On 11/23/2015 08:58 PM, John Rose wrote: > On Nov 23, 2015, at 8:23 AM, Ivan Gerasimov <ivan.gerasi...@oracle.com > <mailto:ivan.gerasi...@oracle.com>> wrote: >> >> Though, it may be better to get yet another pair of eyes. >> >> One minor nit: In the tests, in the summary, it is written, "Test >> Integer.toString method*s*", but only one of the overloads is tested. > > Here's another nit in the tests. > This is supposed to "wiggle around" critical points, which I agree with. > But it only wiggles above: > > 39 while (base < Long.MAX_VALUE / 10) { > 40 for (int c = 1; c < 65536; c++) { > 41 buildAndTest(base + c); > 42 } > 43 base = (base == 0) ? 1 : base * 10; > 44 } > > I suggest: for (int c = -1<<15; c <= 1<<15; c++) > > You'll need to guard the call to buildAndTest to avoid negatives. > You could also start base at, say, 10000.
Okay, here it is (only tests changed): http://cr.openjdk.java.net/~shade/8136500/webrev.06/ Thanks, -Aleksey
