On 2010-Jul-14 19:24:30 -0700, Robert Bradshaw <[email protected]> 
wrote:
>On Wed, Jul 14, 2010 at 1:21 PM, Dr. David Kirkby
><[email protected]> wrote:
>> Peter Jeremy asked me if I'd look at the cephes package, as he was having
>> trouble with it. It's only installed on CYGWIN, so 99% of people are not
>> going to see that cephes is broken.
>>
>> I hacked the package so that it would try to build on any operating system -
>> not just Cygwin.
>>
>> Anyway, it has a few serious problems.
>
>Do you have any better suggestions for c99 support on Cygwin?

The actual math libraries appear to build OK and pass the testing I
have done (at least on FreeBSD).  It appears that base package used
for cephes had deleted a number of source files used to build the
self-tests whilst the makefiles have been taken from a variant that
still expects the self tests to exist - hence the make failures only
affect the (lack of) self tests.

What is more concerning (to me) is that as far as David Kirkby and I
can determine, the building the cephes will always result in errors
that are not detected - and genuine errors would likewise go
undetected.  The failure to pick this up as part of trac #8780 is
indicates a worrying failure of the QA process within Sage and is a
good example of why all spkg-install files should have 'set -e' at the
top.

Likewise, there are no self tests - most versions of cephes come with
self test code so implementing this would be reasonably simple.

-- 
Peter Jeremy

Attachment: pgpumssyfNL1u.pgp
Description: PGP signature

Reply via email to