2.1.0beta3: long hang in genkey1024.test (work around with --debug-quick-random doesn't seem to speed things up?)

[Daniel Kahn Gillmor]

Hi GnuPG folks--

I'm still trying to build gnupg 2.1.0beta3 on a headless i386 debian 
unstable system.  the tests seem to take a really long time; in 
particular, tests/openpgp/genkey1024.test takes ages.  I suspect it's 
trying to pull random data from the kernel, and the headless machine 
just doesn't have the entropy to supply it.

This could potentially be a problem on debian's buildd's which may 
themselves be low on entropy.

I tried patching genkey1024.test to add --debug-quick-random as an 
argument to gpg, but that didn't speed up the test either.

observing the test in action, it appears that gpg is blocked reading 
from file descriptor 3, which is a socket.  (presumably to gpg-agent?  i 
can't find the gpg-agent instance on the build system, but perhaps it's 
just getting late and i'm missing obvious things).

As an interim workaround, i've installed haveged on my build machine to 
keep the kernel's entropy pool well-populated, but i don't think that's 
a viable plan for everyone.

What would be the right way to ask the key generation tests to use 
"quick-random" sources?  I don't think we need real-world entropy for 
these tests, do we?

	--dkg