klapaucius/vector-hashtables

Adding Agda to CI?

Opened this issue · 5 comments

Agda is a big (the biggest?) consumer of our package. It'd be nice to make sure that many little changes we've been applying recently don't break anything particularly badly on their side, and at least give them a heads-up if it breaks.

OTOH Agda is a big project to add to the regular CI carelessly. In theory, we could run it by corn or on merge instead of on pull.

@andreasabel -- any suggestions/advice on this?

@ulysses4ever wrote:

run it by corn

Sorry, my vocabulary fails me here. What does this mean?

Ah, this is probably a typo, "cron". Ok.
Yeah, maybe a cron job would do it. What do you want to test? Just functionality or also performance?

Oh my, sorry: I meant cron, which means on a schedule. GitHub actions can do that afair.

We thought functionality for starters.

Maybe you could have agda and agda-stdlib as submodules pinned to released versions (v2.6.4.3 and v2.0 resp.) and build agda with your master and run it against its standard library.

Running the whole testsuite of Agda is maybe not what you want. This is how we do it: https://github.com/agda/agda/blob/403ee4263e0f14222956e398d2610ae1a4f05467/src/github/workflows/test.yml