Move expensive test to xt/, add malloc-canary preventing false-negatives
authorPeter Rabbitson <ribasushi@cpan.org>
Thu, 3 Mar 2016 14:27:34 +0000 (15:27 +0100)
committerPeter Rabbitson <ribasushi@cpan.org>
Thu, 10 Mar 2016 11:09:05 +0000 (12:09 +0100)
commit24fbd7fba01adfc25b53614f8d713af1bd31ae21
tree510ddf020c8a328b72e0d65bb75f028be5de2000
parentbf3802a96bc090d96e7aa4f0f60d2731d3dbd003
Move expensive test to xt/, add malloc-canary preventing false-negatives

This test was originally written to validate both Devel::GlobalDestruction::PP
and the M.A.D. cyclic reference handler (a4367b26). These days it makes little
sense to run on end-user installs, yet this bizarre test still uncovers weird
problems in the underlying Rube Goldberg machine.

So instead of outright deleting it - move it to xt/ and validate its execution
environment with what is essentially a guarded calloc()

Add a tight-memory travis config to make sure that OOM won't kill the wrong
thing

Read diff under -C
.travis.yml
t/lib/DBICTest/Util.pm
xt/extra/internals/ithread_stress.t [moved from t/51threadnodb.t with 59% similarity]