Hmmm... if one cannot easily use random(3) to implement SRFI 27 and/or
SRFI 194, they are probably too overspecified in one or the other
regard, aren't they?

Marc

Am Fr., 7. Aug. 2020 um 05:45 Uhr schrieb Alex Shinn <xxxxxx@gmail.com>:
>
> Sorry, Chibi goes to extraordinary efforts to make use of the system's random(3),
> which is highly platform-dependent, making accurately reusing the bits tricky.
>
> I should give up and just write a PRNG from scratch to ease maintenance.
>
> In the meantime I've fixed this and added some chi squared tests to guard
> against regressions in likely boundary cases.
>
> --
> Alex
>
> On Fri, Aug 7, 2020 at 6:21 AM Arvydas Silanskas <xxxxxx@gmail.com> wrote:
>>
>> Looks to me it's a regression on chibi, at least for one of the tests I investigated thoroughly. I opened an issue on github. I'll wait for response before investigating other test failures in case the underlying cause is the same
>>
>> 2020-08-06, kt, 22:03 Arvydas Silanskas <xxxxxx@gmail.com> rašė:
>>>
>>> Thanks, I'll have a look
>>>
>>> 2020-08-06, kt, 21:44 Wolfgang Corcoran-Mathe <xxxxxx@sigwinch.xyz> rašė:
>>>>
>>>> Hi all,
>>>>
>>>> A quick run of the tests on a recent Chibi build reveal some assertion
>>>> failures.  Here's the test footer output:
>>>>
>>>> > 2420 out of 2425 (99.8%) tests passed in 52.14049506187439 seconds.
>>>> > 5 failures (0.2%).
>>>> > 16 out of 19 (84.2%) subgroups passed.
>>>>
>>>> Test log attached.
>>>>
>>>> Regards,
>>>>
>>>> --
>>>> Wolfgang Corcoran-Mathe  <xxxxxx@sigwinch.xyz>
>>>>
>>>> "The art of doing mathematics consists in finding that special case
>>>> which contains all the germs of generality." --David Hilbert