Oddities with use of MD5 hashes instead of timestamps about biomake HOT 1 CLOSED

yeah, well spotted. When the file timestamp changes, biomake should recompute the hash (md5_hash) but not ignore the validity condition (md5_valid). The two are stored in the same file, which is possibly a poor design decision, but the rationale is to avoid cluttering the filesystem. See e.g. here for examples

For some reason the code had decided not to read that file when timestamps were stale, which kind of defeats the whole purpose of an MD5 checksum. What it should do is read the file (to load the validity condition) but then recompute the hash.

Really, the root problem was that there was no test! Crafting a test for this was a little devious since it requires calling biomake twice and subtly altering the recipe the second time (so test engine can tell if it's been re-run)

The MD5 tests are all a little devious/hacky, actually, but nothing compared to the lengths we'd have to go to test the various error reporting options ;) a job which i have so far skirted (so e.g. there is no test for #31)

from biomake.