[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: [help] modern packaging of multi-precision scientific library



Hi Julien,

2014-09-03 11:18 GMT+01:00 Julien Puydt <julien.puydt@laposte.net>:
Hi,

Le 03/09/2014 10:24, Ghislain Vaillant a écrit :

Hi everyone,

I am currently working on packaging 2 libraries (potentially a 3rd one too)
which supports multi-precision arithmetic (single, double and long double),
similar to what FFTW does. So from the source package, one can build
different binary package corresponding to each precision. For each
precision, the configure flags and library suffix need to be changed to
produce each version of the library with non-clashing names. Since, the
source package uses autotools, this can be achieved with dh-autoreconf I
suppose.

I could take a similar d/rules as in FFTW but it looks quite dated. I am
sure the package build process could be somewhat modernized and simplified
using dh commands and overrides. However, I am not sure which approach
would be possible or best.

Thoughts anyone ?

I think it could interest upstream ; you could see how things are done in palp's GNUMakefile.am and propose a patch upstream:
 http://anonscm.debian.org/gitweb/?p=debian-science/packages/palp.git

Snark on #debian-science


I fail to understand what you mean. My problem is not with upstream. Upstream
does provide all the necessary stuff to generate different versions of the library for
different precisions. In fact, upstream's build system is heavily inspired from the
FFTW one. You can do something like:

configure && make && make install --> lib<name>.so --> lib<name>-double.deb
configure --enable-single --suffix=f && make && make install --> lib<name>f.so --> lib<name>-single.deb
configure --enable-double --suffix=l && make && make install --> lib<name>l.so --> lib<name>-longdouble.deb

My issue is with being able to translate the build of the different versions of the
library for each supported precision in modern styled d/rules.

Does it sound clear ?

Ghis

Reply to: