Sergey, 2015-04-03 15:16 GMT+02:00 Sergey Bylokhov <sergey.bylok...@oracle.com>:
> Hello. > Is that a problem that with the new code we can get OOM earlier in some > cases? Should we take care of overflow more carefully now? It seems that > ArrayList.grow contains similar logic. > > I advocate I did not test with a small heap (Xmx32m ...) I agree my patch will waste more memory : 1/8th for both arrays (byte[] and float or double[]) instead of 500 byte and 1000 float/double values ! I looked at ArrayList.grow(int) and it deals with integer overflow ie Integer.MAX_VALUE ... I did not imagine such big paths ! But you're right: my code can be improved to deal with OOME and overflow. In case of OOME, it may be possible to try allocating another array with a smaller grow ? Is it what you had in mind ? Cheers, Laurent
