let pari's spkg-configure accept pari 2.13

[dimpase]

some formats have changed, so tests fail, see
#30801 comment:6

Depends on #30801

CC: @tobihan @thierry-FreeBSD @orlitzky @mkoeppe @antonio-rojas @videlec @dkwo

Component: build: configure

Branch/Commit: u/dimpase/build/pari213fix @ ad702bf

Reviewer: Dima Pasechnik

Issue created by migration from https://trac.sagemath.org/ticket/30906

[dimpase]

New commits:

ad702bf

do the bnfisunit check in Pari, not rely on output

[dimpase]

Commit: ad702bf

[dimpase]

Author: Dima Pasechnik

[dimpase]

Branch: u/dimpase/build/pari213fix

[dimpase]

comment:4

ping

[dimpase]

comment:5

on Gentoo this blocks using system pari

[mkoeppe]

comment:7

The pari upgrade to the same version in #30801 seems to require nontrivial changes to Sage code and tests and also involves an upgrade of cypari2. How come we should just be able to be accepting system pari 2.13?

[dimpase]

comment:9

made it conditional on pari update to 2.13

[dimpase]

Dependencies: #30801

[dkwo]

comment:10

On my Void linux machine, what works is

$ echo "bnf = bnfinit(y^4-y-1); bnfisunit(bnf,-y^3+2*y^2-1)==[0,2,0]~" | gp -qf
1

with only one square parenthesis, not two.

[mkoeppe]

comment:12

the two brackets are m4 syntax

[dkwo]

comment:13

I see, thanks.
However, even if I edit the file as per the commit, the bnfisunit test still fails for me with pari 2.13.0 using ./config, even though from a shell I get as above.

What am I missing?

[mkoeppe]

comment:14

Did you run ./bootstrap?

[dkwo]

comment:15

That fixes it, thanks.

[mkoeppe]

comment:17

Looks like #30801 is taking care of this

[mkoeppe]

Changed author from Dima Pasechnik to none

[dimpase]

comment:18

yep. looks like the fix here was copied there.

[dimpase]

Reviewer: Dima Pasechnik