Copy libpari.dll.a on Cygwin

[jpflori]

On Cygwin, PARI produces not only a shared library dll file but also an import dll.a file.
Nonetheless, the dll.a is not copied by the install targets of the Makefile.

And having it would ensure that other shared libraries linked with -lpari would link to libpari.dll rather than libpari.a (and some as eclib won't work with libpari.a anyway, see #13325), because ld looks in order for dll.a, then a, then dll.

This has been included upstream, see git rev 2c34d58be7f844581733e7c2b96f2a60ee3eb82a and ffd02415dde8a48f7c06765635b0457b23ba60e8

Use the spkg at:
http://www.infres.enst.fr/~flori/sage/pari-2.5.2.p2.spkg

Depends on #13534

Upstream: Fixed upstream, but not in a stable release.

CC: @dimpase

Component: porting: Cygwin

Keywords: pari spkg cygwin

Reviewer: Dmitrii Pasechnik, Jean-Pierre Flori

Issue created by migration from https://trac.sagemath.org/ticket/13333

[jpflori]

comment:1

Tentative spkg available at
http://perso.telecom-paritech.fr/~flori/sage/pari-2.5.1.p4.spkg

[jpflori]

comment:2

I'll post an updated spkg based on #13320 when this gets positive review.

[dimpase]

comment:3

Replying to @jpflori:

Tentative spkg available at
http://perso.telecom-paritech.fr/~flori/sage/pari-2.5.1.p4.spkg

should be http://perso.telecom-paristech.fr/~flori/sage/pari-2.5.1.p4.spkg

[jpflori]

comment:4

Updated spkg on top of pari-2.5.2.p0 from #13320:
http://perso.telecom-paristech.fr/~flori/sage/pari-2.5.2.p1.spkg

[jpflori]

Author: Jean-Pierre Flori

[jpflori]

comment:5

No return from upstream yet (which means some weeks now).
If someone thinks the changes made here are logical, let's use the spkg at
http://perso.telecom-paristech.fr/~flori/sage/pari-2.5.2.p1.spkg
in Sage.
At least they make sure that the shared version of Pari is found before the static one when linking (and so surely fix broken pieces of Sage) by only copying a file produced by Pari itself (the only drawback might be that linking should be a little slower, but Cygwin is already slow as hell and I did not notice any difference).

[jpflori]

Work Issues: Rebase on #13314

[jpflori]

comment:6

This should be rebased on top of #13314.

[jpflori]

Changed work issues from Rebase on #13314 to Add PARI fix to #13314

[jpflori]

Changed upstream from Reported upstream. No feedback yet. to Fixed upstream, but not in a stable release.

[jpflori]

comment:7

such a change has been integrated upstream:
http://pari.math.u-bordeaux.fr/archives/pari-dev-1209/msg00041.html
I'll package a version of #13314 inlcuding the corresponding commit (GIT rev 2c34d58be7f8).

[jpflori]

Changed work issues from Add PARI fix to #13314 to none

[jpflori]

comment:8

New spkg including (slightly modified, because some other commits changed the patching context in between) upstream patches at:
http://perso.telecom-paristech.fr/~flori/sage/pari-2.5.2.p2.spkg

[jpflori]

spkg diff, for review only

[jpflori]

comment:10

Attachment: pari-2.5.2.p2.diff.gz

[dimpase]

comment:11

looks good.

[jdemeyer]

Reviewer: Dmitrii Pasechnik

[jdemeyer]

comment:15

In the PARI spkg, patches are supposed to be documented in patches/README.txt. Also, I prefer a more descriptive name for the patch file than "Makefile.SH.patch".

I made those fixes in #13534. You could either make those changes here, or wait for #13534 (then this gets closed as duplicate).

[jpflori]

comment:16

I'd say let's get this merged with #13534, that's perfectly fine for me.

[jdemeyer]

Dependencies: #13534

[jdemeyer]

Changed author from Jean-Pierre Flori to none

[jdemeyer]

Changed reviewer from Dmitrii Pasechnik to Dmitrii Pasechnik, Jean-Pierre Flori