samth
2019-12-1 18:55:06

Unfortunately it doesn’t look like GHA supports that kind of test output


pocmatos
2019-12-1 19:20:43

Thanks for the reference, I will take a look.


pocmatos
2019-12-1 19:21:10

@notjack unsure I understand the comment, you think we shouldn’t speed up the PR tests by not performing a full bootstrap?


pocmatos
2019-12-1 19:22:05

As @samth said, we could add a hook for a comment like full-ci if a reviewer wanted to force a full ci on the pr. (Actions allows that - I think…)


pocmatos
2019-12-1 19:22:41

We could at least test how much faster it gets. Maybe it doesn’t help that much in which case there’s no point to discuss it further.


samth
2019-12-1 19:23:08

it’s definitely worth testing


samth
2019-12-1 19:23:28

the best would be to automatically detect when we need it


pocmatos
2019-12-1 19:24:12

When would that be?


pocmatos
2019-12-1 19:25:22

I assume, if the PR only hits Racket files we don’t need a full bootstrap but might need with changing C code. Is there a more specific rule than this?


pocmatos
2019-12-1 19:25:38

Oh wait - actually not


samth
2019-12-1 19:25:39

right, that’s what I’m thinking


pocmatos
2019-12-1 19:25:50

xform is in racket, in this case we should bootstrap, right/


pocmatos
2019-12-1 19:25:57

?


samth
2019-12-1 19:26:08

although changes to the expander requires re-building (but that changes a C file)


samth
2019-12-1 19:26:33

xform is indeed in Racket (that’s the whole issue) and any changes to that would be like changes to C code


pocmatos
2019-12-1 19:27:21

ah - right. so we could say - a C file change requires a bootstrapping CI, otherwise no bootstrapping CI.


pocmatos
2019-12-1 19:27:38

More rules could be used like, if only a README changes, then no CI is needed?


samth
2019-12-1 19:28:25

yes


pocmatos
2019-12-1 19:29:55

I will setup the basic for now but keep these things in mind for later. I prefer small changes and see how it evolves than going all-in on CI optimizations. :wink:


pocmatos
2019-12-1 19:30:29

also, small changes are easier to manage when not working on something full-time.


samth
2019-12-1 19:31:45

indeed; you’ve already done lots!


samth
2019-12-1 19:32:21

let me know when everything that the Azure builds are doing is replicated on GHA. I’ll turn the Azure builds off then.


pocmatos
2019-12-1 19:32:53

Once Windows CI is in place, https://github.com/racket/racket/pull/2945 we can do that.


pocmatos
2019-12-1 19:33:15

Kids are now in bed. I will grab a pizza quickly and work on that before bed. Be right back.


pocmatos
2019-12-1 23:17:11

samth
2019-12-1 23:30:22

No idea. We’ve never had CI for racketcs on Windows so I’ve never worked with it


pocmatos
2019-12-1 23:38:30

Thanks. I will take another look at it tomorrow.