oh, and based on our previous discussion, the D_*_bits() functions now work on integers stored as doubles, so they should be good up to 2^53. I need to make a unit test to prove that.
2. Re: Changes to casting / OS X fixes
- Posted by jaygade Jul 11, 2013
- 1307 views
Crap. Looks like I broke it anyway. 64-bit was doing good up to this point; but now it fails. Don't use the latest tip yet.
3. Re: Changes to casting / OS X fixes
- Posted by jaygade Jul 12, 2013
- 1251 views
I think that I figured it out. I'll update tomorrow evening (PST).
4. Re: Changes to casting / OS X fixes
- Posted by jaygade Jul 12, 2013
- 1240 views
Okay, updated. On OS X, I'm failing the "normal" failures (t_io.e, t_sockets.e, t_net_http.e). On 64-bit I'm also failing t_rand.e which is my fault I think.
I have more casting changes, as well as build system changes. Should I just make a new branch? My main aims are portability and correctness. I'm not adding any new features, just refactoring.
