Math packages: givaro, fflas-ffpack, linbox

[tornaria]

General

• This is a new package and it conforms to the quality requirements

Have the results of the proposed changes been tested?

• I use the packages affected by the proposed changes on a regular basis and confirm this PR works for me

Notes

• This includes: givaro, fflas-ffpack, linbox.
• Everything is tested and check pass in x86_64, x86_64-musl, i686.
• Except gf2x, everything is nocross.
• I tried to make sure everything works with baseline x86_64 as defined here, essentially it means SSE2 is ok, but SSE3 is not.
• I added a general option native, which defaults to off. When turned on, it will let the libraries compile for the build machine which often results in significant improvements.

EDIT: split in smaller independent PR to ease review

[dkwo]

Great job, I will test soon.

[tornaria]

Nothing relevant changed: I cleaned up some leftover lines in common/shlib.

[dkwo]

Looks good to me.

[Chocimier]

I would rather remove option and make in unconditional. People are free to use XBPS_CFLAGS="-march=native" to get optimized build. Works for this package, should work for others too.

[ericonr]

I think the correct variable is XBPS_TARGET_CFLAGS? At least that's what's used in the build-profiles (the README there needs to be fixed, btw).

I agree that that's a better way of solving it, but it would be nice to make sure it's working. Also see #29732

[Chocimier]

I mean XBPS_CFLAGS in etc/conf

[tornaria]

It might work in this case (assuming the only effect of the --enable-sse* configure flags is to add -msse* to CFLAGS, I'd have to check).

However, some math packages have custom written asm for different architectures so selecting build option "native" will not necessarily be equivalent to just using -march=native.

Another point is that using a build option will make an annotation in the xbps, e.g.:

$ xq givaro | grep native
build-options: ~native 

vs.

$ xq givaro | grep native
build-options: native 

[dkwo]

@tornaria @ericonr Btw, don't forget this one ;)

[tornaria]

@ericonr: rebased with only style changes. Tested with sage-9.4.rc2.

[tornaria]

Latest changes:

• renamed fflas_ffpack to fflas-ffpack since that is the right package name (the tar.bz2 has the wrong name but the tar.gz has the correct one).
• added a patch from Use std::ptrdiff_t for vector iterator difference type linbox-team/linbox#274 which fixes a bug on i686; formerly we disabled a particular test for arch=i686 due to this, but now all tests pass.

[tornaria]

@ericonr: let's do math stuff one at a time. I propose to start with this one which is the older one I've got. I think I applied all the style changes you told me in #30032.

[dkwo]

@ericonr Do you think we can merge some of the Math packages PR, starting from this perhaps? otherwise, rebasing them can be a bit tiring exercise.