Re: Problem with 2.3.3 testing ellglobalred and galois

[Bill Allombert <Bill.Allombert@math.u-bordeaux1.fr>]

On Fri, Apr 04, 2008 at 02:34:34PM -0400, David A. Desrosiers wrote:
> 
> >The 'fixes' are actually optional data packages that are required to 
> >run the tests. The test are in part a validation on the content of 
> >the data packages.
> 
> The packages have been unpacked into the tree, as indicated, creating:
> 
> 	/pari-2.3.3/data/elldata/
> 	/pari-2.3.3/data/galdata/
> 
> >Assuming the test failed because the optional packages where not
> >properly installed:
> >1) You need to run 'make install' before 'make check'.
> 
> Can't do that, because the install target actually tries to install 
> the software into the global $PREFIX specified (/usr/local/ if none 
> was specified). Can't install software that doesn't pass checks, of 
> course.

Well, I proposed a quick fix. If you prefer:

Run make install-data to install the optional packages, before running
test-all.

Actually PARI only require that galdata and elldata are installed before
you use them.

As for the 'of course': since elldata and galdata are totally optionnal,
of course there is no requirement that test-all pass before installing PARI, 
and even with them we never promised that test-all report success on correct
builds.

> >3) The only functions affected are polgalois (need galdata.tgz) and 
> >ellsearch and ellidentify (need elldata.tgz).
> 
> Right, so the "solution" for now is to run 'dobench' instead of 
> 'test-all', because that seems to work on all platforms we're building 
> this on. elldata and galois fail exactly the same on all platforms.

Exactly, you are supposed to use 'make bench' for basic sanity checks.

Cheers,
Bill.