Flaky test in test_pruner.py

[timokau]

While packaging the 0.5.0 update for nixos, I noticed that one of the tests fails some of the time:

=================================== FAILURES ===================================
_________________________________ test_pruner __________________________________

    def test_pruner():
    
        # A dummyPruningParams.run to load tabulated values
        Pruning.run(5, 50, 10*[1.], .5)
    
        for (n, overhead) in dim_oh:
    
            print(" \n ~~~~ Dim %d \n" % n)
    
            M = prepare(n)
            r = [M.get_r(i, i) for i in range(n)]
    
            print(" \n GREEDY")
            radius = gaussian_heuristic(r) * 1.6
            print("pre-greedy radius %.4e" % radius)
            tt = process_time()
>           pruning =Pruning.run(radius, overhead, r, 200, flags=Pruning.ZEALOUS, metric="solutions")

tests/test_pruner.py:37: 
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ 

>   sig_on()
E   RuntimeError: Aborted

src/fpylll/fplll/pruner.pyx:1000: RuntimeError
----------------------------- Captured stdout call -----------------------------
 
 ~~~~ Dim 40 

 
 GREEDY
pre-greedy radius 4.3353e+06
Time 6.0940e-02
post-greedy radius 4.3353e+06
PruningParams<1.600000, (1.00,...,0.36), 203.2640>
cost 8.4996e+04
209
 
 GRADIENT 

radius 4.3353e+06
Time 1.3805e-02
PruningParams<1.600000, (1.00,...,0.36), 203.2640>
cost 8.4996e+04
209
 
 HYBRID 

radius 4.3353e+06
Time 6.1017e-02
PruningParams<1.600000, (1.00,...,0.36), 203.2640>
cost 8.4996e+04
209
 
 ~~~~ Dim 41 

 
 GREEDY
pre-greedy radius 2.9518e+06
----------------------------- Captured stderr call -----------------------------
terminate called after throwing an instance of 'std::runtime_error'
  what():  Concavity says that should not happen.
======================== 1 failed, 64 passed in 37.31s =========================

[malb]

What system/OS is this on? What fraction of tests fails? I can't reproduce it on Debian/64-bit.

[timokau]

Its on nixos (64 bit) and only the one test that I reported fails.

[malb]

I can reproduce it now. I wonder if it is related to fplll/fplll#398

[timokau]

Sounds likely. We can re-test once that is merged.

[malb]

I just ran the tests 200 times and this does indeed seem to fix it.

[timokau]

I can confirm that the PR fixes the issue for me as well. The patch looks harmless, so I'll backport it to fpylll-0.5 package on nixos (NixOS/nixpkgs#74372). Thank you!