Monkey patch two dependencies to use /dev/unrandom over /dev/random

[kaoru]

Proof of concept... I don't recommend or expect you to actually merge this! :-)

---

This makes the tests 11-encrypt.t and 13-keygen.t run consistently
quickly, whereas before they were intermittently much slower.

Before:

$ for i in $(seq 1 10) ; do prove -b t/11-encrypt.t 2>&1 | grep wallclock ; done
Files=1, Tests=36, 0 wallclock secs ( 0.03 usr 0.00 sys + 0.67 cusr 0.01 csys = 0.71 CPU)
Files=1, Tests=36, 1 wallclock secs ( 0.04 usr 0.00 sys + 0.64 cusr 0.02 csys = 0.70 CPU)
Files=1, Tests=36, 80 wallclock secs ( 0.03 usr 0.00 sys + 0.66 cusr 0.01 csys = 0.70 CPU)
Files=1, Tests=36, 87 wallclock secs ( 0.04 usr 0.00 sys + 0.55 cusr 0.00 csys = 0.59 CPU)
Files=1, Tests=36, 67 wallclock secs ( 0.03 usr 0.00 sys + 0.58 cusr 0.00 csys = 0.61 CPU)
Files=1, Tests=36, 89 wallclock secs ( 0.03 usr 0.01 sys + 0.49 cusr 0.02 csys = 0.55 CPU)
Files=1, Tests=36, 83 wallclock secs ( 0.03 usr 0.01 sys + 0.65 cusr 0.02 csys = 0.71 CPU)
Files=1, Tests=36, 80 wallclock secs ( 0.04 usr 0.01 sys + 0.60 cusr 0.01 csys = 0.66 CPU)
Files=1, Tests=36, 34 wallclock secs ( 0.04 usr 0.00 sys + 0.62 cusr 0.03 csys = 0.69 CPU)
Files=1, Tests=36, 32 wallclock secs ( 0.02 usr 0.00 sys + 0.54 cusr 0.01 csys = 0.57 CPU)

After:

$ for i in $(seq 1 10) ; do prove -b t/11-encrypt.t 2>&1 | grep wallclock ; done
Files=1, Tests=36, 1 wallclock secs ( 0.02 usr 0.00 sys + 0.43 cusr 0.01 csys = 0.46 CPU)
Files=1, Tests=36, 0 wallclock secs ( 0.02 usr 0.00 sys + 0.41 cusr 0.00 csys = 0.43 CPU)
Files=1, Tests=36, 1 wallclock secs ( 0.02 usr 0.00 sys + 0.40 cusr 0.01 csys = 0.43 CPU)
Files=1, Tests=36, 0 wallclock secs ( 0.03 usr 0.00 sys + 0.40 cusr 0.00 csys = 0.43 CPU)
Files=1, Tests=36, 1 wallclock secs ( 0.03 usr 0.00 sys + 0.42 cusr 0.02 csys = 0.47 CPU)
Files=1, Tests=36, 0 wallclock secs ( 0.03 usr 0.00 sys + 0.40 cusr 0.01 csys = 0.44 CPU)
Files=1, Tests=36, 1 wallclock secs ( 0.03 usr 0.00 sys + 0.40 cusr 0.00 csys = 0.43 CPU)
Files=1, Tests=36, 0 wallclock secs ( 0.02 usr 0.01 sys + 0.39 cusr 0.01 csys = 0.43 CPU)
Files=1, Tests=36, 1 wallclock secs ( 0.02 usr 0.01 sys + 0.41 cusr 0.01 csys = 0.45 CPU)
Files=1, Tests=36, 0 wallclock secs ( 0.03 usr 0.00 sys + 0.41 cusr 0.00 csys = 0.44 CPU)

[rfdrake]

I'm not a developer of the project, but if you're mainly concerned with test cases then I would monkey patch there instead of in the modules themselves. Alternatively, you could write a module that patched when it was loaded to override the other modules, then run something like:

prove -lv -MMonkey::URandom t/test.t

The real problem is that these kinds of patches are unsafe and hard to maintain through upstream changes. It would be nice if Crypt::Random::Util would let you set the order, or if the other module would use a shared "our" variable as the source for the random file. Then you could just patch that line instead of rewriting the whole function.