On 22.09.2014, at 16:00, Henrik Johansen <henrik.s.johan...@veloxit.no> wrote:
> I am sceptical to the use of Floats. > Even using doubles, you've lost nanosecond accuracy when you get up to a > number of seconds corresponding to a year: > > 31536000.000000001 = 31536000.0 true > I don't see how the summing done in basicSeconds can possibly add up to > correct values… Hm, good point. The intent (as I understand) was to get nano second precision. So either we go back to seconds or we use nano seconds (as integers) instead of floats (under the assumption that this change actually will be integrated…) > > Cheers, > Henry > > On 18 Sep 2014, at 9:56 , Max Leske <maxle...@gmail.com> wrote: > >> Another issue for cleanup review. >> >> https://pharo.fogbugz.com/f/cases/13229/Year-arithmetic-broken >> >> >> Ported from Pharo 30. >> >> Made a few additional changes to #hash and #printOn: methods that no longer >> worked, added class comment to AnnualDuration. >> >> NOTE: I haven't done a full review on this and the change of seconds from >> integer to floats can have quite serious implications. Please look at this >> carefully. >> >> >> >> Name: SLICE-Issue-13229-Year-arithmetic-broken-MaxLeske.1 >> Author: MaxLeske >> Time: 18 September 2014, 9:50:41.500402 pm >> UUID: 44e5effc-4344-4aec-8cf5-3cb5d92ad73e >> Ancestors: >> Dependencies: Kernel-MaxLeske.1827, KernelTests-MaxLeske.678 >> >> Port from Pharo 30. >> >> - Added class comment to AnnualDuration. >> - adapted #hash and #printOn: methods that no longer worked because of >> seconds being floats >
