Presumably with option (2) we would still run the Harmony Classlib and DRLVM test suites as part of the build? If so, then (2) would be my preference.
2008/11/12 Aleksey Shipilev <[EMAIL PROTECTED]>: > Tim, I see the good point in your explanation too. > > So we need to consider three options: > Option 1. Go with r711744 as M8. It is already tested, so just solidify > build. > Option 2. Fix H6013, declare r711744 + H6013 as M8, presume the > impact locality, solidify the build. > Option 3. Fix H6013, declare r711744 + H6013 as M8, re-spin the > tests, solidify the build. > > I'm voting for (3). I would be glad to be proved wrong on my concerns, > actually I would be pleased with that :) > Maybe just arrange a vote again? > > Thanks, > Aleksey. > > On Wed, Nov 12, 2008 at 3:39 PM, Tim Ellison <[EMAIL PROTECTED]> wrote: >> Aleksey Shipilev wrote: >>> On Wed, Nov 12, 2008 at 2:17 PM, Tim Ellison <[EMAIL PROTECTED]> wrote: >>>> Can you think of a situation when the null check will introduce some >>>> instability or regression? >>> I actually persuaded by Chunrong's point -- that's double checking, so >>> no problems should occur. >>> >>> As for introducing new bugs, consider the issue described in >>> HARMONY-6013 is really covering some other deadly issue. Consider the >>> workload where NPE is not firing because of H6013, >> >> ...but the test doesn't silently work without the NPE, it causes a trap. >> >> So we know that our tests don't currently cover the situation where we >> would now expect to get a NPE, or they would be trapping today, right? >> >>> so after H6013 gets >>> fixed the control flow in that workload is going differ than in tested >>> M8. As many uses of the helper, as many the chances the control flow >>> differs. Having that, we can't say the change is minor. >> >> I appreciate that the code will appear in many places, but I think it is >> localized and we know the situation doesn't occur in current testing. >> >> That said, I'd rather run the two days + testing again rather than spend >> two days arguing about it :-) >> >>> If I will be >>> able eventually to say that similar changes are "limited >>> impact"-issues, then you should employ me as oracle tester <g> :) >> >> lol >> >>> Of course, that's the speculation if this is actually a double null >>> checking. >>> I just want not to guess while talking about milestones. >> >> ack - like I said, if people think we should re-spin the build and >> retest, then I'm ok with that too. It would be the conservative approach. >> >> Regards, >> Tim >> > -- Unless stated otherwise above: IBM United Kingdom Limited - Registered in England and Wales with number 741598. Registered office: PO Box 41, North Harbour, Portsmouth, Hampshire PO6 3AU
