New issue 2904: Jit and the copy module causes miscompilation
https://bitbucket.org/pypy/pypy/issues/2904/jit-and-the-copy-module-causes
Björn Lindqvist:
See the following test code:
```
#!python
import copy
def eva(board, w):
height = len(board) - 1
width = len(board[0])
eb = copy.copy(list(reversed(board[0:height])))
ch = [0]*width
for c in range(width):
# Works if I add the assert!!!
#assert height == 20
h = height-1
while h > 0:
if eb[h][c] != 0:
break
h -= 1
ch[c] = h + 1
return max(ch)
board = [
[0, 0, 0, 0, 0, 0, 0, 0, 0, 0], [0, 0, 0, 0, 0, 0, 0, 0, 0, 0], [0, 0, 0,
0, 0, 0, 0, 0, 0, 0], [0, 0, 0, 0, 0, 0, 0, 0, 0, 0], [0, 0, 0, 0, 0, 0, 0, 0,
0, 0], [0, 0, 0, 0, 0, 0, 0, 0, 0, 0], [0, 0, 0, 0, 0, 0, 0, 0, 0, 0], [0, 0,
0, 0, 0, 0, 0, 0, 0, 0], [0, 0, 0, 0, 0, 0, 0, 0, 0, 0], [0, 0, 0, 0, 0, 0, 0,
0, 0, 0], [0, 0, 0, 0, 0, 0, 0, 0, 0, 0], [0, 0, 0, 0, 0, 0, 0, 0, 0, 0], [0,
0, 0, 0, 0, 0, 0, 0, 0, 0], [0, 0, 0, 0, 0, 0, 0, 0, 0, 0], [0, 0, 0, 0, 0, 0,
0, 0, 0, 0], [0, 0, 0, 0, 0, 0, 0, 0, 0, 0], [0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
[0, 0, 0, 1, 0, 0, 0, 0, 0, 0], [4, 0, 0, 1, 1, 0, 0, 0, 0, 0], [4, 4, 4, 1, 0,
0, 0, 0, 0, 0], [1, 1, 1, 1, 1, 1, 1, 1, 1, 1]
]
for x in range(50):
print(eva(list(board), [1.0, 1.0, 5.0, 2.0, 1.0, 1.0, 1.0]))
```
When running it with pypy it correctly prints 3 for about 40 loops. Then the
JIT runs and after that it prints 2 instead. Python ofc prints the right answer
3 all the time. pypy3 version 3.5.3 and python3 3.7.0.
_______________________________________________
pypy-issue mailing list
[email protected]
https://mail.python.org/mailman/listinfo/pypy-issue
