New defsystem 3.6i (Was Re: OpenMCL support Re: [Maxima] Installation of maxima-5.9.2 on Mac OS X)

Hi,

Here goes the summary of the failed test on OpenMCL.

Running tests in rtest3:
********************** Problem 70 ***************
Input:
block([fpprec : 35], ev(ans, x : 1, bfloat) - ev(ans, x : 0, bfloat))


Result:
4.5994649519620995822199830432826167b-1

This differed from the expected result:
5.779160182042402b-1

107/108 tests passed.

In addition, it fails trig related things:

Running tests in rtest_trig:
********************** Problem 55 ***************
********************** Problem 60 ***************
********************** Problem 61 ***************
********************** Problem 65 ***************
********************** Problem 69 ***************

I omitted the details of rtest_trig failures since they include rather
larger expressions.

Here is the tested environment:
iMac G5 (first generation)
Mac OS X 10.4.5
OpenMCL 0.14.3 (included in Fink)
Maxima from CVS tree (tagged RELEASE-5_9_3-BRANCH)

On 2006/02/26, at 8:50 AM, Robert Dodier wrote:

> Hello Yasuaki,
>
>> Let me add few words on openmcl support situation. With defsystem  
>> 3.6i
>> and recent patches I committed to CVS tree, openmcl certainly  
>> compiles
>> the maxima 5.9.3 branch, and it works. However, I still see some  
>> tests
>> do fail. Failed tests are unfortunately not limited to trig related.
>
> Can you summarize the problems which you see in OpenMCL ?
>
>> If SCL has been a supported platform till today and we keep its
>> support level as other platforms, I would recommend to stay with
>> current defsystem for 5.9.3.
>
> Agreed, changing defsystem is too big a change for the 5.9.3 branch.
> However if someone wants to work on the main branch
> it is OK with me.
>
> best,
> Robert Dodier