Pass -no-undefined flag to autotool based spkg

[Jean-Pierre Flori]

A bunch of spkg using autotools do not build shared libraries of Cygwin because -no-undefined flag is not passed to libtool:

• cddlib (update in #13026, we ship a lot of patches for that one, so if I report upstream for that particular issue, what for the others? opened #15872),
• eclib (fixed in #13325 with new upstream version),
• ecm (fixed upstream, see ​https://gforge.inria.fr/scm/viewvc.php?view=rev&root=ecm&revision=2304),
• givaro (updated in #9511 and #13164, try ​http://www.infres.enst.fr/~flori/sage/givaro-3.7.1.p0.spkg, reported upstream, got in, see ​https://forge.imag.fr/scm/browser.php?group_id=187),
• glpk (updated in #12703, reported upstream, some feedback then nothing, see ​http://lists.gnu.org/archive/html/help-glpk/2012-12/msg00008.html, my last answer does not appear (although it was sent to marco and the list which must be somehow filtered), new discussion at ​http://lists.gnu.org/archive/html/help-glpk/2014-11/msg00000.html),
• gsl (mentions of -no-undefined in src code but only for MinGW, reported upstream at ​https://savannah.gnu.org/bugs/index.php?37894, you can try ​http://boxen.math.washington.edu/home/jpflori/gsl-1.15.p1.spkg, this is #14096, bumped upstream one year later, fixed upstream: ​http://git.savannah.gnu.org/cgit/gsl.git/commit/?id=1d002ee931b6a19333416ba8c4c4cccbc009b6a9),
• iml (updated in #748, try non updated at ​http://www.infres.enst.fr/~flori/sage/iml-1.0.1.p15.spkg, not sure upstream still exists, opened #15839, actually fixed upstream with 1.0.4 update, see #16076),
• libfplll (see #13804 for a quick (different problem) fix and #12835 for the fact we ship a really old version, should update it before thinking about the problem here, fixed in new upstream version 4.0.3),
• linbox (updated in #12883, try ​http://www.infres.enst.fr/~flori/sage/linbox-1.3.2.p0.spkg, reported upstream, fixed, see ​https://groups.google.com/d/topic/linbox-devel/sHan2BB8HVM/discussion),
• lrcalc (fixed in #13839 which should be half upstream),
• mpfi (fixed upstream, see ​https://gforge.inria.fr/scm/viewvc.php?view=rev&root=mpfi&revision=656),
• ntl (broken and fixed in #11635, no real upstream anymore as far as I understand, in fact Shoup is working on NTL again and I've made him aware of that),
• libgap (guess we are more or less upstream, see #14038)
• gf2x (fixed in #15014, reported upstream, I'll provide a proper fix when I have time)

[Jean-Pierre Flori]

Updates for

• cddlib in #13026,
• eclib in #13325,
• ecm,
• givaro in #9511,
• glpk in #12703,
• gsl,
• iml in #748,
• libfplll,
• linbox in #12883,
• lrcalc,
• mpfi,
• ntl in #11635.

[Jean-Pierre Flori]

For the updated fixed linbox, you'll need first iml, and then updated fixed givaro together with dependencies at #12840 and #12841.

[Jean-Pierre Flori]

Wrong button...

[Jean-Pierre Flori]

The Linbox spkg from #12883 changed, so the one linked here is now outdated.

[Jean-Pierre Flori]

The linbox spkg is now rebased on top of #12883, but nothing is committed. By the way, the spkg there lacks an hg tag.

[Jean-Pierre Flori]

Spkg diff, for review only.

[Jean-Pierre Flori]

Reported upstream for givaro, mpfi, linbox, gsl.

[Jean-Pierre Flori]

Reported for glpk and libfplll as well.

[Jean-Pierre Flori]

Contacted upstream ecm as well.

[Jean-Pierre Flori]

Reported for lrcalc as well.

[Jean-Pierre Flori]

MPFI is fixed upstream, see: ​https://gforge.inria.fr/scm/viewvc.php?view=rev&root=mpfi&revision=656

Givaro is fixed upstream: ​https://forge.imag.fr/scm/browser.php?group_id=187

I'm working on lrcalc as now Sage is more or less upstream. (I guess IML will be the same.)

And I'll provide quickly a patch for upstream ECM and it should get merged.

Feedback from fplll which I guess will be fixed in a couple of weeks in the new version (so we'll have to update to get a shared library).

Feedback from GSL and GLPK as well.

No feedback from LinBox? yet.

[Jean-Pierre Flori]

I've open a ticket for lrcalc, see #13839. Spkg coming soon.

[Jean-Pierre Flori]

ECM is fixed upstream. See ​https://gforge.inria.fr/scm/viewvc.php?view=rev&root=ecm&revision=2304

[Karl-Dieter Crisman]

JP, if you could be sure to somehow indicate on tickets you open for this that this is the metaticket for them?

[Jean-Pierre Flori]

Replying to kcrisman:

JP, if you could be sure to somehow indicate on tickets you open for this that this is the metaticket for them?

I'll try :)

[Jean-Pierre Flori]

Libgap needs fixing as well.

[Jean-Pierre Flori]

libgap ticket at #14038

[Jean-Pierre Flori]

Tentative gsl fixed and dirty spkg which seems to let some doctests which failed before pass at ​http://boxen.math.washington.edu/home/jpflori/gsl-1.15.p1.spkg

[Dima Pasechnik]

As we observed on #14038, this change can break OSX. So these changes should be made CYGWIN-only, or non-OSX only. (I don't know if there is any advantage to use -no-undefined, e.g. performance-wise).

[Jean-Pierre Flori]

GSL was #14096.

[Jean-Pierre Flori]

libfplll has been fixed upstream in version 4.0.3.

[Jean-Pierre Flori]

#15014 is for gf2x, this prevents building NTL on Cygwin...

[Jean-Pierre Flori]

For the record, the spkg needing patches are as far as I know:

• iml
• glpk
• cddlib

[Jean-Pierre Flori]

Opened #15839 for IML.

[Jean-Pierre Flori]

GSL has been fixed upstream! ​http://git.savannah.gnu.org/cgit/gsl.git/commit/?id=1d002ee931b6a19333416ba8c4c4cccbc009b6a9

[Jean-Pierre Flori]

Opened #15872 for cddlib.

[Jean-Pierre Flori]

All packages we currently ship are ok upstream or configured or patched by us.