On Thu 2016-09-01 10:52:47 -0400, gregor herrmann wrote: > On Tue, 30 Aug 2016 08:05:14 -0400, Daniel Kahn Gillmor wrote: > Maybe we should think about this for the test suite temporarily. > Or maybe we can trick gpg-agent into working during build / in a > chroot somehow. And stopping it afterwards. Etc. ... > Probably the latter, since the tests work in general ... > > --allow-preset-passphrase for the agent and gpg-preset-passphrase(1) > look interesting.
An even better way to work around pinentry-related failures in test suites is to supply a phony pinentry and point to it. This ensures that the gpg codepaths in use in the test suite are the same ones that are used by the code in operation, with a change only to the pinentry. GnuPG itself does this with a minimal fake-pinentry.c for its own test suite: https://anonscm.debian.org/git/pkg-gnupg/gnupg2.git/tree/tests/openpgp/fake-pinentry.c I imagine for perl we'd want to use a fake-pinentry written in perl so that we don't introduce new system-level dependencies for the test suite. I'll take a crack at this for libmail-gnupg-perl and see if i can sort something out. --dkg
signature.asc
Description: PGP signature
