Re: build error
- Posted by mattlewis (admin) Apr 04, 2013
- 1610 views
mattlewis said...
At one point, I ran "make all" directly on the pi using a binary posted by rkdavis (IIRC). I didn't time it, but "all day" seems like a decent estimation.
Using a fresh interpreter and a translated eutest, it took about 33 minutes to run the interpreted tests. Down to these failures:
Test results summary:
FAIL: t_de_memory.e
FAIL: t_serialize.e
FAIL: t_socket.e
FAIL: t_c_slice_past_end_double.e
Files (run: 153) (failed: 4) (97% success)
There was a problem with poke4. In the C implementation, we were casting a double to an unsigned integer. When poking a large negative number, however, this would fail on ARM. So something like -0x4000_0000 was communicated as 0 to the back end.
I cannot say that this was causing any of Jacques' problems, but it was a serious problem.
Matt
