On Mon 2017-01-23 19:21:30 -0500, Santiago Vila wrote: >> > I wonder, however, why the tests need to generate a key at all, >> > just to discard it after the package is built. Would not be possible >> > to use a pregenerated key for the tests? >> >> gpgme exercises functionality of the GnuPG suite, including key >> generation. If we were to use a pregenerated key, we might as well just >> skip the test suite :) > > Well, it could be argued that the key creation part should be already > covered by the test suite of gnupg itself.
hm, that wouldn't test whether gpgme is actually doing the right thing to drive gpg, though. I think we do still need that test in gpgme, just like we'd need a higher-level test for a mail user agent that was supposed to generate keys, or to do anything else that involves randomness. Build processes and testing should in general avoid the use of randomness, but a test suite that tries to exercise code that should use randomness in the real world (even at other lower levels) should still be allowed access to that randomness during the compile-time testing. no? --dkg