Spent my weekend rewriting GMime's test-suite to be nearly fully automated. I say "nearly" because it's not completely done yet.
I've currently got fully automated tests for streams (concat stream has its very own), iconv utility functions, address parser, rfc2047/rfc2184 header encoders/decoders, quoted-string utilities, mbox parser, gpg context, and the pgp/mime utility classes.
The stream (including concat) tests generate their own test-case data and most of the others have built-in test data, however the mbox parser and pgp tests require manually created test data (the pgp tests just need pre-generated keys).
I could probably include the pre-generated gpg pub/private keys in the tarball distribution, but the test data for the mbox parser is around 11 megs, so I'm not sure how to approach that - I suppose I could tarball it up separately and upload it somewhere and upload a new version whenever I need to change it (or add more test data?).
I was kind of surprised that the only MIME parser test-suite messages I could find out there on the internet was the bundle of messages I used to host on my personal Ximian website (many of which I think I got from jwz?), put together for Camel a number of years ago (which is what I've been using to regression test GMime as well - only, until now, it hadn't been fully automated).
Do authors of other MIME parsers out there simply not have test suites for their parsers?
On a related note, I've just released GMime 2.2.5 with a few fixes (including some fixes for bugs I found only via the new automated test suite).