GitLab

Housekeeping fixes

See merge request !31

I suppose that the MPI prototypes have become const-correct as a result of one of the latest standards updates.

simplify ci structure, get rid of intermediary containers.

See merge request !30

It seems too hard to make sure that we don't run out of disk on runners,
and this saves very little in the end.

have to push to master, otherwise this won't get any exposure of course.

The "notiming" threshold wasn't correctly used.

Add -t command line option that disables correctness checks and loops arithmetic 1024 times

See merge request !23

G++10 fixes

See merge request !28

Bucket sieve projective primes; fix attempt for #30012

See merge request !27

when the prime is fixed and the test driver script auto-selects the
lingen binary, then we must have the test depend on the binary that gets
selected, of course.

These "If proj and r > 0, then r == 1/p (mod p^2)" comments are just
bogus.

 - we're not bucket-sieving powers at the moment, so proj implies r==0
   anyway.
 - if we have proj and r>0, i.e. if we're considering the factor base
   entry p=4 r=6, which encodes (1:2) in P1(Z/4), then we get r=2, and
   this encodes the p-lattice with basis vectors (2,0) and (1,2). Which
   by no means triggers a non-trivial gcd, and anyway it's by
   construction. No element that we're ever going to consider on the
   projective line will encode a lattice with a non-trivial gcd of the
   coordinates.

Ci more features (late fixes)

See merge request !25