Generating keys in tests is very slow

[wearpants]

Generating a keypair in tests is extremely slow. This is related to #18 & #23.

Ideally, we reduce # of tests that generate keys in the first place. At a quick glance, the only test that seems to require generating a new key is GnuPGTest.test_gen_key.

Key generation is slow b/c the kernel lacks entropy (this is often a problem on servers & VMs without a lot of I/O). Possible approaches:

• run find / in a background process to thrash the disk. Ick.
• reduce keypair size (need more research)
• configure things to use an alternate source of entropy (hardware RNG, whatever)

Alternately, the slow tests could be moved to a separate file, so that they can be run selectively.

[mfb]

would running haveged in the test environment be enough of a solution?

[wearpants]

Cool, that did the trick :-). I'll make a note in the docs

http://www.issihosts.com/haveged/

[garrettr]

Additionally, you could consider using a smaller key size (say RSA-1024) for tests only. These are much faster to generate than 2048 or 4096 bit keys. We do this for the Securedrop unit tests.