The build system is currently built around local runs, which means it's okay if it falls over from time to time. We don't need to be quite as explicit as the old liveblog (especially since we kept tuning down the warnings that freak people out), but we should have some degree of fault tolerance built in. Questions we should answer:
Should we turn on --force for Grunt, causing it to continue running if tasks fail?
If the doc is broken, does it make sense to build from the last good copy or should we just abort, since the effect is the same?
What should be rebuilt on the timer? Just HTML, or should we also be building styles/scripts even though they don't change much?
I feel good about this--errors in parsing do not over the liveblog, and errors in templating halt the process before publishing. The systemd unit then restarts five seconds after the crash.
The build system is currently built around local runs, which means it's okay if it falls over from time to time. We don't need to be quite as explicit as the old liveblog (especially since we kept tuning down the warnings that freak people out), but we should have some degree of fault tolerance built in. Questions we should answer:
--force
for Grunt, causing it to continue running if tasks fail?