Re: Math::Pari in turmoil

[A complimentary Cc of this posting was sent to
sisyphus
<sisyphus359@xxxxxxxxx>], who wrote in article <9447a4da-5926-4ace-8be7-d59e82eb03fc@xxxxxxxxxxxxxxxxxxxxxxxxxxxx>:
Math::Pari depends on (auto-)download of pari-2.1.7.tgz.

I have to ask (mainly out of curiosity) - do you have any plans to
change that ? Or will Math::Pari, from here to eternity, always depend
on pari-2.1.7 ?

I gather the "update" to 2.3.4 support must be not-all-that-trivial,
or you would have done it already.

2.3.4 works with a 3-line patch, 5 failing subtests, and missing
"high-level" APIs.

(I got a lot of unreadable build failure reports from OS X people,
who, of course, do not know how to read the docs. So it may be
different on OS X...)

One of failing subtest may be hard to trace (numeric integration).
The rest, I expect, are more or less trivial. Missing API may be
tricky since GP/PARI people removed my patches...

I have tried (and failed) to build Math-Pari-2.010800 against 2.3.4.
The 'make' process runs fine - but there are numerous 'make test'
failures (on perl-5.10.0, btw), most of which don't obviously relate
to the fact that I've attempted to build against 2.3.4. Afaict, most
of the failures appear to be purely perl issues. Building against
2.1.7 works flawlessly, however.

One may need to manually apply some patches from `patches' directory [+],
and people report that this user's advice:

*) 2.3.4: I solved it by adding these lines on paricfg.h:

#define PARI_VERSION_CODE 131844
#define PARI_VERSION(a,b,c) (((a) << 16) + ((b) << 8) + (c))
#define PARI_VERSION_SHIFT 8

helps them too.

[+] Math::PariBuild::patches_for() method uses

'2.1.7' => [ ($^O =~ /darwin/i ? 'patches/diff_2.1.6_no-common' : ()),
'patches/patch-pari-unnormalized-float',
'patches/diff_2.1.7_-O',
'patches/diff_2.1.7_restart'],

I expect the darwin stuff should be propagated to 2.3.* as well
(since it is a bug in configuration of gcc).

Hope this helps,
Ilya
.

Relevant Pages