martinrb at google.com
Wed Jun 23 18:15:12 PDT 2010
On Tue, Jun 22, 2010 at 10:01, Alan Bateman <Alan.Bateman at oracle.com> wrote:
> Xueming Shen wrote:
>> The only possible cause of its failure on a particular platform is
>> timeout, this test takes long cpu time
>> on slow machine. I tweak it a little, it runs about 30% faster now. It
>> passes the last several jprt runs
>> on all platforms.
FindDecoderBugs was a very useful test, but IIRC the 5-minute runtime
is overkill for finding regressions, and you can't exhaustively test
all possible byte sequences anyways. It should be possible to use my
pet strategy of generating a random subset of interesting byte
// Can you spare a week of CPU time?
// testExhaustively(cs, tester, prefix, 3);
so the test could have a 10-second mode and a 10-minute mode.
But I'm not volunteering.
BTW, Thank y'all very much for making these test robustness
improvements. They are appreciated.
More information about the core-libs-dev