Re: Removal requests submitted for CERNLIB packages
Hi Bastien,
On Wed, Mar 9, 2011 at 1:01 AM, roucaries bastien
<roucaries.bastien+debian@gmail.com> wrote:
> On Tue, Mar 8, 2011 at 8:05 PM, Andreas Tille <andreas@an3as.eu> wrote:
>> On Tue, Mar 08, 2011 at 03:25:02PM +0100, Bastien ROUCARIES wrote:
>>> I have just created a cernlib.git under debian science with dsfg clean source
>>
>> Great. Just ping the list once you run into any blockers or if you need
>> a sponsor.
>
> My goal is to refactor cernlib support and to extract 64 bits ugliness
> to some libraries.
>
> I have only interest in mathlib and I suppose they are safe (ping kevin)
I'd think (hope) that mathlib would be 64-bit safe in nearly all
cases. Note that the existing (just-removed) packages run a set of
test suites from debian/rules, and the debian build errors out if the
test suite fails. The test suite for mathlib was pretty extensive.
The only thing that might be a potential issue would be if any of the
Fortran functions take an array by address as input.
I seem to remember there were a few functions in mathlib that took
multiple arrays as input, then calculated the element-by-element
differences (?) using some other function, to which was passed the
difference between pointer addresses of the two arrays. That may not
be 64-bit safe. I'm afraid my memory on this is very hazy. It might
possibly be addressed already in one of the existing debian/patches.
best regards,
--
Kevin B. McCarty <kmccarty@gmail.com>
WWW: http://www.starplot.org/
GPG: public key ID 4F83C751
Reply to: